Hiba Arnaout, Anmol Goel, H. Andrew Schwartz, Steffen T. Eberhardt, Dana Atzil-Slonim, Gavin Doherty, Brian Schwartz, Wolfgang Lutz, Tim Althoff, Munmun De Choudhury, Hamidreza Jamalabadi, Raj Sanjay Shah, Flor Miriam Plaza-del-Arco, Dirk Hovy, Maria Liakata, Iryna Gurevych. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Adithya V Ganesan, Vasudha Varadarajan, Oscar Kjell, Whitney Ringwald, Scott M. Feltman, Benjamin J. Luft, Roman Kotov, Ryan L. Boyd, H. Andrew Schwartz. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Nikita Soni, Dhruv Vijay Kunjadiya, Pratham Piyush Shah, Dikshya Mohanty, H. Andrew Schwartz, Niranjan Balasubramanian. Proceedings of the 30th Conference on Computational Natural Language Learning. 2026.

Siddharth Mangalik, Adithya V Ganesan, Abigail B. Wheeler, Nicholas Kerry, Jeremy D. W. Clifton, H. Schwartz, Ryan L. Boyd. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Current speech encoding pipelines often rely on an additional text-based LM to get robust representations of human communication, even though SotA speech-to-text models often have a LM within. This work proposes an approach to improve the LM within an audio model such that the subsequent text-LM is unnecessary. We introduce WhiSPA (Whisper with Semantic and Psychological Alignment), which leverages a novel audio training objective: contrastive loss with a language model embedding as a teacher. Using over 500k speech segments from mental health audio interviews, we evaluate the utility of aligning Whisper’s latent space with semantic representations from a text autoencoder (SBERT) and lexically derived embeddings of basic psychological dimensions: emotion and personality. Over self-supervised affective tasks and downstream psychological tasks, WhiSPA surpasses current speech encoders, achieving an average error reduction of 73.4% and 83.8%, respectively. WhiSPA demonstrates that it is not always necessary to run a subsequent text LM on speech-to-text output in order to get a rich psychological representation of human communication.

State-of-the-art automatic speech recognition (ASR) models like Whisper, perform poorly on atypical speech, such as that produced by individuals with dysarthria.Past works for atypical speech have mostly investigated fully personalized (or idiosyncratic) models, but modeling strategies that can both generalize and handle idiosyncracy could be more effective for capturing atypical speech.To investigate this, we compare four strategies: (a) normative models trained on typical speech (no personalization), (b) idiosyncratic models completely personalized to individuals, (c) dysarthric-normative models trained on other dysarthric speakers, and (d) dysarthric-idiosyncratic models which combine strategies by first modeling normative patterns before adapting to individual speech.In this case study, we find the dysarthricidiosyncratic model performs better than idiosyncratic approach while requiring less than half as much personalized data (36.43WER with 128 train size vs 36.99 with 256).Further, we found that tuning the speech encoder alone (as opposed to the LM decoder) yielded the best results reducing word error rate from 71% to 32% on average.Our findings highlight the value of leveraging both normative (crossspeaker) and idiosyncratic (speaker-specific) patterns to improve ASR for underrepresented speech populations. 1 * Equal contribution 1 Github: VishnuRaja98/Dysarthric-Speech-Transcription

The proliferation of ideological movements into extremist factions via social media has become a global concern. While radicalization has been studied extensively within the context of specific ideologies, our ability to accurately characterize extremism in more generalizable terms remains underdeveloped. In this paper, we propose a novel method for extracting and analyzing extremist discourse across a range of online ideological community forums. By focusing on verbal behavioral signatures of extremist traits, we develop a framework for quantifying extremism at both user and community levels. Our research identifies 11 distinct factors, which we term "The Extremist Eleven," as a generalized psychosocial model of extremism. Applying our method to various online communities, we demonstrate an ability to characterize ideologically diverse communities across the 11 extremist traits. We demonstrate the power of this method by analyzing user histories from members of the incel community. We find that our framework accurately predicts which users join the incel community up to 10 months before their actual entry with an AUC of > 0.6, steadily increasing to AUC ~ 0.9 three to four months before the event. Further, we find that upon entry into an ideological forum, the users tend to maintain their level of extremist traits within the community, while still remaining distinguishable from the general online discourse. Our findings contribute to the study of extremism by introducing a more holistic, cross-ideological approach that transcends traditional, trait-specific models.

Nikita Soni, H. Andrew Schwartz, João Sedoc, Niranjan Balasubramanian. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

model. While we found both methods to improve over non-adaptive baselines, We found ALIRT to be the most accurate and scalable, achieving the highest accuracy with fewer questions (e.g., Pearson r ≈ 0.93 after only 3 questions as compared to typically needing at least 7 questions). In general, adaptive language-based assessments of depression and anxiety were able to utilize a smaller sample of language without compromising validity or large computational costs.

