Daniel Preoțiuc-Pietro

Go to Year: 2024 | 2023 | 2022 | 2021 | 2020 | 2019 | 2018 | 2017 | 2016 | 2015 | 2014 | 2013 | 2012

2024

• Modeling and Detecting Company Risks from News: A Case Study in Bloomberg News

  Jiaxin Pei, Soumya Vadlamannati, Liang-Kang Huang, Daniel Preoțiuc-Pietro, Xinyu Hua

  NAACL Industry Track

  [Abstract]

  Identifying risks associated with a company is important to investors and the well-being of the overall financial market. In this study, we build a computational framework to automatically extract company risk factors from news articles. Our newly proposed schema comprises seven distinct aspects, such as supply chain, regulations, and competitions. We sample and annotate 744 news articles and benchmark various machine learning models. While large language models have achieved huge progress in various types of NLP tasks, our experiment shows that zero-shot and few-shot prompting state-of-the-art LLMs (e.g. LLaMA-2) can only achieve moderate to low performances in identifying risk factors. And fine-tuned pre-trained language models are performing better on most of the risk factors. Using this model, we analyze over 277K Bloomberg news articles and demonstrate that identifying risk factors from news could provide extensive insight into the operations of companies and industries.
• Leveraging Contextual Information for Effective Entity Salience Detection

  Rajarshi Bhowmik, Marco Ponza, Atharva Tendle, Anant Gupta, Rebecca Jiang, Xingyu Lu, Qian Zhao, Daniel Preoțiuc-Pietro

  NAACL Findings

  [Abstract]

  In text documents such as news articles, the content and key events usually revolve around a subset of all the entities mentioned in a document. These entities, often deemed as salient entities, provide useful cues of the aboutness of a document to a reader. Identifying the salience of entities was found helpful in several downstream applications such as search, ranking, and entity-centric summarization, among others. Prior work on salient entity detection mainly focused on machine learning models that require heavy feature engineering. We show that fine-tuning medium-sized language models with a cross-encoder style architecture yields substantial performance gains over feature engineering approaches. To this end, we conduct a comprehensive benchmarking of four publicly available datasets using models representative of the medium-sized pre-trained language model family. Additionally, we show that zero-shot prompting of instruction-tuned language models yields inferior results, indicating the task's uniqueness and complexity.
• Who is bragging more online? A large scale analysis of bragging in social media

  Mali Jin, Seza A. Dogruoz, Daniel Preoţiuc-Pietro, Nikolaos Aletras

  LREC-COLING

  [Abstract]

  Bragging is the act of uttering statements that are likely to be positively viewed by others and it is extensively employed in human communication with the aim to build a positive self-image of one self. Social media is a natural platform for users to employ bragging in order to gain admiration, respect, attention and followers from their audiences. Yet, little is known about the scale of bragging online and its characteristics. This paper employs computational linguistics to conduct the first large scale study of bragging behavior on Twitter in the U.S. by focusing on its overall prevalence, temporal dynamics and impact of demographic factors. Our study shows that the prevalence of bragging has been decreasing over time within the same population of users and that younger, more educated and popular users in the U.S. are more likely to brag. Finally, we conduct an extensive linguistics analysis to unveil specific bragging themes associated with different user traits. The study aims to offer an initial understanding of bragging behavior online in computational linguistics and to pave the way for further research.
• Unsupervised Contrast-Consistent Ranking with Language Models

  Niklas Stoehr, Pengxiang Cheng, Jing Wang, Daniel Preoțiuc-Pietro, Rajarshi Bhowmik

  EACL

  [Abstract]

  Language models contain ranking-based knowledge and are powerful solvers of in-context ranking tasks. For instance, they may have parametric knowledge about the ordering of countries by size or may be able to rank reviews by sentiment. Recent work focuses on pairwise, pointwise, and listwise prompting techniques to elicit a language model's ranking knowledge. However, we find that even with careful calibration and constrained decoding, prompting-based techniques may not always be self-consistent in the rankings they produce. This motivates us to explore an alternative approach that is inspired by an unsupervised probing method called Contrast-Consistent Search (CCS). The idea is to train a probing model guided by a logical constraint: a model's representation of a statement and its negation must be mapped to contrastive true-false poles consistently across multiple statements. We hypothesize that similar constraints apply to ranking tasks where all items are related via consistent pairwise or listwise comparisons. To this end, we extend the binary CCS method to Contrast-Consistent Ranking (CCR) by adapting existing ranking methods such as the Max-Margin Loss, Triplet Loss, and Ordinal Regression objective. Our results confirm that, for the same language model, CCR probing outperforms prompting and even performs on a par with prompting much larger language models.
• Improving Multimodal Classification of Social Media Posts by Leveraging Image-Text Auxiliary Tasks

  Danae Sánchez Villegas, Daniel Preoțiuc-Pietro, Nikolaos Aletras

  EACL Findings

  [Abstract]

  Effectively leveraging multimodal information from social media posts is essential to various downstream tasks such as sentiment analysis, sarcasm detection and hate speech classification. However, combining text and image information is challenging because of the idiosyncratic cross-modal semantics with hidden or complementary information present in matching image-text pairs. In this work, we aim to directly model this by proposing the use of two auxiliary losses jointly with the main task when fine-tuning any pre-trained multimodal model. Image-Text Contrastive (ITC) brings imagetext representations of a post closer together and separates them from different posts, capturing underlying dependencies. Image-Text Matching (ITM) facilitates the understanding of semantic correspondence between images and text by penalizing unrelated pairs. We combine these objectives with five multimodal models, demonstrating consistent improvements across four popular social media datasets. Furthermore, through detailed analysis, we shed light on the specific scenarios and cases where each auxiliary task proves to be most effective.

2023

• EntSUMv2: Dataset, Models and Evaluation for More Abstractive Entity-Centric Summarization

  Dhruv Mehra, Lingjue Xie, Ella Hofmann-Coyle, Mayank Kulkarni, Daniel Preoțiuc-Pietro

  EMNLP

  [Abstract]

  Entity-centric summarization is a form of controllable summarization that aims to generate a summary for a specific entity given a document. Concise summaries are valuable in various real-life applications as they enable users to quickly grasp the main points of the document focusing on an entity of interest. This paper presents EntSUMv2, a more abstractive version of the original entity-centric EntSUM summarization data set. In EntSUMv2 the annotated summaries are intentionally made shorter to benefit more specific and useful entity-centric summaries for downstream users. We conduct extensive experiments on this data set using multiple abstractive summarization approaches that employ supervised fine-tuning or large-scale instruction tuning. Additionally, we perform comprehensive human evaluation that incorporates metrics for measuring crucial facets. These metrics provide a more fine-grained interpretation of the current state-of-the-art systems and highlight areas for future improvement.
• Filling in the White Space: Spatial Interpolation with Gaussian Processes and Social Media Data

  Salvatore Giorgi, Johannes C. Eichstaedt, Daniel Preoțiuc-Pietro, Jacob R. Gardner, H. Andrew Schwartz, Lyle H. Ungar

  Current Research in Ecological and Social Psychology

  [Abstract]

  Full national coverage below the state level is difficult to attain through survey-based data collection. Even the largest survey-based data collections, such as the CDC's Behavioral Risk Factor Surveillance System or the Gallup-Healthways Well-being Index (both with more than 300,000 responses p.a.) only allow for the estimation of annual averages for about 260 out of roughly U.S. 3,000 counties when a threshold of 300 responses per county is used. Using a relatively high threshold of 300 responses gives substantially higher convergent validity--higher correlations with health variables--than lower thresholds but covers a reduced and biased sample of the population. We present principled methods to interpolate spatial estimates and show that including large-scale geotagged social media data can increase interpolation accuracy. In this work, we focus on Gallup-reported life satisfaction, a widely-used measure of subjective well-being. We use Gaussian Processes (GP), a formal Bayesian model, to interpolate life satisfaction, which we optimally combine with estimates from low-count data. We interpolate over several spaces (geographic and socioeconomic) and extend these evaluations to the space created by variables encoding language frequencies of approximately 6 million geotagged Twitter users. We find that Twitter language use can serve as a rough aggregate measure of socioeconomic and cultural similarity, and improves upon estimates derived from a wide variety of socioeconomic, demographic, and geographic similarity measures. We show that applying Gaussian Processes to the limited Gallup data allows us to generate estimates for a much larger number of counties while maintaining the same level of convergent validity with external criteria (i.e., N = 1,133 vs. 2,966 counties). This work suggests that spatial coverage of psychological variables can be reliably extended through Bayesian techniques while maintaining out-of-sample prediction accuracy and that Twitter language adds important information about cultural similarity over and above traditional socio-demographic and geographic similarity measures. Finally, to facilitate the adoption of these methods, we have also open-sourced an online tool that researchers can freely use to interpolate their data across geographies.
• Analyzing and Predicting Persistence of News Tweets

  Maggie Liu, Jing Wang, Daniel Preoțiuc-Pietro

  AACL

  [Abstract]

  News is read and consumed differently based on its topic and timeliness to the reader. Some stories attract readers immediately after they are published, while others capture readership consistently over multiple days after their publication, regardless of their overall popularity. This paper studies this less explored facet of news story consumption, which we name persistence, operationalized as the time for a story to reach a certain percent of its total interest. In particular, we study persistence though a novel, publicly available data set of news tweets from 353 news outlets. We perform an extensive linguistic analysis of news persistence in social media to uncover the underlying topical and stylistic cues that impact short- or long-term interest in a story. We train several models for predicting news persistence that achieve predictive performance of up to 0.353 Spearman correlation when extrapolating to tweets from days unseen in training and retain significant predictive performance even on tweets from accounts unseen in training. The ability to predict news persistence can be useful in several practical applications that drive news and social media consumption including alerting, search ranking or recommendations.
• Efficient Zero-Shot Cross-lingual Inference via Retrieval

