Emotion corpora are typically sampled based on keyword/hashtag search or by asking study participants to generate textual instances. In any case, these corpora are not uniform samples representing the entirety of a domain. We hypothesize that this practice of data acquision leads to unrealistic correlations between overrepresented topics in these corpora that harm the generalizability of models. Such topic bias could lead to wrong predictions for instances like “I organized the service for my aunt’s funeral.” when funeral events are overpresented for instances labeled with sadness, despite the emotion of pride being more appropriate here. In this paper, we study this topic bias both from the data and the modeling perspective. We first label a set of emotion corpora automatically via topic modeling and show that emotions in fact correlate with specific topics. Further, we see that emotion classifiers are confounded by such topics. Finally, we show that the established debiasing method of adversarial correction via gradient reversal mitigates the issue. Our work points out issues with existing emotion corpora and that more representative resources are required for fair evaluation of models predicting affective concepts from text.
Distorted science communication harms individuals and society as it can lead to unhealthy behavior change and decrease trust in scientific institutions. Given the rapidly increasing volume of science communication in recent years, a fine-grained understanding of how findings from scientific publications are reported to the general public, and methods to detect distortions from the original work automatically, are crucial. Prior work focused on individual aspects of distortions or worked with unpaired data. In this work, we make three foundational contributions towards addressing this problem: (1) annotating 1,600 instances of scientific findings from academic papers paired with corresponding findings as reported in news articles and tweets wrt. four characteristics: causality, certainty, generality and sensationalism; (2) establishing baselines for automatically detecting these characteristics; and (3) analyzing the prevalence of changes in these characteristics in both human-annotated and large-scale unlabeled data. Our results show that scientific findings frequently undergo subtle distortions when reported. Tweets distort findings more often than science news reports. Detecting fine-grained distortions automatically poses a challenging task. In our experiments, fine-tuned task-specific models consistently outperform few-shot LLM prompting.
There is a mismatch between psychological and computational studies on emotions. Psychological research aims at explaining and documenting internal mechanisms of these phenomena, while computational work often simplifies them into labels. Many emotion fundamentals remain under-explored in natural language processing, particularly how emotions develop and how people cope with them. To help reduce this gap, we follow theories on coping, and treat emotions as strategies to cope with salient situations (i.e., how people deal with emotion-eliciting events). This approach allows us to investigate the link between emotions and behavior, which also emerges in language. We introduce the task of coping identification, together with a corpus to do so, constructed via role-playing. We find that coping strategies realize in text even though they are challenging to recognize, both for humans and automatic systems trained and prompted on the same task. We thus open up a promising research direction to enhance the capability of models to better capture emotion mechanisms from text.
The statement “The earth is flat” is factually inaccurate, but if someone truly believes and argues in its favor, it is not deceptive. Research on deception detection and fact checking often conflates factual accuracy with the truthfulness of statements. This assumption makes it difficult to (a) study subtle distinctions and interactions between the two and (b) gauge their effects on downstream tasks. The belief-based deception framework disentangles these properties by defining texts as deceptive when there is a mismatch between what people say and what they truly believe. In this study, we assess if presumed patterns of deception generalize to German language texts. We test the effectiveness of computational models in detecting deception using an established corpus of belief-based argumentation. Finally, we gauge the impact of deception on the downstream task of fact checking and explore if this property confounds verification models. Surprisingly, our analysis finds no correlation with established cues of deception. Previous work claimed that computational models can outperform humans in deception detection accuracy, however, our experiments show that both traditional and state-of-the-art models struggle with the task, performing no better than random guessing. For fact checking, we find that natural language inference-based verification performs worse on non-factual and deceptive content, while prompting large language models for the same task is less sensitive to these properties.
Labeling corpora constitutes a bottleneck to create models for new tasks or domains. Large language models mitigate the issue with automatic corpus labeling methods, particularly for categorical annotations. Some NLP tasks such as emotion intensity prediction, however, require text regression, but there is no work on automating annotations for continuous label assignments. Regression is considered more challenging than classification: The fact that humans perform worse when tasked to choose values from a rating scale lead to comparative annotation methods, including best–worst scaling. This raises the question if large language model-based annotation methods show similar patterns, namely that they perform worse on rating scale annotation tasks than on comparative annotation tasks. To study this, we automate emotion intensity predictions and compare direct rating scale predictions, pairwise comparisons and best–worst scaling. We find that the latter shows the highest reliability. A transformer regressor fine-tuned on these data performs nearly on par with a model trained on the original manual annotations.