Vasudha Varadarajan, Swanie Juhng, Syeda Mahwish, Xiaoran Liu, Jonah Luby, Christian Luhmann, H. Andrew Schwartz. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Adithya V Ganesan, Matthew Matero, Aravind Reddy Ravula, Huy Vu, H. Andrew Schwartz. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Lexical phenomena, such as clusters of words, disseminate through social networks at different rates but most models of diffusion focus on the discrete adoption of new lexical phenomena (i.e. new topics or memes). It is possible much of lexical diffusion happens via the changing rates of existing word categories or concepts (those that are already being used, at least to some extent, regularly) rather than new ones. In this study we introduce a new metric, contrastive lexical diffusion (CLD) coefficient, which attempts to measure the degree to which ordinary language (here clusters of common words) catch on over friendship connections over time. For instance topics related to meeting and job are found to be sticky, while negative thinking and emotion, and global events, like ‘school orientation’ were found to be less sticky even though they change rates over time. We evaluate CLD coefficient over both quantitative and qualitative tests, studied over 6 years of language on Twitter. We find CLD predicts the spread of tweets and friendship connections, scores converge with human judgments of lexical diffusion (r=0.92), and CLD coefficients replicate across disjoint networks (r=0.85). Comparing CLD scores can help understand lexical diffusion: positive emotion words appear more diffusive than negative emotions, first-person plurals (we) score higher than other pronouns, and numbers and time appear non-contagious.

Not all documents are equally important.Language processing is increasingly finding use as a supplement for questionnaires to assess psychological attributes of consenting individuals, but most approaches neglect to consider whether all documents of an individual are equally informative.In this paper, we present a novel model that uses message-level attention to learn the relative weight of users' social media posts for assessing their five factor personality traits.We demonstrate that models with message-level attention outperform those with word-level attention, and ultimately yield stateof-the-art accuracies for all five traits by using both word and message attention in combination with past approaches (an average increase in Pearson r of 2.5%).In addition, examination of the high-signal posts identified by our model provides insight into the relationship between language and personality, helping to inform future work.

An increasing number of natural language processing papers address the effect of bias on predictions, introducing mitigation techniques at different parts of the standard NLP pipeline (data and models).However, these works have been conducted individually, without a unifying framework to organize efforts within the field.This situation leads to repetitive approaches, and focuses overly on bias symptoms/effects, rather than on their origins, which could limit the development of effective countermeasures.In this paper, we propose a unifying predictive bias framework for NLP.We summarize the NLP literature and suggest general mathematical definitions of predictive bias.We differentiate two consequences of bias: outcome disparities and error disparities, as well as four potential origins of biases: label bias, selection bias, model overamplification, and semantic bias.Our framework serves as an overview of predictive bias in NLP, integrating existing work into a single structure, and providing a conceptual baseline for improved frameworks.

Human natural language is mentioned at a specific point in time while human emotions change over time.While much work has established a strong link between language use and emotional states, few have attempted to model emotional language in time.Here, we introduce the task of affective language forecasting -predicting future change in language based on past changes of language, a task with real-world applications such as treating mental health or forecasting trends in consumer confidence.We establish some of the fundamental autoregressive characteristics of the task (necessary history size, static versus dynamic length, varying time-step resolutions) and then build on popular sequence models for words to instead model sequences of language-based emotion in time.Over a novel Twitter dataset of 1,900 users and weekly + daily scores for 6 emotions and 2 additional linguistic attributes, we find a novel dual-sequence GRU model with decayed hidden states achieves best results (r = .66).We make our anonymized dataset as well as task setup and evaluation code available for others to build on.

Correcting errors in entered text is a common task but usually diffcult to perform on mobile devices due to tedious cursor navigation steps. In this paper, we present JustCorrect, an intelligent post hoc text correction technique for smartphones. To make a correction, the user simply types the correct text at the end of their current input, and JustCorrect will automatically detect the error and apply the correction in the form of an insertion or a substitution. In this way, manual navigation steps are bypassed, and the correction can be committed with a single tap. We solved two critical problems to support JustCorrect: (1) Correction Algorithm: we propose an algorithm that infers the user's correction intention from the last typed word. (2) Input Modalities: our study revealed that both tap and gesture were suitable input modalities for performing JustCorrect. Based on our fndings, we integrated JustCorrect into a soft keyboard. Our user studies show that using JustCorrect reduces the text correction time by 12.8% over the stock Android keyboard and by 9.7% over the "Type, then Correct" text correction technique by Zhang et al. (2019). Overall, JustCorrect complements existing post hoc text correction techniques, making error correction more automatic and intelligent.