  Genta Winata, Lingjue Xie, Karthik Radhakrishnan, Yifan Gao, Daniel Preoțiuc-Pietro

  AACL

  [Abstract]

  Resources for building NLP applications, such as data and models, are usually only created and curated for a limited set of high resource languages. Thus, the ability to transfer knowledge to a new language is a key way in which to enable access to NLP technology for a wider population. This paper presents a framework to perform zero-shot inference in a target language by using cross-lingual retrieval from another language where limited annotated data for a comparable domain is available. Results on two large-scale multilingual data sets show that, in this setup, this framework improves over fine-tuning multilingual models or translating annotated data and achieves results relatively close to fine-tuning the model on the target language directly. These results show that models can be transferred efficiently across languages for a given task and domain, even for languages not covered by multilingual model training approaches.
• Overcoming Catastrophic Forgetting in Massively Multilingual Continual Learning

  Genta Indra Winata, Lingjue Xie, Karthik Radhakrishnan, Shijie Wu, Xisen Jin, Pengxiang Cheng, Mayank Kulkarni, Daniel Preoțiuc-Pietro

  ACL Findings

  [Abstract]

  Real-life multilingual systems should be able to efficiently incorporate new languages as data distributions fed to the system evolve over time. To do this, systems need to handle the issue of catastrophic forgetting, where the model performance drops for languages or tasks seen further in its past. In this paper, we study catastrophic forgetting, as well as methods to minimize this, in a massively multilingual continual learning framework involving up to 51 languages and covering both classification and sequence labeling tasks. We present LR ADJUST, a learning rate scheduling method that is simple yet effective in preserving new information without strongly overwriting past knowledge. Furthermore, we show that this method is effective across multiple continual learning approaches. Finally, we provide further insights into the dynamics of catastrophic forgetting in this massively multilingual setup.
• Towards a Unified Multi-Domain Multilingual Named Entity Recognition Model

  Mayank Kulkarni, Daniel Preoțiuc-Pietro, Karthik Radhakrishnan, Genta Indra Winata, Shijie Wu, Lingjue Xie, Shaohua Yang

  EACL

  [Abstract]

  Named Entity Recognition is a key Natural Language Processing task whose performance is sensitive to choice of genre and language. A unified NER model across multiple genres and languages is more practical and efficient by leveraging commonalities across genres or languages. In this paper, we propose a novel setup for NER which includes multi-domain and multilingual training and evaluation across 13 domains and 4 languages. We explore a range of approaches to building a unified model using domain and language adaptation techniques. Our experiments highlight multiple nuances to consider while building a unified model, including that naive data pooling fails to obtain good performance, that domain-specific adaptations are more important than language-specific ones and that including domain-specific adaptations in a unified model nears the performance of training multiple dedicated monolingual models at a fraction of their parameter count.
• Dataless Knowledge Fusion by Merging Weights of Language Models

  Xisen Jin, Xiang Ren, Daniel Preoţiuc-Pietro, Pengxiang Cheng

  ICLR

  [Abstract] [Code]

  Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a dataless knowledge fusion method that merges models in their parameter space, guided by weights that minimize prediction differences between the merged model and the individual models. Over a battery of evaluation settings, we show that the proposed method significantly outperforms baselines such as Fisher-weighted averaging or model ensembling. Further, we find that our method is a promising alternative to multi-task learning that can preserve or sometimes improve over the individual models without access to the training data. Finally, model merging is more efficient than training a multi-task model, thus making it applicable to a wider set of scenarios.

2022

• Extractive Entity-Centric Summarization as Sentence Selection using Bi-Encoders

  Ella Hofmann-Coyle, Mayank Kulkarni, Lingjue Xie, Mounica Maddela, Daniel Preoţiuc-Pietro

  AACL

  [Abstract]

  Entity-centric summarization is a type of controllable summarization that produces a summary of a document that is specific to a given target entity. Extractive summarization approaches posses multiple advantages over abstractive methods such as preserving factuality and can be directly used in downstream tasks like target-based sentiment analysis or incorporated into search applications. In this paper, we explore methods to solve this task by recasting it as a sentence selection task, as supported by the EntSUM data set. We use methods inspired from information retrieval, where the input to the model is a pair representing a sentence from the original document and the target entity, in place of the query. We explore different architecture variants and loss functions in this framework with results showing an up to 5.8 F1 improvement over past state-of-the-art and outperforming the competitive entity-centric Lead 3 heuristic by 1.1 F1. In addition, we also demonstrate similarly strong results on the related task of salient sentence selection for an entity.
• Cross-lingual Few-Shot Learning on Unseen Languages

  Genta Winata, Shijie Wu, Mayank Kulkarni, Thamar Solorio, Daniel Preoţiuc-Pietro

  AACL

  [Abstract]

  Large pre-trained language models (LMs) have demonstrated the ability to obtain good performance on downstream tasks with limited examples in cross-lingual settings. However, this was mostly studied for relatively resource-rich languages, where at least enough unlabeled data is available to be included in pre-training a multilingual language model. In this paper, we explore the problem of cross-lingual transfer in low-resource languages, where no unlabeled data is available for the pre-trained model. We use a downstream sentiment analysis task across 12 languages to analyze the effectiveness of several few-shot learning strategies across the three major types of model architectures and their learning dynamics. We also compare strategies for selecting languages for transfer and contrast findings across languages seen in pre-training compared to those that are not. Our results contribute to the body of knowledge on cross-lingual models for low-resource settings which is paramount to increasing coverage, diversity, and equity in access to NLP technology.
• Combining Humor and Sarcasm for Improving Political Parody Detection

  Xiao Ao, Danae Sanchez Villegas, Daniel Preoţiuc-Pietro, Nikolaos Aletras

  NAACL

  [Abstract]

  Parody is a figurative device used for mimicking entities for comedic or critical purposes. Parody is intentionally humorous and often involves sarcasm. This paper explores jointly modelling these figurative tropes with the goal of improving performance of political parody detection in tweets. To this end, we present a multi-encoder model that combines three parallel encoders to enrich parody-specific representations with humor and sarcasm information. Experiments on a publicly available data set of political parody tweets demonstrate that our approach outperforms previous state-of-the-art methods.
• Automatic Identification and Classification of Bragging in Social Media

  Mali Jin, Seza A. Dogruoz, Daniel Preoţiuc-Pietro, Nikolaos Aletras

  ACL

  [Abstract]

  Bragging is a speech act employed with the goal of constructing a favorable self-image through positive statements about oneself. It is widespread in daily communication and especially popular in social media, where users aim to build a positive image of their persona directly or indirectly. In this paper, we present the first large scale study of bragging in computational linguistics, building on previous research in linguistics and pragmatics. To facilitate this, we introduce a new publicly available data set of tweets annotated for bragging and their types. We empirically evaluate different transformer based models injected with linguistic information in (a) binary bragging classification, i.e., if tweets contain bragging statements or not; and (b) multi-class bragging type prediction including not bragging. Our results show that our models can predict bragging with macro F1 up to 72.42 and 35.95 in the binary and multi-class classification tasks respectively. Finally, we present an extensive linguistic and error analysis of bragging prediction to guide future research on this topic.
• EntSUM: A Data Set for Entity-Centric Summarization

  Mounica Madella. Mayank Kulkarni, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Data]

  Controllable summarization aims to provide summaries that take into account user-specified aspects and preferences to better assist them with their information need, as opposed to the standard summarization setup which build a single generic summary of a document. We introduce a human-annotated data set (EntSUM) for controllable summarization with a focus on named entities as the aspects to control. We conduct an extensive quantitative analysis to motivate the task of entity-centric summarization and show that existing methods for controllable summarization fail to generate entity-centric summaries. We propose extensions to state-of-the-art summarization approaches that achieve substantially better results on our data set. Our analysis and results show the challenging nature of this task and of the proposed data set.

2021

• Identifying Named Entities as they are Typed

  Ravneet Arora, Chen-Tse Tsai, Daniel Preoţiuc-Pietro

  EACL

  [Abstract] [Slides]

  Identifying named entities in written text is an essential component of the text processing pipeline used in applications such as text editors to gain a better understanding of the semantics of the text. However, the typical experimental setup for evaluating Named Entity Recognition (NER) systems is not directly applicable to systems that process text in real time as the text is being typed. Evaluation is performed on a sentence level assuming the end-user is willing to wait until the entire sentence is typed for entities to be identified and further linked to identifiers or co-referenced. We introduce a novel experimental setup for NER systems for applications where decisions about named entity boundaries need to be performed in an online fashion. We study how state-of-the-art methods perform under this setup in multiple languages and propose adaptations to these models to suit this new experimental setup. Experimental results show that the best systems that are evaluated on each token after its typed, reach performance within 1-5 F1 points of systems that are evaluated at the end of the sentence. These show that entity recognition can be performed in this setup and open up the development of other NLP tools in a similar setup.