Verifying biomedical claims fails if no evidence can be discovered. In these cases, the fact-checking verdict remains unknown and the claim is unverifiable. To improve this situation, we have to understand if there are any claim properties that impact its verifiability. In this work we assume that entities and relations define the core variables in a biomedical claim’s anatomy and analyze if their properties help us to differentiate verifiable from unverifiable claims. In a study with trained annotation experts we prompt them to find evidence for biomedical claims, and observe how they refine search queries for their evidence search. This leads to the first corpus for scientific fact verification annotated with subject–relation–object triplets, evidence documents, and fact-checking verdicts (the BEAR-FACT corpus). We find (1) that discovering evidence for negated claims (e.g., X–does-not-cause–Y) is particularly challenging. Further, we see that annotators process queries mostly by adding constraints to the search and by normalizing entities to canonical names. (2) We compare our in-house annotations with a small crowdsourcing setting where we employ both medical experts and laypeople. We find that domain expertise does not have a substantial effect on the reliability of annotations. Finally, (3), we demonstrate that it is possible to reliably estimate the success of evidence retrieval purely from the claim text (.82F1), whereas identifying unverifiable claims proves more challenging (.27F1)
Many individuals affected by Social Anxiety Disorder turn to social media platforms to share their experiences and seek advice. This includes discussing the potential benefits of engaging with outdoor environments. As part of #SMM4H 2024, Shared Task 3 focuses on classifying the effects of outdoor spaces on social anxiety symptoms in Reddit posts. In our contribution to the task, we explore the effectiveness of domain-specific models (trained on social media data – SocBERT) against general domain models (trained on diverse datasets – BERT, RoBERTa, GPT-3.5) in predicting the sentiment related to outdoor spaces. Further, we assess the benefits of augmenting sparse human-labeled data with synthetic training instances and evaluate the complementary strengths of domain-specific and general classifiers using an ensemble model. Our results show that (1) fine-tuning small, domain-specific models generally outperforms large general language models in most cases. Only one large language model (GPT-4) exhibits performance comparable to the fine-tuned models (52% F1). Further, we find that (2) synthetic data does improve the performance of fine-tuned models in some cases, and (3) models do not appear to complement each other in our ensemble setup.
Corpora that are the fundament for toxicity detection contain such expressions typically directed against a target individual or group, e.g., people of a specific gender or ethnicity. Prior work has shown that the target identity mention can constitute a confounding variable. As an example, a model might learn that Christians are always mentioned in the context of hate speech. This misguided focus can lead to a limited generalization to newly emerging targets that are not found in the training data. In this paper, we hypothesize and subsequently show that this issue can be mitigated by considering targets on different levels of specificity. We distinguish levels of (1) the existence of a target, (2) a class (e.g., that the target is a religious group), or (3) a specific target group (e.g., Christians or Muslims). We define a target label hierarchy based on these three levels and then exploit this hierarchy in an adversarial correction for the lowest level (i.e. (3)) while maintaining some basic target features. This approach does not lower the toxicity detection performance but increases the generalization to targets not being available at training time.
In sentiment analysis of longer texts, there may be a variety of topics discussed, of entities mentioned, and of sentiments expressed regarding each entity. We find a lack of studies exploring how such texts express their sentiment towards each entity of interest, and how these sentiments can be modelled. In order to better understand how sentiment regarding persons and organizations (each entity in our scope) is expressed in longer texts, we have collected a dataset of expert annotations where the overall sentiment regarding each entity is identified, together with the sentence-level sentiment for these entities separately. We show that the reader’s perceived sentiment regarding an entity often differs from an arithmetic aggregation of sentiments at the sentence level. Only 70% of the positive and 55% of the negative entities receive a correct overall sentiment label when we aggregate the (human-annotated) sentiment labels for the sentences where the entity is mentioned. Our dataset reveals the complexity of entity-specific sentiment in longer texts, and allows for more precise modelling and evaluation of such sentiment expressions.