While most mortality rates have decreased in the US, maternal mortality has increased and is among the highest of any OECD nation. Extensive public health research is ongoing to better understand the characteristics of communities with relatively high or low rates. In this work, we explore the role that social media language can play in providing insights into such community characteristics. Analyzing pregnancy-related tweets generated in US counties, we reveal a diverse set of latent topics including Morning Sickness, Celebrity Pregnancies, and Abortion Rights. We find that rates of mentioning these topics on Twitter predicts maternal mortality rates with higher accuracy than standard socioeconomic and risk variables such as income, race, and access to health-care, holding even after reducing the analysis to six topics chosen for their interpretability and connections to known risk factors. We then investigate psychological dimensions of community language, finding the use of less trustful, more stressed, and more negative affective language is significantly associated with higher mortality rates, while trust and negative affect also explain a significant portion of racial disparities in maternal mortality. We discuss the potential for these insights to inform actionable health interventions at the community-level.

Prior research has shown that TV content affects what people do on the web, particularly in the minutes after a TV ad airs, when online searches for the advertised product spike. To study this, researchers have typically focused on the total volume of search queries that include any and all keywords associated with the brands and products in TV ads. We argue that a granular consideration of search queries would be beneficial for two reasons. First, focusing on relevant keywords reduces measurement error, which can hinder identification of a significant effect from TV ads. Second, by grouping queries into themes based on semantic similarity, marketers can, for example, explore consumer intent in searches, or whether response manifests in queries with objective `value'. Our nuanced proposed approach considers query activity more broadly. Leveraging data from Bing and iSpotTV for 12 product campaigns, we first demonstrate that the outcome measure (i.e., response variable) can significantly influence conclusions about whether an ad has influenced search behavior. Second, we present a theme-based difference-in-differences method to determine the associations between TV ads and individual queries and query groups. We do this by first exploring the effects of TV ads on distinct, individual queries, and then aggregating queries into themes reflecting customer intentions, instead of aggregating all queries under a product or brand name umbrella. Our approach has implications for researchers who can improve measurement in search response; for marketers who can evaluate whether and how marketing messages resonate with consumers; and for sponsored search advertisers who can determine which keywords and queries they should bid on in the moments after their TV ads air. In addition, our method can be applied outside of the advertising context to understand how user generated content is impacted by events in general. Finally, we have developed a standalone tool that has been deployed to by Marketers internal to our company to generate reports to assess the success of their TV ad campaigns and has been deployed to our advertising customers by the Bing Ads Sales team to provide business intelligence11.

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 = .73to .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.

Individuals express their locus of control, or "control", in their language when they identify whether or not they are in control of their circumstances.Although control is a core concept underlying rhetorical style, it is not clear whether control is expressed by how or by what authors write.We explore the roles of syntax and semantics in expressing users' sense of control -i.e.being "controlled by" or "in control of" their circumstances-in a corpus of annotated Facebook posts.We present rich insights into these linguistic aspects and find that while the language signaling control is easy to identify, it is more challenging to label it is internally or externally controlled, with lexical features outperforming syntactic features at the task.Our findings could have important implications for studying selfexpression in social media.

Understanding causal explanations -reasons given for happenings in one's life -has been found to be an important psychological factor linked to physical and mental health.Causal explanations are often studied through manual identification of phrases over limited samples of personal writing.Automatic identification of causal explanations in social media, while challenging in relying on contextual and sequential cues, offers a larger-scale alternative to expensive manual ratings and opens the door for new applications (e.g.studying prevailing beliefs about causes, such as climate change).Here, we explore automating causal explanation analysis, building on discourse parsing, and presenting two novel subtasks: causality detection (determining whether a causal explanation exists at all) and causal explanation identification (identifying the specific phrase that is the explanation).We achieve strong accuracies for both tasks but find different approaches best: an SVM for causality prediction (F 1 = 0.791) and a hierarchy of Bidirectional LSTMs for causal explanation identification (F 1 = 0.853).Finally, we explore applications of our complete pipeline (F 1 = 0.868), showing demographic differences in mentions of causal explanation and that the association between a word and sentiment can change when it is used within a causal explanation.

Predictive models over social media language have shown promise in capturing community outcomes, but approaches thus far largely neglect the socio-demographic context (e.g.age, education rates, race) of the community from which the language originates.For example, it may be inaccurate to assume people in Mobile, Alabama, where the population is relatively older, will use words the same way as those from San Francisco, where the median age is younger with a higher rate of college education.In this paper, we present residualized factor adaptation, a novel approach to community prediction tasks which both (a) effectively integrates community attributes, as well as (b) adapts linguistic features to community attributes (factors).We use eleven demographic and socioeconomic attributes, and evaluate our approach over five different community-level predictive tasks, spanning health (heart disease mortality, percent fair/poor health), psychology (life satisfaction), and economics (percent housing price increase, foreclosure rate).Our evaluation shows that residualized factor adaptation significantly improves 4 out of 5 community-level outcome predictions over prior state-of-the-art for incorporating sociodemographic contexts.