2020

• Fact vs. Opinion: the Role of Argumentative Features in News Classification

  Tariq Alhindi, Smaranda Muresan, Daniel Preoţiuc-Pietro

  COLING

  [Abstract] [Slides]

  Editorial news articles aim to persuade readers of an opinion and, in an era of widespread digital misinformation, it is paramount for readers to be able to easily distinguish these from news stories reporting factual events. Argumentative discourse is a key distinctive feature of editorials and likely more resilient across different topics or publishers. We study classifying articles into news stories and opinion using models that aim to supplement the article content representation with argumentative features. We show that argumentative features outperform linguistic features used previously and improve on fine-tuned transformer-based models when tested on data from publishers unseen in training. Automatically flagging argumentative news aids applications such as fact-checking of fact-based articles and separate them form opinion-based ones that have the writer's personal point of view.
• Point-of-Interest Type Inference from Social Media Text

  Danae Sanchez Villegas, Daniel Preoţiuc-Pietro, Nikolaos Aletras

  AACL-IJCNLP

  [Abstract] [Data] [Slides]

  Physical places help shape how we perceive the experiences we have there. For the first time, we study the relationship between social media text and the type of the place from where it was posted, whether a park, restaurant, or someplace else. To facilitate this, we introduce a novel data set of ~200,000 English tweets published from 2,761 different points-of-interest in the U.S., enriched with place type information. We train classifiers to predict the type of the location a tweet was sent from that reach a macro F1 of 43.67 across eight classes and uncover the linguistic markers associated with each type of place. The ability to predict semantic place information from a tweet has applications in recommendation systems, personalization services and cultural geography. Data is available here: https://archive.org/details/poi-data
• Multi-Domain Named Entity Recognition with Genre-Aware and Agnostic Inference

  Jing Wang, Mayank Kulkarni, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Slides] [Video]

  Named entity recognition is a key component of many text processing pipelines and it is thus essential for these models to be robust to different types of input. However, domain transfer of NER models with data from multiple genres has yet to be studied. To this end, we conduct for the first time NER in three predictive setups on data from: a) multiple domains; b) multiple domains where the genre label is unknown at inference time; c) domains not encountered in training. We introduce a new architecture tailored to this task by using shared and private domain parameters and multi-task learning. This consistently outperforms all other baseline and competitive methods on all three experimental setups, with differences ranging between +1.95 to +3.11 average F1 across multiple genres when compared to standard approaches. These results illustrate the challenges that need to be taken into account when building real-world NLP applications that are robust to various types of text and the methods that can help, at least partially, alleviate these issues.
• Temporally-informed Analysis of Named Entity Recognition

  Shruti Rijhwani, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Data] [Slides] [Video]

  Natural language processing models are often built to make predictions on text data that changes over time as a result of changes in language use or of the information described in text. However, experimental results on existing data sets are seldom reported by taking into account the timestamp of the document being processed. We analyze and propose methods that make better use of temporally-diverse training data, with a focus on the popular task of named entity recognition. To support these experiments, we introduce a novel data set of English tweets annotated with named entities. We empirically demonstrate how the temporal information of documents can be used to obtain better models on new data, as compared to models that disregard temporal information. Our analysis give insights into why this information is useful, in the hope of informing potential avenues of improvement for other NLP tasks under a similar experimental setup.
• Analyzing Political Parody in Social Media

  Antonios Maronikolakis, Danae Sanchez Villegas, Daniel Preoţiuc-Pietro and Nikolaos Aletras

  ACL

  [Abstract] [Data] [Slides] [Video]

  Parody is a figurative device used to imitate an entity for comedic or critical purposes and represents a widespread phenomenon in social media through many popular parody accounts. In this paper, we present the first computational study of parody. We introduce a new publicly available data set of tweets from real politicians and their corresponding parody accounts. We run a battery of supervised machine learning models for automatically detecting parody tweets with an emphasis on robustness by testing on tweets from accounts unseen in training, across different genders and across countries. Our results show that political parody tweets can be predicted with an accuracy up to 90%. Finally, we identify the markers of parody through a linguistic analysis. Beyond research in linguistics and political communication, accurately and automatically detecting parody is important to improving fact checking for journalists and analytics such as sentiment analysis through filtering out parodical utterances.

2019

• Categorizing and Inferring the Relationship between the Text and Image of Twitter Posts

  Alakananda Vempala, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Data] [Slides] [Video]

  Text in social media posts is frequently accompanied by images in order to provide content, supply context, or to express feelings. This paper studies how the meaning of the entire tweet is composed through the relationship between its textual content and its image. We build and release a data set of image tweets annotated with four classes which express the whether the text or the image provides additional information to the other modality. We show that by combining the text and image information, we can build a machine learning approach that accurately distinguishes between the relationship types. Further, we derive insights into how these relationships are materialized through text and image content analysis and how they are impacted by user demographic traits. These methods can be used in several downstream applications including pre-training image tagging models, collecting distantly supervised data for image captioning, and can be directly used in end-user applications to optimize screen estate by automatically hiding images when these contain redundant information to the text.
• Automatically Identifying Complaints in Social Media

  Daniel Preoţiuc-Pietro, Mihaela Găman, Nikolaos Aletras

  ACL

  [Abstract] [Data] [Slides] [Video]

  Complaining is a basic speech act regularly used in human and computer mediated communication to express a negative mismatch between reality and expectations in a particular situation. Automatically identifying complaints in social media is of utmost importance for organizations or brands to improve the customer experience or in developing dialogue systems for handling and responding to complaints. In this paper, we introduce the first systematic analysis of complaints in computational linguistics. We collect a new annotated data set of written complaints expressed on Twitter. We present an extensive linguistic analysis of complaining as a speech act in social media and train strong feature-based and neural models of complaints across nine domains achieving a predictive performance of up to 79 F1 using distant supervision.
• Analyzing Linguistic Differences between Owner and Staff Attributed Tweets

  Daniel Preoţiuc-Pietro, Rita Devlin Marier

  ACL

  [Abstract] [Data] [Slides] [Video]

  Research on social media has to date assumed that all posts from an account are authored by the same person. In this study, we challenge this assumption and study the linguistic differences between posts signed by the account owner or attributed to their staff. We introduce a novel data set of tweets posted by U.S. politicians who self-reported their tweets using a signature. We analyze the linguistic topics and style features that distinguish the two types of tweets. Predictive results show that we are able to predict owner and staff attributed tweets with good accuracy, even when not using any training data from that account.
• Multi-task Pairwise Neural Ranking for Hashtag Segmentation

  Mounica Maddela, Wei Xu, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Code & Data] [Demo] [Poster]

  Hashtags are often employed on social media and beyond to add metadata to a textual utterance with the goal of increasing discoverability, aiding search, or providing additional semantics. However, the semantic content of hashtags is not straightforward to infer as these represent ad-hoc conventions which frequently include multiple words joined together and can include abbreviations and unorthodox spellings. We build a dataset of 12,594 hashtags split into individual segments and propose a set of approaches for this task by framing it as a pairwise ranking problem between candidate segmentations. Our novel neural approaches demonstrate 24.6% error reduction in hashtag segmentation accuracy compared to the current state-of-the-art method. Finally, we demonstrate that a deeper understanding of hashtag semantics obtained through segmentation is useful for downstream applications such as sentiment analysis, for which we achieved a 2.6% increase in average recall on the benchmark SemEval 2017 sentiment analysis dataset.
• What Twitter Profile and Posted Images Reveal about Depression and Anxiety

  Sharath Chandra Guntuku, Daniel Preoţiuc-Pietro, Johannes Eichstaedt, Lyle Ungar

  ICWSM

  [Abstract]

  Previous work has found strong links between the choice of social media images and users' emotions, demographics and personality traits. In this study, we examine which attributes of profile and posted images are associated with depression and anxiety of Twitter users. We used a sample of 28,749 Facebook users to build a language prediction model of survey-reported depression and anxiety, and validated it on Twitter on a sample of 887 users who had taken anxiety and depression surveys. We then applied it to a different set of 4,132 Twitter users to impute language-based depression and anxiety labels, and extracted interpretable features of posted and profile pictures to uncover the associations with users' depression and anxiety, controlling for demographics. For depression, we find that profile pictures suppress positive emotions rather than display more negative emotions, likely because of social media self-presentation biases. They also tend to show the single face of the user (rather than show her in groups of friends), marking increased focus on the self, emblematic for depression. Posted images are dominated by grayscale and low aesthetic cohesion across a variety of image features. Profile images of anxious users are similarly marked by grayscale and low aesthetic cohesion, but less so than those of depressed users. Finally, we show that image features can be used to predict depression and anxiety, and that multitask learning that includes a joint modeling of demographics improves prediction performance. Overall, we find that the image attributes that mark depression and anxiety offer a rich lens into these conditions largely congruent with the psychological literature, and that images on Twitter allow inferences about the mental health status of users.
• Predicting and Analyzing Language Specificity in Social Media Posts

  Yifan Gao, Yang Zhong, Daniel Preoţiuc-Pietro, Junyi Jessy Li

  AAAI

  [Abstract] [Code & Data] [Slides]

  In computational linguistics, specificity quantifies how much detail is engaged in text. It is an important characteristic of speaker intention and language style, and is useful in NLP applications such as summarization and argumentation mining. Yet to date, expert-annotated data for sentence-level specificity are scarce and confined to the news genre. In addition, systems that predict sentence specificity are classifiers trained to produce binary labels (general or specific). We collect a dataset of over 7,000 tweets annotated with specificity on a fine-grained scale. Using this dataset, we train a supervised regression model that accurately estimates specificity in social media posts, reaching a mean absolute error of 0.3578 (for ratings on a scale of 1-5) and 0.73 Pearson correlation, significantly improving over baselines and previous sentence specificity prediction systems. We also present the first large-scale study revealing the social, temporal and mental health factors underlying language specificity on social media.