If a person firmly believes in a non-factual statement, such as “The Earth is flat”, and argues in its favor, there is no inherent intention to deceive. As the argumentation stems from genuine belief, it may be unlikely to exhibit the linguistic properties associated with deception or lying. This interplay of factuality, personal belief, and intent to deceive remains an understudied area. Disentangling the influence of these variables in argumentation is crucial to gain a better understanding of the linguistic properties attributed to each of them. To study the relation between deception and factuality, based on belief, we present the DeFaBel corpus, a crowd-sourced resource of belief-based deception. To create this corpus, we devise a study in which participants are instructed to write arguments supporting statements like “eating watermelon seeds can cause indigestion”, regardless of its factual accuracy or their personal beliefs about the statement. In addition to the generation task, we ask them to disclose their belief about the statement. The collected instances are labelled as deceptive if the arguments are in contradiction to the participants’ personal beliefs. Each instance in the corpus is thus annotated (or implicitly labelled) with personal beliefs of the author, factuality of the statement, and the intended deceptiveness. The DeFaBel corpus contains 1031 texts in German, out of which 643 are deceptive and 388 are non-deceptive. It is the first publicly available corpus for studying deception in German. In our analysis, we find that people are more confident in the persuasiveness of their arguments when the statement is aligned with their belief, but surprisingly less confident when they are generating arguments in favor of facts. The DeFaBel corpus can be obtained from https://www.ims.uni-stuttgart.de/data/defabel .
Emotion analysis often involves the categorization of isolated textual units, but these are parts of longer discourses, like dialogues or stories. This leads to two different established emotion classification setups: (1) Classification of a longer text into one or multiple emotion categories. (2) Classification of the parts of a longer text (sentences or utterances), either (2a) with or (2b) without consideration of the context. None of these settings, does, however, enable to answer the question which emotion is presumably experienced at a specific moment in time. For instance, a customer’s request of “My computer broke.” would be annotated with anger. This emotion persists in a potential follow-up reply “It is out of warranty.” which would also correspond to the global emotion label. An alternative reply “We will send you a new one.” might, in contrast, lead to relief. Modeling these label relations requires classification of textual parts under consideration of the past, but without access to the future. Consequently, we propose a novel annotation setup for emotion categorization corpora, in which the annotations reflect the emotion up to the annotated sentence. We ensure this by uncovering the textual parts step-by-step to the annotator, asking for a label in each step. This perspective is important to understand the final, global emotion, while having access to the individual sentence’s emotion contributions to this final emotion. In modeling experiments, we use these data to check if the context is indeed required to automatically predict such cumulative emotion progressions.
The term emotion analysis in text subsumes various natural language processing tasks which have in common the goal to enable computers to understand emotions. Most popular is emotion classification in which one or multiple emotions are assigned to a predefined textual unit. While such setting is appropriate for identifying the reader’s or author’s emotion, emotion role labeling adds the perspective of mentioned entities and extracts text spans that correspond to the emotion cause. The underlying emotion theories agree on one important point; that an emotion is caused by some internal or external event and comprises several subcomponents, including the subjective feeling and a cognitive evaluation. We therefore argue that emotions and events are related in two ways. (1) Emotions are events; and this perspective is the fundament in natural language processing for emotion role labeling. (2) Emotions are caused by events; a perspective that is made explicit with research how to incorporate psychological appraisal theories in NLP models to interpret events. These two research directions, role labeling and (event-focused) emotion classification, have by and large been tackled separately. In this paper, we contextualize both perspectives and discuss open research questions.
Existing fact-checking models for biomedical claims are typically trained on synthetic or well-worded data and hardly transfer to social media content. This mismatch can be mitigated by adapting the social media input to mimic the focused nature of common training claims. To do so, Wührl and Klinger (2022a) propose to extract concise claims based on medical entities in the text. However, their study has two limitations: First, it relies on gold-annotated entities. Therefore, its feasibility for a real-world application cannot be assessed since this requires detecting relevant entities automatically. Second, they represent claim entities with the original tokens. This constitutes a terminology mismatch which potentially limits the fact-checking performance. To understand both challenges, we propose a claim extraction pipeline for medical tweets that incorporates named entity recognition and terminology normalization via entity linking. We show that automatic NER does lead to a performance drop in comparison to using gold annotations but the fact-checking performance still improves considerably over inputting the unchanged tweets. Normalizing entities to their canonical forms does, however, not improve the performance.