We pose the general task of user-factor adaptation -adapting supervised learning models to real-valued user factors inferred from a background of their language, reflecting the idea that a piece of text should be understood within the context of the user that wrote it.We introduce a continuous adaptation technique, suited for real-valued user factors that are common in social science and bringing us closer to personalized NLP, adapting to each user uniquely.We apply this technique with known user factors including age, gender, and personality traits, as well as latent factors, evaluating over five tasks: POS tagging, PP-attachment, sentiment analysis, sarcasm detection, and stance detection.Adaptation provides statistically significant benefits for 3 of the 5 tasks: up to +1.2 points for PP-attachment, +3.4 points for sarcasm, and +3.0 points for stance.

Recommendations are often rated for their subjective quality, but few researchers have studied quality in terms of objective utility.We explore quality assessment with respect to both subjective (i.e.users' ratings) and objective (i.e., did it influence?did it improve decisions?)metrics in a massive online geopolitical forecasting system, ultimately comparing linguistic characteristics of each quality metric.Using a variety of features, we predict all types of quality with better accuracy than the simple yet strong baseline of recommendation length.For example, more complex sentence constructions, as evidenced by subordinate conjunctions, are characteristic of recommendations leading to objective improvements in forecasting.Our analyses also reveal rater biases; for example, forecasters are subjectively biased in favor of recommendations mentioning business deals and material things, even though such recommendations do not indeed prove any more useful objectively.

Several studies have demonstrated how language models of user attributes, such as personality, can be built by using the Facebook language of social media users in conjunction with their responses to psychology questionnaires. It is challenging to apply these models to make general predictions about attributes of communities, such as personality distributions across US counties, because it requires 1. the potentially inavailability of the original training data because of privacy and ethical regulations, 2. adapting Facebook language models to Twitter language without retraining the model, and 3. adapting from users to county-level collections of tweets. We propose a two-step algorithm, Target Side Domain Adaptation (TSDA) for such domain adaptation when no labeled Twitter/county data is available. TSDA corrects for the different word distributions between Facebook and Twitter and for the varying word distributions across counties by adjusting target side word frequencies; no changes to the trained model are made. In the case of predicting the Big Five county-level personality traits, TSDA outperforms a state-of-the-art domain adaptation method, gives county-level predictions that have fewer extreme outliers, higher year-to-year stability, and higher correlation with county-level outcomes.

Psychological analysis of language has repeatedly shown that an individual's rate of mentioning 1st person singular pronouns predicts a wealth of important demographic and psychological factors.However, these analyses are performed out of context -syntactic and semantic -which may change the magnitude or even direction of such relationships.In this paper, we put "pronouns in their context", exploring the relationship between self-reference and age, gender, and depression depending on syntactic position and verbal governor.We find that pronouns are overall more predictive when taking dependency relations and verb semantic categories into account, and, the direction of the relationship can change depending on the semantic class of the verbal governor.

Laura Smith, Salvatore Giorgi, Rishi Solanki, Johannes Eichstaedt, H. Andrew Schwartz, Muhammad Abdul-Mageed, Anneke Buffone, Lyle Ungar. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016.

H. Andrew Schwartz, Gregory Park, Maarten Sap, Evan Weingarten, Johannes Eichstaedt, Margaret Kern, David Stillwell, Michal Kosinski, Jonah Berger, Martin Seligman, Lyle Ungar. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2015.

Maarten Sap, Gregory Park, Johannes Eichstaedt, Margaret Kern, David Stillwell, Michal Kosinski, Lyle Ungar, Hansen Andrew Schwartz. Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2014.

This paper describes a linguistic text mining tool for analyzing problem reports in aerospace engineering and safety organizations. The Semantic Trend Analysis Tool (STAT) helps analysts find and review recurrences, similarities and trends in problem reports. The tool is being used to analyze engineering discrepancy reports at NASA Johnson Space Center. The tool has been augmented with a statistical natural language parser that also resolves parsing gaps and identifies verb arguments and adjuncts. The tool uses an aerospace ontology augmented with features of taxonomies and thesauruses. The ontology defines hierarchies of problem types, equipment types and function types. STAT uses the output of the parser and the aerospace ontology to identify words and phrases in problem report descriptions that refer to types of hazards, equipment damage, performance deviations or functional impairments. Tool performance has been evaluated on 120 problem descriptions from problem reports, with encouraging results.