2018

• The Impact of Actively Open-minded Thinking on Social Media Communication

  Jordan Carpenter, Daniel Preoţiuc-Pietro, Jenna Clark, Lucie Flekova, Laura Smith, Margaret L. Kern, Anneke Buffone, Lyle Ungar, Martin Seligman

  Judgement & Decision Making

  [Abstract]

  Online, social media communication is often ambiguous, and it can encourage speed and inattentiveness. We investigated whether Actively Open Minded Thinking (AOT), a dispositional willingness to seek out new or potentially threatening information, may help users avoid these pitfalls. In Study 1, we determined that correctly assessing social media authors’ traits was positively predicted by raters’ AOT. In Study 2, we used data-driven methods to devise a three-dimensional picture of online behaviors of people high or low in AOT, finding that AOT is associated with thoughtful, nuanced, idiosyncratic actions and with resisting the typically fast pace of online interactions. AOT may be an important factor in accurate, socially responsible online behavior.
• Facebook Language Predicts Depression in Medical Records

  Johannes Eichstaedt, Robert Smith, Raina Merchant, Lyle Ungar, Patrick Crutchley, Daniel Preoţiuc-Pietro, David Asch, H. Andrew Schwartz

  PNAS

  [Abstract]

  Depression, the most prevalent mental illness, is underdiagnosed and undertreated, highlighting the need to extend the scope of current screening methods. Here, we use language from Facebook posts of consenting individuals to predict depression recorded in electronic medical records. We accessed the history of Facebook statuses posted by 683 patients visiting a large urban academic emergency department, 114 of whom had a diagnosis of depression in their medical records. Using only the language preceding their first documentation of a diagnosis of depression, we could identify depressed patients with fair accuracy [area under the curve (AUC) = 0.69], approximately matching the accuracy of screening surveys benchmarked against medical records. Restricting Facebook data to only the 6 months immediately preceding the first documented diagnosis of depression yielded a higher prediction accuracy (AUC = 0.72) for those users who had sufficient Facebook data. Significant prediction of future depression status was possible as far as 3 months before its first documentation. We found that language predictors of depression include emotional (sadness), interpersonal (loneliness, hostility), and cognitive (preoccupation with the self, rumination) processes. Unobtrusive depression assessment through social media of consenting individuals may become feasible as a scalable complement to existing screening and monitoring procedures.
• Why Swear? Analyzing and Inferring the Intentions of Vulgar Expressions

  Eric Holgate, Isabel Cachola, Daniel Preoţiuc-Pietro, Junyi Jessy Li

  EMNLP

  [Abstract] [Code & Data] [Slides]

  Vulgar words are employed in language use for several different functions, ranging from expressing aggression to signaling group identity or the informality of the communication. This versatility of usage of a restricted set of words is challenging for downstream applications and has yet to be studied quantitatively or using natural language processing techniques. We introduce a novel data set of 7,800 tweets from users with known demographic traits where all instances of vulgar words are annotated with one of the six categories of vulgar word use. Using this data set, we present the first analysis of the pragmatic aspects of vulgarity and how they relate to social factors. We build a model able to predict the category of a vulgar word based on the immediate context it appears in with 67.4 macro F1 across six classes. Finally, we demonstrate the utility of modeling the type of vulgar word use in context by using this information to achieve state-of-the-art performance in hate speech detection on a benchmark data set.
• The Remarkable Benefit of User-Level Aggregation for Lexical-based Population-Level Predictions

  Salvatore Giorgi, Daniel Preoţiuc-Pietro, Anneke Buffone, Daniel Rieman, Lyle Ungar, H. Andrew Schwartz

  EMNLP

  [Abstract]

  Nowcasting based on social media text promises to provide unobtrusive and near real-time predictions of community-level outcomes. These outcomes are typically regarding people, but the data is often aggregated without regard to users in the Twitter populations of each community. This paper describes a simple yet effective method for building community-level models using Twitter language aggregated by user. Results on four different U.S. county-level tasks, spanning demographic, health, and psychological outcomes show large and consistent improvements in prediction accuracies (e.g. from Pearson r = .73 to .82 for median income prediction or r = .37 to .47 for life satisfaction prediction) over the standard approach of aggregating all tweets. We make our aggregated and anonymized community-level data, derived from 37 billion tweets – over 1 billion of which were mapped to counties, available for research.
• Cross-Platform and Cross-Interaction Study of User Personality based on Images on Twitter and Flickr

  Zahra Riahi Samani, Sharath Chandra Guntuku, Mohsen Ebrahimi Moghaddam, Daniel Preoţiuc-Pietro, Lyle Ungar

  PLoS ONE

  [Abstract]

  Assessing the predictive value of different social media platforms is important to understand the variation in how users reveal themselves across multiple platforms. Most social media platforms allow users to interact in multiple ways: by posting content to the platform, liking others’ posts, or building a user profile. While prior studies offer insights into how language use differs across platforms, differences in image usage is less well understood. In this study, we analyzed variation in image content with user personality across three interaction types (posts, likes and profile images) and two platforms, using a unique data set of users who are active on both Twitter and Flickr. Usage patterns on these two social media platforms revealed different aspects of users’ personality. Cross-platform data fusion is thus shown to improve personality prediction performance.
• User-Level Race and Ethnicity Predictors from Twitter Text

  Daniel Preoţiuc-Pietro, Lyle Ungar

  COLING

  [Abstract] [Poster]

  User demographic inference from social media text has the potential to improve a range of downstream applications, including real-time passive polling or quantifying demographic bias. This study focuses on developing models for user-level race and ethnicity prediction. We introduce a data set of users who self-report their race/ethnicity through a survey, in contrast to previous approaches that use distantly supervised data or perceived labels. We develop predictive models from text which accurately predict the membership of a user to the four largest racial and ethnic groups with up to .884 AUC and make these available to the research community.
• Expressively vulgar: The socio-dynamics of vulgarity and its effects on sentiment analysis in social media

  Isabel Cachola, Eric Holgate, Daniel Preoţiuc-Pietro, Junyi Jessy Li

  COLING

  [Abstract] [Code & Data] [Slides]

  Vulgarity is a common linguistic expression and is used to perform several linguistic functions. Understanding their usage can aid both linguistic and psychological phenomena as well as benefit downstream natural language processing applications such as sentiment analysis. This study performs a large-scale, data-driven empirical analysis of vulgar words using social media data. We analyze the socio-cultural and pragmatic aspects of vulgarity using tweets from users with known demographics. Further, we collect sentiment ratings for vulgar tweets to study the relationship between the use of vulgar words and perceived sentiment and show that explicitly modeling vulgar words can boost sentiment analysis performance.

2017

• Controlling Human Perception of Basic User Traits

  Daniel Preoţiuc-Pietro, Sharath Chandra Guntuku, Lyle Ungar

  EMNLP

  [Abstract] [Poster]

  Much of our online communication is text-mediated and, lately, more common with automated agents. Unlike interacting with humans, these agents currently do not tailor their language to the type of person they are communicating to. In this pilot study, we measure the extent to which human perception of basic user trait information – gender and age – is controllable through text. Using automatic models of gender and age prediction, we estimate which tweets posted by a user are more likely to mis-characterize his traits. We perform multiple controlled crowdsourcing experiments in which we show that we can reduce the human prediction accuracy of gender to almost random – a > 20% drop in accuracy. Our experiments show that it is practically feasible for multiple applications such as text generation, text summarization or machine translation to be tailored to specific traits and perceived as such.
• Beyond Binary Labels: Political Ideology Prediction of Twitter Users

  Daniel Preoţiuc-Pietro, Ye Liu, Daniel Hopkins, Lyle Ungar

  ACL

  [Abstract] [Slides] [Video]

  Automatic political orientation prediction from social media posts has to date proven successful only in distinguishing between publicly declared liberals and conservatives in the US. This study examines users’ political ideology using a seven-point scale which enables us to identify politically moderate and neutral users – groups which are of particular interest to political scientists and pollsters. Using a novel data set with political ideology labels self-reported through surveys, our goal is two-fold: a) to characterize the groups of politically engaged users through language use on Twitter; b) to build a fine-grained model that predicts political ideology of unseen users. Our results identify differences in both political leaning and engagement and the extent to which each group tweets using political keywords. Finally, we demonstrate how to improve ideology prediction accuracy by exploiting the relationships between the user groups.
• Personality Driven Differences in Paraphrase Preference

  Daniel Preoţiuc-Pietro, Jordan Carpenter, Lyle Ungar

  Workshop on Natural Language Processing and Computational Social Science (NLP+CSS), ACL

  [Abstract] [Data] [Slides]

  Personality plays a decisive role in how people behave in different scenarios, including online social media. Researchers have used such data to study how personality can be predicted from language use. In this paper, we study phrase choice as a particular stylistic linguistic difference, as opposed to the mostly topical differences identified previously. Building on previous work on demographic preferences, we quantify differences in paraphrase choice from a massive Facebook data set with posts from over 115,000 users. We quantify the predictive power of phrase choice in user profiling and use phrase choice to study psycholinguistic hypotheses. This work is relevant to future applications that aim to personalize text generation to specific personality types.
• Personality Profiles of Users Sharing Animal-related Content on Social Media

  Courtney Hagan, Jordan Carpenter, Lyle Ungar, Daniel Preoţiuc-Pietro

  Anthrozoos

  [Abstract]