Emotion analysis in text is an area of research that encompasses a set of various natural language processing (NLP) tasks, including classification and regression settings, as well as structured prediction tasks like role labelling or stimulus detection. In this tutorial, we provide an overview of research from emotion psychology which sets the ground for choosing adequate NLP methodology, and present existing resources and classification methods used for emotion analysis in texts. We further discuss appraisal theories and how events can be interpreted regarding their presumably caused emotion and briefly introduce emotion role labelling. In addition to these technical topics, we discuss the use cases of emotion analysis in text, their societal impact, ethical considerations, as well as the main challenges in the field.
The most prominent tasks in emotion analysis are to assign emotions to texts and to understand how emotions manifest in language. An important observation for natural language processing is that emotions can be communicated implicitly by referring to events alone, appealing to an empathetic, intersubjective understanding of events, even without explicitly mentioning an emotion name. In psychology, the class of emotion theories known as appraisal theories aims at explaining the link between events and emotions. Appraisals can be formalized as variables that measure a cognitive evaluation by people living through an event that they consider relevant. They include the assessment if an event is novel, if the person considers themselves to be responsible, if it is in line with their own goals, and so forth. Such appraisals explain which emotions are developed based on an event, for example, that a novel situation can induce surprise or one with uncertain consequences could evoke fear. We analyze the suitability of appraisal theories for emotion analysis in text with the goal of understanding if appraisal concepts can reliably be reconstructed by annotators, if they can be predicted by text classifiers, and if appraisal concepts help to identify emotion categories. To achieve that, we compile a corpus by asking people to textually describe events that triggered particular emotions and to disclose their appraisals. Then, we ask readers to reconstruct emotions and appraisals from the text. This set-up allows us to measure if emotions and appraisals can be recovered purely from text and provides a human baseline to judge a model’s performance measures. Our comparison of text classification methods to human annotators shows that both can reliably detect emotions and appraisals with similar performance. Therefore, appraisals constitute an alternative computational emotion analysis paradigm and further improve the categorization of emotions in text with joint models.
Emotions, which are responses to salient events, can be realized in text implicitly, for instance with mere references to facts (e.g., “That was the beginning of a long war”). Interpreting affective meanings thus relies on the readers background knowledge, but that is hardly modeled in computational emotion analysis. Much work in the field is focused on the word level and treats individual lexical units as the fundamental emotion cues in written communication. We shift our attention to word relations. We leverage Frame Semantics, a prominent theory for the description of predicate-argument structures, which matches the study of emotions: frames build on a “semantics of understanding” whose assumptions rely precisely on peoples world knowledge. Our overarching question is whether and to what extent the events that are represented by frames possess an emotion meaning. To carry out a large corpus-based correspondence analysis, we automatically annotate texts with emotions as well as with FrameNet frames and roles, and we analyze the correlations between them. Our main finding is that substantial groups of frames have an emotional import. With an extensive qualitative analysis, we show that they capture several properties of emotions that are purported by theories from psychology. These observations boost insights on the two strands of research that we bring together: emotion analysis can profit from the event-based perspective of frame semantics; in return, frame semantics gains a better grip of its position vis-à-vis emotions, an integral part of word meanings.