  Animal preferences are thought to be linked with more salient psychological traits of people and most research examining owner personality as a differentiating factor has obtained mixed results. The rise in usage of social networks offers users a new medium in which users broadcast their preferences and activities, including about animals. In two studies, the first on Facebook status updates and the second on images shared on Twitter, we revisited the link between user Big Five personality traits and animal preference, specifically focusing on cats and dogs. We used automatic content analysis of text and images to unobtrusively measure preference for animals online using large data sets. Results from Study 1 indicated that those who mentioned ownership of a cat (by using the phrase ‘my cat’) in their status updates were more open to experience, introverted, neurotic and less conscientious when compared to the general population, while users mentioning ownership of a dog (by using ‘my dog’) were only less conscientious compared to the rest of the population. Study 2 foundfinds that users who featured either cat or dog images in their tweets are more neurotic, less conscientious and less agreeable than those who do not. In addition, posting images containing cats was specific to users higher in openness, while posting images featuring dogs was associated with users higher in extraversion. These findings taken together align with some previous findings on the relationship between owner personality and animal preference, additionally highlighting some social media specific behaviors.
• Predicting Emotional Word Ratings using Distributional Representations and Signed Clustering

  Joao Sedoc, Daniel Preoţiuc-Pietro, Lyle Ungar

  EACL

  [Abstract]

  Inferring the emotional content of words is important for text-based sentiment analysis, dialogue systems and psycholinguistics, but word ratings are expensive to collect at scale and across languages or domains. We develop a method that automatically extends word-level ratings to unrated words using signed clustering of vector space word representations along with affect ratings. We use our method to determine a word's valence and arousal, which determine its position on the circumplex model of affect, the most popular dimensional model of emotion. Our method achieves superior out-of-sample word rating prediction on both affective dimensions across three different languages when compared to state-of-the-art word similarity based methods. Our method can assist building word ratings for new languages and improve downstream tasks such as sentiment analysis and emotion detection.
• Studying Personality through the Content of Posted and Liked Images on Twitter

  Sharath Chandra Guntuku, Weisi Lin, Jordan Carpenter, Wee Keong Ng, Lyle Ungar, Daniel Preoţiuc-Pietro

  Web Science

  [Abstract]

  Interacting with images through social media has become widespread due to ubiquitous Internet access and multimedia enabled devices. Through images, users generally present their daily activities, preferences or interests. This study aims to identify the way and extent to which personality differences measured as using the Big Five model are related to online image posting and liking. In two experiments, the larger consisting of ~$1.5 million Twitter images both posted and liked by ~4,000 users, we extract interpretable semantic concepts using large-scale image content analysis and analyze differences specific of each personality trait. Predictive results show that image content can predict personality traits, and that there can be significant performance gain by fusing the signal from both posted and liked images.

2016

• Studying the Dark Triad of Personality using Twitter Behavior

  Daniel Preoţiuc-Pietro, Jordan Carpenter, Salvatore Giorgi, Lyle Ungar

  CIKM

  [Abstract] [Data] [Models] [Slides]

  Research into the darker traits of human nature is growing in interest especially in the context of increased social media usage. This allows users to express themselves to a wider online audience. We study the extent to which the standard model of dark personality – the dark triad – consisting of narcissism, psychopathy and Machiavellianism, is related to observable Twitter behavior such as platform usage, posted text and profile image choice. Our results show that we can map various behaviors to psychological theory and study new aspects related to social media usage. Finally, we build a machine learning algorithm that predicts the dark triad of personality in out-of-sample users with reliable accuracy.
• Analysing Biases in Human Perception of User Age and Gender from Text

  Lucie Flekova, Jordan Carpenter, Salvatore Giorgi, Lyle Ungar, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Poster]

  User traits disclosed through written text, such as age and gender, can be used to personalize applications such as recommender systems or conversational agents. However, human perception of these traits is not perfectly aligned with reality. In this paper, we conduct a large-scale crowdsourcing experiment on guessing age and gender from tweets. We systematically analyze the quality and possible biases of these predictions. We identify the textual cues which lead to miss-assessments of traits or make workers more or less confident in their choice. Our study demonstrates that differences between real and perceived traits are noteworthy and elucidates inaccurately used stereotypes in human perception.
• Predicting Judicial Decisions of the European Court of Human Rights: A Natural Language Processing Perspective

  Nikolaos Aletras, Dimitrios Tsarapatsanis, Daniel Preoţiuc-Pietro, Vasileios Lampos

  PeerJ Computer Science

  [Abstract] [Data]

  Recent advances in Natural Language Processing and Machine Learning provide us with the tools to build predictive models that can be used to unveil patterns driving judicial decisions. This can be useful, for both lawyers and judges, as an assisting tool to rapidly identify cases and extract patterns which lead to certain decisions. This paper presents the first systematic study on predicting the outcome of cases tried by the European Court of Human Rights based solely on textual content. We formulate a binary classification task where the input of our classifiers is the textual content extracted from a case and the target output is the actual judgment as to whether there has been a violation of an article of the convention of human rights. Textual information is represented using contiguous word sequences, i.e. N-grams, and topics. Our models can predict the court’s decisions with a strong accuracy (79% on average). Our empirical analysis indicates that the formal facts of a case are the most important predictive factor. This is consistent with the theory of legal realism suggesting that judicial decision-making is significantly affected by the stimulus of the facts. We also observe that the topical content of a case is another important feature in this classification task and explore this relationship further by conducting a qualitative analysis
• Sub-Story Detection in Twitter with Hierarchical Dirichlet Processes

  Srijith P.K., Kalina Bontcheva, Mark Hepple, Daniel Preoţiuc-Pietro

  Information Processing and Management

  [Abstract]

  Social media has now become the de facto information source on real world events. The challenge, however, due to the high volume and velocity nature of social media streams, is in how to follow all posts pertaining to a given event over time – a task referred to as story detection. Moreover, there are often several different stories pertaining to a given event, which we refer to as sub-stories and the corresponding task of their automatic detection – as sub-story detection. This paper proposes hierarchical Dirichlet processes (HDP), a probabilistic topic model, as an effective method for automatic sub-story detection. HDP can learn sub-topics associated with sub-stories which enables it to handle subtle variations in sub-stories. It is compared with state-of-the-art story detection approaches based on locality sensitive hashing and spectral clustering. We demonstrate the superior performance of HDP for sub-story detection on real world Twitter data sets using various evaluation measures. The ability of HDP to learn sub-topics helps it to recall the sub-stories with high precision. This has resulted in an improvement of up to 60% in the F-score performance of HDP based sub-story detection approach compared to standard story detection approaches. A similar performance improvement is also seen using an information theoretic evaluation measure proposed for the sub-story detection task. Another contribution of this paper is in demonstrating that considering the conversational structures within the Twitter stream can bring up to 200% improvement in sub-story detection performance.
• Real Men don’t say 'cute': Using Automatic Language Analysis to Isolate Inaccurate Aspects of Stereotypes

  Jordan Carpenter, Daniel Preoţiuc-Pietro, Lucie Flekova, Salvatore Giorgi, Courtney Hagan, Margaret Kern, Anneke Buffone, Lyle Ungar, Martin Seligman

  Social Psychological and Personality Science

  [Abstract] [Supplemental Materials]

  People associate certain behaviors with certain social groups. These stereotypical beliefs consist of both accurate and inaccurate associations. Using large-scale, data driven methods with social media as a context, we isolate stereotypes by using verbal expression. Across four social categories: gender, age, education level, and political orientation. We identify words and phrases that lead people to incorrectly guess the social category of the writer. Although raters often correctly categorize authors, they overestimate the importance of some stereotype-congruent signal. Findings suggest that data-driven approaches might be a valuable and ecologically valid tool for identifying even subtle aspects of stereotypes and highlighting the facets that are exaggerated or misapplied.
• Exploring Stylistic Variation with Age and Income on Twitter

  Lucie Flekova, Lyle Ungar, Daniel Preoţiuc-Pietro

  ACL

  [Abstract] [Slides]

  Writing style allows NLP tools to adjust to the traits of an author. In this paper, we explore the relation between stylistic and syntactic features and authors’ age and income. We confirm our hypothesis that for numerous feature types writing style is predictive of income even beyond age. We analyze the predictive power of writing style features in a regression task on two data sets of around 5,000 Twitter users each. Additionally, we use our validated features to study daily variations in writing style of users from distinct income groups. Temporal stylistic patterns not only provide novel psychological insight into user behavior, but are useful for future research and applications in social media.
• #Microposts2016 – 6th Workshop on ‘Making Sense of Microposts’

  Amparo Elisabeth Cano, Daniel Preoţiuc-Pietro, Danica Radovanovic, Katrin Weller, Aba-Sah Dadzie

  WWW

  [Abstract]

  #Microposts2016, the 6th workshop on Making Sense of Microposts, is summarised by the sub-theme: big things come in small packages. The workshop serves as a forum to discuss and promote research on the generation, analysis and reuse of Microposts – small chunks of information published on social media and messaging platforms. Low effort and cost to publish Microposts gives a voice to all, across differences in expertise, socio-cultural, generational and economic spheres, covering a wide swathe of topics, posted in the moment and on the go, during events, crises and personal experiences. While the usual suspects, including Twitter, Facebook, Instagram and Pinterest continue to dominate, especially as services are merged or shared across platforms, newer players such as WhatsApp, Vine, Meerkat and Yik Yak are growing in popularity, with increased access to fast, high capacity networks and advanced small, personal devices. #Microposts2016 solicited participation from Computer Science and other relevant fields, with a focus on interdisciplinary work. Starting in 2015, the workshop includes a track dedicated to encouraging research employing methods for analysis of Microposts in the Social Sciences.
• Analyzing Personality through Social Media Profile Picture Choice

  Liu Leqi, Daniel Preoţiuc-Pietro, Zahra Riahi, Mohsen E. Moghaddam, Lyle Ungar

  ICWSM

  [Abstract] [Data] [Slides]

  The content of images users post to their social media is driven in part by personality. In this study, we analyze how Twitter profile images vary with the personality of the users posting them. In our main analysis, we use profile images from over 66,000 users whose personality we estimate based on their tweets. To facilitate interpretability, we focus our analysis on aesthetic and facial features and control for demographic variation in image features and personality. Our results show significant differences in profile picture choice between personality traits, and that these can be harnessed to predict personality traits with robust accuracy. For example, agreeable and conscientious users display more positive emotions in their profile pictures, while users high in openness prefer more aesthetic photosd
• Discovering User Attribute Stylistic Differences via Paraphrasing

  Daniel Preoţiuc-Pietro, Wei Xu, Lyle Ungar

  AAAI

  [Abstract] [Data] [Slides]

  User attribute prediction from social media text has proven successful and useful for downstream tasks. In previous studies, user trait differences have been limited primarily to the presence or absence of words that indicate topical preferences. In this study, we aim to find linguistic style distinctions across three different user attributes: gender, age and occupational class. By combining paraphrases with a simple yet effective method, we capture a wide set of stylistic differences that are exempt from topic bias. We show their predictive power in user profiling, conformity with human perception and psycholinguistic hypotheses, and potential use in generating natural language tailored to specific user traits.
• An Empirical Exploration of Moral Foundations Theory in Partisan News Sources

  Dean Fulgoni, Jordan Carpenter, Lyle Ungar, Daniel Preoţiuc-Pietro

  LREC

  [Abstract] [Slides]

  News sources frame issues in different ways in order to appeal or control the perception of their readers. We present a large scale study of news articles from partisan sources in the US across a variety of different issues. We first highlight that differences between sides exist by predicting the political leaning of articles of unseen political bias. Framing can be driven by different types of morality that each group values. We emphasize differences in framing of different news building on the moral foundations theory quantified using hand crafted lexicons. Our results show that partisan sources frame political issues differently both in terms of words usage and through the moral foundations they relate to.
• Modelling Valence and Arousal in Facebook posts

  Daniel Preoţiuc-Pietro, Andrew H. Schwartz, Gregory Park, Johannes Eichstaedt, Margaret Kern, Lyle Ungar, Elisabeth Shulman

  Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA), NAACL

  [Abstract] [Data] [Slides]

  Access to expressions of subjective personal posts increased with the popularity of Social Media. However, most of the work in sentiment analysis focuses on predicting only valence from text and usually targeted at a product, rather than affective states. In this paper, we introduce a new data set of 2895 Social Media posts rated by two psychologically-trained annotators on two separate ordinal nine-point scales. These scales represent valence (or sentiment) and arousal (or intensity), which defines each post’s position on the circumplex model of affect, a well-established system for describing emotional states (Russell, 1980; Posner et al., 2005). The data set is used to train prediction models for each of the two dimensions from text which achieve high predictive accuracy – correlated at r = :65 with valence and r = :85 with arousal annotations. Our data set offers a building block to a deeper study of personal affect as expressed in social media. This can be used in applications such as mental illness detection or in automated large-scale psychological studies.
• Studying the temporal dynamics of word co-occurrences: An application to event detection

  Daniel Preoţiuc-Pietro, Srijith P.K., Mark Hepple, Trevor Cohn

  LREC

  [Abstract] [Code & Data] [Slides]

  Streaming media provides a number of unique challenges for computational linguistics. This paper studies the temporal variation in word co-occurrence statistics, with application to event detection. We develop a spectral clustering approach to find groups of mutually informative terms occurring in discrete time frames. Experiments on large datasets of tweets show that these groups identify key real world events as they occur in time, despite no explicit supervision. The performance of our method rivals state-of-the-art methods for event detection on F-score, obtaining higher recall at the expense of precision.

2015

• Studying User Income through Language, Behaviour and Affect in Social Media

  Daniel Preoţiuc-Pietro, Svitlana Volkova, Vasileios Lampos, Yoram Bachrach, Nikolaos Aletras

  PLoS ONE

  [Abstract] [Data]

  Automatically inferring user demographics from social media posts is useful for both social science research and a range of downstream applications in marketing and politics. We present the first extensive study where user behaviour on Twitter is used to build a predictive model of income. We apply non-linear methods for regression, i.e. Gaussian Processes, achieving strong correlation between predicted and actual user income. This allows us to shed light on the factors that characterise income on Twitter and analyse their interplay with user emotions and sentiment, perceived psycho-demographics and language use expressed through the topics of their posts. Our analysis uncovers correlations between different feature categories and income, some of which reflect common belief e.g. higher perceived education and intelligence indicates higher earnings, known differences e.g. gender and age differences, however, others show novel findings e.g. higher income users express more fear and anger, whereas lower income users express more of the time emotion and opinions.
• Analysing domain suitability of a sentiment lexicon by identifying distributionally bipolar words

  Lucie Flekova, Eugen Ruppert, Daniel Preoţiuc-Pietro

  Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA), EMNLP

  [Abstract] [Slides]

  Contemporary sentiment analysis approaches rely heavily on lexicon based methods. This is mainly due to their simplicity, although the best empirical results can be achieved by more complex techniques. We introduce a method to assess suitability of generic sentiment lexicons for a given domain, namely to identify frequent bigrams where a polar word switches polarity. Our bigrams are scored using Lexicographers Mutual Information and leveraging large automatically obtained corpora. Our score matches human perception of polarity and demonstrates improvements in classification results using our enhanced context-aware method. Our method enhances the assessment of lexicon based sentiment detection algorithms and can be further used to quantify ambiguous words.
• An analysis of the user occupational class through Twitter content

  Daniel Preoţiuc-Pietro, Vasileios Lampos, Nikolaos Aletras

  ACL

  [Abstract] [Data] [Slides]

  Social media content can be used as a complementary source to the traditional methods for extracting and studying collective social attributes. This study focuses on the prediction of the occupational class for a public user profile. Our analysis is conducted on a new annotated corpus of Twitter users, their respective job titles, posted textual content and platform-related attributes. We frame our task as classification using latent feature representations such as word clusters and embeddings. The employed linear and, especially, non-linear methods can predict a user’s occupational class with strong accuracy for the coarsest level of a standard occupation taxonomy which includes nine classes. Combined with a qualitative assessment, the derived results confirm the feasibility of our approach in inferring a new user attribute that can be embedded in a multitude of downstream applications.
• Analyzing crowdsourced assessment of user traits through Twitter posts

  Lucie Flekova, Daniel Preoţiuc-Pietro, Jordan Carpenter, Salvatore Giorgi, Lyle Ungar

  HCOMP 2015 (Work-in-Progress)

  [Abstract] [Poster] [Supplementary Materials]

  Social media allows any user to express themselves to the public through posting content. Using a crowdsourcing experiment, we aim to quantify and analyze which human attributes lead to better perceptions of the true identity of others. Using tweet content from a set of users with known age and gender information, we ask workers to rate their perception of these traits and we analyze those results in relation to the crowdsourcing workers’ age and gender. Results show that female workers are both more confident and more accurate at reporting gender, and workers in their thirties were most accurate but least confident for rating age. Our study is a first step in identifying the worker traits which contribute to a better understanding of others through their posted text content. Our findings help to identify the types of workers best suited for certain tasks.
• Mental Illness Detection at the World Well-Being Project for the CLPsych 2015 Shared Task

  Daniel Preoţiuc-Pietro, Maarten Sap, Andrew H. Schwartz, Lyle Ungar

  Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (CLPysch), NAACL

  [Abstract]

  This article is a system description and report on the submission of the World Well-Being Project from the University of Pennsylvania in the 'CLPsych 2015' shared task. The goal of the shared task was to automatically determine Twitter users who self-reported having one of two mental illnesses: post traumatic stress disorder (PTSD) and depression. Our system employs user metadata and textual features derived from Twitter posts. To reduce the feature space and avoid data sparsity, we consider several word clustering approaches. We explore the use of linear classifiers based on different feature sets as well as a combination use a linear ensemble. This method is agnostic of illness specific features, such as lists of medicines, thus making it readily applicable in other scenarios. Our approach ranked second in all tasks on average precision and showed best results at .1 false positive rates.
• The Role of Personality, Age and Gender in Tweeting about Mental Illnesses

  Daniel Preoţiuc-Pietro, Johannes Eichstaedt, Gregory Park, Maarten Sap, Laura Smith, Victoria Tobolsky, Andrew H. Schwartz, Lyle Ungar

  Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (CLPsych), NAACL

  [Abstract] [Slides]

  Mental illnesses, such as depression and post traumatic stress disorder (PTSD), are highly underdiagnosed globally. Populations sharing similar demographics and personality traits are known to be more at risk than others. In this study, we characterise the language use of users disclosing their mental illness on Twitter. Language-derived personality and demographic estimates show surprisingly strong performance in distinguishing users that tweet a diagnosis of depression or PTSD from random controls, reaching an area under the receiver operating characteristic curve – AUC – of around .8 in all our binary classification tasks. In fact, when distinguishing users disclosing depression from those disclosing PTSD, the single feature of estimated age shows nearly as strong performance (AUC = .806) as using thousands of topics (AUC = .819) or tens of thousands of n-grams (AUC = .812). We also find that differential language analyses, controlled for demographics, recover many symptoms associated with the mental illnesses in the clinical literature.