People differ fundamentally in what motivates them to pursue a goal and how they approach it. For instance, some people seek growth and show eagerness, whereas others prefer security and are vigilant. The concept of regulatory focus is employed in psychology, to explain and predict this goal-directed behavior of humans underpinned by two unique motivational systems – the promotion and the prevention system. Traditionally, text analysis methods using closed-vocabularies are employed to assess the distinctive linguistic patterns associated with the two systems. From an NLP perspective, automatically detecting the regulatory focus of individuals from text provides valuable insights into the behavioral inclinations of the author, finding its applications in areas like marketing or health communication. However, the concept never made an impactful debut in computational linguistics research. To bridge this gap we introduce the novel task of regulatory focus classification from text and present two complementary German datasets – (1) experimentally generated event descriptions and (2) manually annotated short social media texts used for evaluating the generalizability of models on real-world data. First, we conduct a correlation analysis to verify if the linguistic footprints of regulatory focus reported in psychology studies are observable and to what extent in our datasets. For automatic classification, we compare closed-vocabulary-based analyses with a state-of-the-art BERT-based text classification model and observe that the latter outperforms lexicon-based approaches on experimental data and is notably better on out-of-domain Twitter data.
Verbal deception has been studied in psychology, forensics, and computational linguistics for a variety of reasons, like understanding behaviour patterns, identifying false testimonies, and detecting deception in online communication. Varying motivations across research fields lead to differences in the domain choices to study and in the conceptualization of deception, making it hard to compare models and build robust deception detection systems for a given language. With this paper, we improve this situation by surveying available English deception datasets which include domains like social media reviews, court testimonials, opinion statements on specific topics, and deceptive dialogues from online strategy games. We consolidate these datasets into a single unified corpus. Based on this resource, we conduct a correlation analysis of linguistic cues of deception across datasets to understand the differences and perform cross-corpus modeling experiments which show that a cross-domain generalization is challenging to achieve. The unified deception corpus (UNIDECOR) can be obtained from https://www.ims.uni-stuttgart.de/data/unidecor.
Models for affective text generation have shown a remarkable progress, but they commonly rely only on basic emotion theories or valance/arousal values as conditions. This is appropriate when the goal is to create explicit emotion statements (“The kid is happy.”). Emotions are, however, commonly communicated implicitly. For instance, the emotional interpretation of an event (“Their dog died.”) does often not require an explicit emotion statement. In psychology, appraisal theories explain the link between a cognitive evaluation of an event and the potentially developed emotion. They put the assessment of the situation on the spot, for instance regarding the own control or the responsibility for what happens. We hypothesize and subsequently show that including appraisal variables as conditions in a generation framework comes with two advantages. (1) The generation model is informed in greater detail about what makes a specific emotion and what properties it has. This leads to text generation that better fulfills the condition. (2) The variables of appraisal allow a user to perform a more fine-grained control of the generated text, by stating properties of a situation instead of only providing the emotion category. Our Bart and T5-based experiments with 7 emotions (Anger, Disgust, Fear, Guilt, Joy, Sadness, Shame), and 7 appraisals (Attention, Responsibility, Control, Circumstance, Pleasantness, Effort, Certainty) show that (1) adding appraisals during training improves the accurateness of the generated texts by 10 pp in F1. Further, (2) the texts with appraisal variables are longer and contain more details. This exemplifies the greater control for users.
Conditional natural language generation methods often require either expensive fine-tuning or training a large language model from scratch. Both are unlikely to lead to good results without a substantial amount of data and computational resources. Prompt learning without changing the parameters of a large language model presents a promising alternative. It is a cost-effective approach, while still achieving competitive results. While this procedure is now established for zero- and few-shot text classification and structured prediction, it has received limited attention in conditional text generation. We present the first automatic prompt optimization approach for emotion-conditioned text generation with instruction-fine-tuned models. Our method uses an iterative optimization procedure that changes the prompt by adding, removing, or replacing tokens. As objective function, we only require a text classifier that measures the realization of the conditional variable in the generated text. We evaluate the method on emotion-conditioned text generation with a focus on event reports and compare it to manually designed prompts that also act as the seed for the optimization procedure. The optimized prompts achieve 0.75 macro-average F1 to fulfill the emotion condition in contrast to manually designed seed prompts with only 0.22 macro-average F1.