2014

• Extracting socioeconomic patterns from the news: Modelling text and outlet importance jointly

  Vasileios Lampos, Daniel Preoţiuc-Pietro, Sina Samangooei, Douwe Gelling, Trevor Cohn

  Workshop on Language Technologies and Computational Social Science (LACSS), ACL

  [Abstract] [Data] [Poster]

  Information from news articles can be used to study correlations between textual discourse and socioeconomic patterns. This work focuses on the task of understanding how words contained in the news as well as the news outlets themselves may relate to a set of indicators, such as economic sentiment or unemployment rates. The bilinear nature of the applied regression model facilitates learning jointly word and outlet importance, supervised by these indicators. By evaluating the predictive ability of the extracted features, we can also assess their relevance to the target socioeconomic phenomena. Therefore, our approach can be formulated as a potential NLP tool, particularly suitable to the computational social science community, as it can be used to interpret connections between vast amounts of textual content and measurable society driven factors.
• Predicting and characterising user impact on Twitter

  Vasileios Lampos, Nikolaos Aletras, Daniel Preoţiuc-Pietro, Trevor Cohn

  EACL

  [Abstract] [Poster]

  The open structure of online social networks and their uncurated nature give rise to problems of user credibility and influence. In this paper, we address the task of predicting the impact of Twitter users based only on features under their direct control, such as usage statistics and the text posted in their tweets.We approach the problem as regression and apply linear as well as nonlinear learning methods to predict a user impact score, estimated by combining the numbers of the user’s followers, followees and listings. The experimental results point out that a strong prediction performance is achieved, especially for models based on the Gaussian Processes framework. Hence, we can interpret various modelling components, transforming them into indirect ‘suggestions’ for impact boosting.
• Tools for mining non-stationary data - v2. Clustering models for discovery of regional and demographic variation v2

  Daniel Preoţiuc-Pietro, Sina Samangooei, Andrea Varga, Douwe Gelling, Trevor Cohn, Mahesan Niranjan

  Public Deliverable for Trendminer Project.

  [Abstract]

  This document presents advanced research and software development work for Task 3.2 on tools for mining non-stationary data and for Task 3.3 on clustering models integrating regional and demographic information for the aim of understanding streaming data. First, for modelling non-stationary data, a research experiment is presented for categorising and forecasting word frequency patterns using Gaussian Processes, with an emphasis on word periodicities. A new soft clustering method based on topic models is introduced, which learns topics and their temporal profile jointly. For using regional and demographic user information, the predictive model presented in previous work (Samangooei et al., 2013) is extended. This is used to identify differences in voting intention between different regions of the United Kingdom and different genders. For discovering specific regional clusters, the soft clustering technique is extended to learn the topics, their regional and temporal profile jointly. Finally, the predictive and clustering models developed on social media data are applied to a news summary dataset where richer linguistic features are also used.
• Temporal models of streaming social media data

  Daniel Preoţiuc-Pietro

  University of Sheffield

  [Abstract]

  There are significant temporal dependencies between online behaviour and occurring real world activities. Particularly in text modelling, these are usually ignored or at best dealt with in overly simplistic ways such as assuming smooth variation with time. Social media is a new data source which present collective behaviour much more richly than traditional sources, such as newswire, with a finer time granularity, timely reflection of activities, multiple modalities and large volume. Analysing temporal patterns in this data is important in order to discover newly emerging topics, periodic occurrences and correlation or causality to real world indicators or human behaviour patterns. With these opportunities come many challenges, both engineering (i.e. data volume and processing) and algorithmic, namely the inconsistency and short length of the messages and the presence of large amounts of irrelevant messages to our goal. Equipped with a better understanding of the dynamics of the complex temporal dependencies, tasks such as classification can be augmented to provide temporally aware responses. In this thesis we model the temporal dynamics of social media data. We first show that temporality is an important characteristic of this type of data. Further comparisons and correlation to real world indicators show that this data gives a timely reflection of real world events. Our goal is to use these variations to discover emerging or recurring user behaviours. We consider both the use of words and user behaviour in social media. With these goals in mind, we adapt existing and build novel machine learning techniques. These span a wide range of models: from Markov models to regularised regression models and from evolutionary spectral clustering which models smooth temporal variation to Gaussian Process regression which can identify more complex temporal patterns. We introduce approaches which discover and predict words, topics or behaviours that change over time or occur with some regularity. These are modeled for the first time in the NLP literature by using Gaussian Processes. We demonstrate that we can effectively pick out patterns, including periodicities, and achieve state-of-the-art forecasting results. We show that this performance gain transfers to improve tasks which do not take temporal information in account. Further analysed is how temporal variation in the text can be used to discover and track new content. We develop a model that exploits the variation in word co-occurrences for clustering over time. Different collection and processing tools, as well as several datasets of social media data have been developed and published as open-source software. The thesis posits that temporal analysis of data, from social media in particular, provides us with insights into real-world dynamics. Incorporating this temporal information into other applications can benefit standard tasks in natural language processing and beyond.

2013

• A temporal model of text periodicities using Gaussian Processes

  Daniel Preoţiuc-Pietro, Trevor Cohn

  EMNLP

  [Abstract] [Data] [Poster]

  Temporal variations of text are usually ignored in NLP applications. However, text use changes with time, which can affect many applications. In this paper we model periodic distributions of words over time. Focusing on hashtag frequency in Twitter, we first automatically identify the periodic patterns. We use this for regression in order to forecast the volume of a hashtag based on past data. We use Gaussian Processes, a state-of-the-art bayesian non-parametric model, with a novel periodic kernel. We demonstrate this in a text classification setting, assigning the tweet hashtag based on the rest of its text. This method shows significant improvements over competitive baselines.
• Mining user behaviours: A study of check-in patterns in Location Based Social Networks

  Daniel Preoţiuc-Pietro, Trevor Cohn

  Web Science

  [Abstract] [Data] [Poster]

  Understanding the patterns underlying human mobility is of an essential importance to applications like recommender systems. In this paper we investigate the behaviour of around 10,000 frequent users of Location Based Social Networks (LBSNs) making use of their full movement patterns. We analyse the metadata associated with the whereabouts of the users, with emphasis on the type of places and their evolution over time. We uncover patterns across different temporal scales for venue category usage. Then, focusing on individual users, we apply this knowledge in two tasks: 1) clustering users based on their behaviour and 2) predicting users’ future movements. By this, we demonstrate both qualitatively and quantitatively that incorporating temporal regularities is beneficial for making better sense of user behaviour.
• Where's @wally: A classification approach to geolocating users based on their social ties

  Dominic Rout, Daniel Preoţiuc-Pietro, Bontcheva Kalina, Trevor Cohn

  HT

  [Abstract] [Data] [Slides]

  This paper presents an approach to geolocating users of online social networks, based solely on their ‘friendship’ connections. We observe that users interact more regularly with those closer to themselves and hypothesise that, in many cases, a person’s social network is sufficient to reveal their location. The geolocation problem is formulated as a classification task, where the most likely city for a user without an explicit location is chosen amongst the known locations of their social ties. Our method uses an SVM classifier and a number of features that reflect different aspects and characteristics of Twitter user networks. The SVM classifier is trained and evaluated on a dataset of Twitter users with known locations. Our method outperforms a state-of-the-art method for geolocating users based on their social ties.
• Exploring venue-based city-to-city similarity measures

  Daniel Preoţiuc-Pietro, Justin Cranshaw, Tae Yano

  Workshop on Urban Computing (UrbComp), SIGKDD 2013

  [Abstract]

  In this work we explore the use of incidentally generated social network data for the folksonomic characterization of cities by the types of amenities located within them. Using data collected about venue categories in various cities, we examine the effect of different granularities of spatial aggregation and data normalization when representing a city as a collection of its venues. We introduce three vector-based representations of a city, where aggregations of the venue categories are done within a grid structure, within the city’s municipal neighborhoods, and across the city as a whole. We apply our methods to a novel dataset consisting of Foursquare venue data from 17 cities across the United States, totaling over 1 million venues. Our preliminary investigation demonstrates that different assumptions in the urban perception could lead to qualitative, yet distinctive, variations in the induced city description and categorization.
• Clustering models for discovery of regional and demographic variation

  Daniel Preoţiuc-Pietro, Sina Samangooei, Vasileios Lampos, Trevor Cohn, Nick Gibbins, Mahesan Niranjan

  Public Deliverable for Trendminer Project

  [Abstract]

  This deliverable describes the first experiments for clustering in social media streams. Our main goal in these experiments was the constructions of scalable clustering algorithms capable of dealing with extremely large datasets as well as streaming contexts. In particular, two algorithms, spectral clustering and DBSCAN, are described. The experiments are conducted on two different scenarios. First is clustering multi-modal objects into events whilst second is clustering words in a Twitter corpora with Austrian data. The former was submitted to MediaEval 2013 Multimedia Social Event Detection benchmark at which it achieved the best results submitted in 2013, whilst the later represents a study of word clustering in the context of a Trend Miner use case.
• A user-centric model of voting intention from Social Media

  Vasileios Lampos, Daniel Preoţiuc-Pietro, Trevor Cohn

  ACL

  [Abstract] [Slides]

  Social Media contain a multitude of user opinions which can be used to predict realworld phenomena in many domains including politics, finance and health. Most existing methods treat these problems as linear regression, learning to relate word frequencies and other simple features to a known response variable (e.g., voting intention polls or financial indicators). These techniques require very careful filtering of the input texts, as most Social Media posts are irrelevant to the task. In this paper, we present a novel approach which performs high quality filtering automatically, through modelling not just words but also users, framed as a bilinear model with a sparse regulariser. We also consider the problem of modelling groups of related output variables, using a structured multi-task regularisation method. Our experiments on voting intention prediction demonstrate strong performance over large-scale input from Twitter on two distinct case studies, outperforming competitive baselines.

2012

• Trendminer: an architecture for real time analysis of social media text

  Daniel Preoţiuc-Pietro, Sina Samangooei, Trevor Cohn, Nick Gibbins, Mahesan Niranjan

  Workshop on Real-Time Analysis and Mining of Social Streams (RAMSS), ICWSM

  [Abstract]

  The emergence of online social networks (OSNs) and the accompanying availability of large amounts of data, pose a number of new natural language processing (NLP) and computational challenges. Data from OSNs is different to data from traditional sources (e.g. newswire). The texts are short, noisy and conversational. Another important issue is that data occurs in a real-time streams, needing immediate analysis that is grounded in time and context. In this paper we describe a new open-source framework for efficient text processing of streaming OSN data (available at www.trendminer-project.eu). Whilst researchers have made progress in adapting or creating text analysis tools for OSN data, a system to unify these tasks has yet to be built. Our system is focused on a real world scenario where fast processing and accuracy is paramount. We use the MapReduce framework for distributed computing and present running times for our system in order to show that scaling to online scenarios is feasible. We describe the components of the system and evaluate their accuracy. Our system supports easy integration of future modules in order to extend its functionality
• Unsupervised document zone identification using probabilistic graphical models

  Andrea Varga, Daniel Preoţiuc-Pietro, Fabio Ciravegna

  LREC

  [Abstract]

  Document zone identification aims to automatically classify sequences of text-spans (e.g. sentences) within a document into predefined zone categories. Current approaches to document zone identification mostly rely on supervised machine learning methods, which require a large amount of annotated data, which is often difficult and expensive to obtain. In order to overcome this bottleneck, we propose graphical models based on the popular Latent Dirichlet Allocation (LDA) model. The first model, which we call zoneLDA aims to cluster the sentences into zone classes using only unlabelled data. We also study an extension of zoneLDA called zoneLDAb, which makes distinction between common words and non-common words within the different zone types. We present results on two different domains: the scientific domain and the technical domain. For the latter one we propose a new document zone classification schema, which has been annotated over a collection of 689 documents, achieving a Kappa score of 85%. Overall our experiments show promising results for both of the domains, outperforming the baseline model. Furthermore, on the technical domain the performance of the models are comparable to the supervised approach using the same feature sets. We thus believe that graphical models are a promising avenue of research for automatic document zoning.
• Regression models of trends in streaming data

  Sina Samangooei, Daniel Preoţiuc-Pietro, Jing Li, Mahesan Niranjan, Nick Gibbins, Mahesan Niranjan

  Public Deliverable for Trendminer Project

  [Abstract]

  The aim of work package 3 is to develop machine learning methods capable of modelling real-valued time series based on contemporaneous textual input. The text is derived from social media, initially using the streaming feed of public Twitter status messages, and the time series are derived from the two use cases of the project: political polling and financial markets. This first deliverable includes our initial work in constructing tools and basic machine learning components for developing such models. The central component of the deliverable is the processing pipeline which takes a stream of Twitter status messages (‘tweets’), to which it applies a series of language processing algorithms – namely tokenization, language detection, word stemming and sentiment analysis. The processed data is then aggregated over a time-period to produce features suitable for describing movements in a time-series, e.g., word frequencies or aggregate sentiment relating to a given person, party or company. The sheer scale of Twitter data complicates even this simple level of processing, necessitating a parallel architecture in order to perform these steps in a real-time setting. For this reason we have developed our pipeline for feature processing and machine learning in the Apache Hadoop framework, an open source implementation of Google’s MapReduce parallel architecture.
• Gamma Knife surgery for pediatric arteriovenous malformations: a 25-year retrospective study

  Eduard B. Dinca, Patricia de Lacy, John Yianni, Jeremy Rowe, Matthias W. R. Radatz, Daniel Preoţiuc-Pietro, Andras A. Kemeny

  Journal of Neurosurgery: Pediatrics

  [Abstract]

  OBJECT
  The authors present their 25-year experience in treating pediatric arteriovenous malformations (AVMs) to allow comparisons with other historic studies and data in adults.

  METHODS
  Data were collected from a prospectively maintained departmental database selected for age and supplemented by case note review and telephone interviews as appropriate.

  RESULTS
  Three hundred sixty-three patients, ages 1–16 years (mean ± SD, 12 ± 3.2 years), underwent 410 treatments; 4 had planned 2-stage treatments and 43 were retreated subsequent to an initial partial response. Fifty-eight percent received general anesthesia for the procedure. Sixteen percent had previously undergone embolization. The most common presenting symptoms were as follows: hemorrhage (80.2%), epilepsy (8.3%; overall seizure prevalence 19.9%), and migrainous headaches (6.3%). Only 0.28% of the AVMs were incidental findings. The mean lesion volume was 3.75 ± 5.3 cm3 (range 0.01–32.8 cm3), with a median Spetzler-Martin grade of III (range I–V). The mean peripheral (therapeutic) dose was 22.7 ± 2.3 Gy (range 15–25 Gy), corresponding to a mean maximum dose of 43.6 ± 6 Gy (range 25–51.4 Gy).

  The obliteration rate was 71.3% in patients who received one treatment and 62.5% for retreated patients, with a mean obliteration time of 32.4 and 79.6 months, respectively. The overall obliteration rate was 82.7%. No follow-up data are as yet available for the 4 patients who underwent the staged treatments. Only 4 patients received peripheral doses below 20 Gy, and the AVM was obliterated in 3 of these patients. The other patients received 20, 22.5, or 25 Gy and had obliteration rates of 82.6%, 77.7%, and 86.3%, respectively. The bleeding rate postradiosurgery was 2.2%, and the cumulative complication rate was 3.6%, with radionecrosis being the most common complication (1.1%).

  CONCLUSIONS
  Surprisingly, there was no correlation (p = 0.43) between outcome and radiosurgical dose when that dose was between 20 and 25 Gy, thus suggesting that the lower of these 2 doses may be effective. Radiosurgery for pediatric AVM is safe and effective.

• Survival and complications following Gamma Knife radiosurgery or enucleation for ocular melanoma: a 20-year experience

  Eduard Dinca, John Yianni, Jeremy Rowe, Matthias Radatz, Daniel Preoţiuc-Pietro, Paul Rundle, Ian Rennie, Andras Kemeny

  Acta Neurochirurgica

  [Abstract]

  Background

  We present our experience in treating ocular melanoma at the National Centre for Stereotactic Radiosurgery in Sheffield, UK over the last 20 years.

  Method

  We analysed 170 patients treated with Gamma Knife radiosurgery, recorded the evolution of visual acuity and complication rates, and compared their survival with 620 patients treated with eye enucleation. Different peripheral doses (using the 50% therapeutic isodose) were employed: 50-70 Gy for 24 patients, 45 Gy for 71 patients, 35 Gy for 62 patients.

  Findings

  There was no significant difference in survival between the 35-Gy, 45-Gy and 50– to 70-Gy groups when compared between themselves (p = 0.168) and with the enucleation group (p = 0.454). The 5-year survival rates were: 64% for 35 Gy, 62.71% for 45 Gy, 63.6% for 50–70 Gy and 65.2% for enucleated patients. Clinical variables influencing survival for radiosurgery patients were tumour volume (p = 0.014) and location (median 66.4 vs 37.36 months for juxtapapillary vs peripheral tumours, respectively; p = 0.001), while age and gender did not prove significant. Regarding complications, using 35 Gy led to more than a 50% decrease, when compared with the 45-Gy dose, in the incidence of cataract, glaucoma and retinal detachment. Retinopathy, optic neuropathy and vitreous haemorrhage were not significantly influenced. Blindness decreased dramatically from 83.7% for 45 Gy to 31.4% for 35 Gy (p = 0.006), as well as post-radiosurgery enucleation: 23.9% for 45 Gy vs 6.45% for 35 Gy (p = 0.018). Visual acuity, recorded up to 5 years post-radiosurgery, was significantly better preserved for 35 Gy than for 45 Gy (p = 0.0003).

  Conclusions
  Using 35 Gy led to a dramatic decrease in complications, vision loss and salvage enucleation, while not compromising patient survival.

• Unsupervised word sense disambiguation with N-gram features

  Daniel Preoţiuc-Pietro, Florentina Hristea

  Artificial Intelligence Review

  [Abstract]

  The present paper concentrates on the issue of feature selection for unsupervised word sense disambiguation (WSD) performed with an underlying Naïve Bayes model. It introduces web N-gram features which, to our knowledge, are used for the first time in unsupervised WSD. While creating features from unlabeled data, we are “helping” a simple, basic knowledge-lean disambiguation algorithm to significantly increase its accuracy as a result of receiving easily obtainable knowledge. The performance of this method is compared to that of others that rely on completely different feature sets. Test results concerning nouns, adjectives and verbs show that web N-gram feature selection is a reliable alternative to previously existing approaches, provided that a “quality list” of features, adapted to the part of speech, is used.