The task of natural language inference (NLI), to decide if a hypothesis entails or contradicts a premise, received considerable attention in recent years. All competitive systems build on top of contextualized representations and make use of transformer architectures for learning an NLI model. When somebody is faced with a particular NLI task, they need to select the best model that is available. This is a time-consuming and resource-intense endeavour. To solve this practical problem, we propose a simple method for predicting the performance without actually fine-tuning the model. We do this by testing how well the pre-trained models perform on the aNLI task when just comparing sentence embeddings with cosine similarity to what kind of performance is achieved when training a classifier on top of these embeddings. We show that the accuracy of the cosine similarity approach correlates strongly with the accuracy of the classification approach with a Pearson correlation coefficient of 0.65. Since the similarity is orders of magnitude faster to compute on a given dataset (less than a minute vs. hours), our method can lead to significant time savings in the process of model selection.
Emotion classification in NLP assigns emotions to texts, such as sentences or paragraphs. With texts like “I felt guilty when he cried”, focusing on the sentence level disregards the standpoint of each participant in the situation: the writer (“I”) and the other entity (“he”) could in fact have different affective states. The emotions of different entities have been considered only partially in emotion semantic role labeling, a task that relates semantic roles to emotion cue words. Proposing a related task, we narrow the focus on the experiencers of events, and assign an emotion (if any holds) to each of them. To this end, we represent each emotion both categorically and with appraisal variables, as a psychological access to explaining why a person develops a particular emotion. On an event description corpus, our experiencer-aware models of emotions and appraisals outperform the experiencer-agnostic baselines, showing that disregarding event participants is an oversimplification for the emotion detection task.
During the first two years of the COVID-19 pandemic, large volumes of biomedical information concerning this new disease have been published on social media. Some of this information can pose a real danger, particularly when false information is shared, for instance recommendations how to treat diseases without professional medical advice. Therefore, automatic fact-checking resources and systems developed specifically for medical domain are crucial. While existing fact-checking resources cover COVID-19 related information in news or quantify the amount of misinformation in tweets, there is no dataset providing fact-checked COVID-19 related Twitter posts with detailed annotations for biomedical entities, relations and relevant evidence. We contribute CoVERT, a fact-checked corpus of tweets with a focus on the domain of biomedicine and COVID-19 related (mis)information. The corpus consists of 300 tweets, each annotated with named entities and relations. We employ a novel crowdsourcing methodology to annotate all tweets with fact-checking labels and supporting evidence, which crowdworkers search for online. This methodology results in substantial inter-annotator agreement. Furthermore, we use the retrieved evidence extracts as part of a fact-checking pipeline, finding that the real-world evidence is more useful than the knowledge directly available in pretrained language models.
Emotion classification is often formulated as the task to categorize texts into a predefined set of emotion classes. So far, this task has been the recognition of the emotion of writers and readers, as well as that of entities mentioned in the text. We argue that a classification setup for emotion analysis should be performed in an integrated manner, including the different semantic roles that participate in an emotion episode. Based on appraisal theories in psychology, which treat emotions as reactions to events, we compile an English corpus of written event descriptions. The descriptions depict emotion-eliciting circumstances, and they contain mentions of people who responded emotionally. We annotate all experiencers, including the original author, with the emotions they likely felt. In addition, we link them to the event they found salient (which can be different for different experiencers in a text) by annotating event properties, or appraisals (e.g., the perceived event undesirability, the uncertainty of its outcome). Our analysis reveals patterns in the co-occurrence of people’s emotions in interaction. Hence, this richly-annotated resource provides useful data to study emotions and event evaluations from the perspective of different roles, and it enables the development of experiencer-specific emotion and appraisal classification systems.
Text mining and information extraction for the medical domain has focused on scientific text generated by researchers. However, their access to individual patient experiences or patient-doctor interactions is limited. On social media, doctors, patients and their relatives also discuss medical information. Individual information provided by laypeople complements the knowledge available in scientific text. It reflects the patient’s journey making the value of this type of data twofold: It offers direct access to people’s perspectives, and it might cover information that is not available elsewhere, including self-treatment or self-diagnose. Named entity recognition and relation extraction are methods to structure information that is available in unstructured text. However, existing medical social media corpora focused on a comparably small set of entities and relations. In contrast, we provide rich annotation layers to model patients’ experiences in detail. The corpus consists of medical tweets annotated with a fine-grained set of medical entities and relations between them, namely 14 entity (incl. environmental factors, diagnostics, biochemical processes, patients’ quality-of-life descriptions, pathogens, medical conditions, and treatments) and 20 relation classes (incl. prevents, influences, interactions, causes). The dataset consists of 2,100 tweets with approx. 6,000 entities and 2,200 relations.
False medical information on social media poses harm to people’s health. While the need for biomedical fact-checking has been recognized in recent years, user-generated medical content has received comparably little attention. At the same time, models for other text genres might not be reusable, because the claims they have been trained with are substantially different. For instance, claims in the SciFact dataset are short and focused: “Side effects associated with antidepressants increases risk of stroke”. In contrast, social media holds naturally-occurring claims, often embedded in additional context: "‘If you take antidepressants like SSRIs, you could be at risk of a condition called serotonin syndrome’ Serotonin syndrome nearly killed me in 2010. Had symptoms of stroke and seizure.” This showcases the mismatch between real-world medical claims and the input that existing fact-checking systems expect. To make user-generated content checkable by existing models, we propose to reformulate the social-media input in such a way that the resulting claim mimics the claim characteristics in established datasets. To accomplish this, our method condenses the claim with the help of relational entity information and either compiles the claim out of an entity-relation-entity triple or extracts the shortest phrase that contains these elements. We show that the reformulated input improves the performance of various fact-checking models as opposed to checking the tweet text in its entirety.
Authors of posts in social media communicate their emotions and what causes them with text and images. While there is work on emotion and stimulus detection for each modality separately, it is yet unknown if the modalities contain complementary emotion information in social media. We aim at filling this research gap and contribute a novel, annotated corpus of English multimodal Reddit posts. On this resource, we develop models to automatically detect the relation between image and text, an emotion stimulus category and the emotion class. We evaluate if these tasks require both modalities and find for the image–text relations, that text alone is sufficient for most categories (complementary, illustrative, opposing): the information in the text allows to predict if an image is required for emotion understanding. The emotions of anger and sadness are best predicted with a multimodal model, while text alone is sufficient for disgust, joy, and surprise. Stimuli depicted by objects, animals, food, or a person are best predicted by image-only models, while multimodal mod- els are most effective on art, events, memes, places, or screenshots.
People associate affective meanings to words - “death” is scary and sad while “party” is connotated with surprise and joy. This raises the question if the association is purely a product of the learned affective imports inherent to semantic meanings, or is also an effect of other features of words, e.g., morphological and phonological patterns. We approach this question with an annotation-based analysis leveraging nonsense words. Specifically, we conduct a best-worst scaling crowdsourcing study in which participants assign intensity scores for joy, sadness, anger, disgust, fear, and surprise to 272 non-sense words and, for comparison of the results to previous work, to 68 real words. Based on this resource, we develop character-level and phonology-based intensity regressors. We evaluate them on both nonsense words and real words (making use of the NRC emotion intensity lexicon of 7493 words), across six emotion categories. The analysis of our data reveals that some phonetic patterns show clear differences between emotion intensities. For instance, s as a first phoneme contributes to joy, sh to surprise, p as last phoneme more to disgust than to anger and fear. In the modelling experiments, a regressor trained on real words from the NRC emotion intensity lexicon shows a higher performance (r = 0.17) than regressors that aim at learning the emotion connotation purely from nonsense words. We conclude that humans do associate affective meaning to words based on surface patterns, but also based on similarities to existing words (“juy” to “joy”, or “flike” to “like”).
Machine-learned models for author profiling in social media often rely on data acquired via self-reporting-based psychometric tests (questionnaires) filled out by social media users. This is an expensive but accurate data collection strategy. Another, less costly alternative, which leads to potentially more noisy and biased data, is to rely on labels inferred from publicly available information in the profiles of the users, for instance self-reported diagnoses or test results. In this paper, we explore a third strategy, namely to directly use a corpus of items from validated psychometric tests as training data. Items from psychometric tests often consist of sentences from an I-perspective (e.g., ‘I make friends easily.’). Such corpora of test items constitute ‘small data’, but their availability for many concepts is a rich resource. We investigate this approach for personality profiling, and evaluate BERT classifiers fine-tuned on such psychometric test items for the big five personality traits (openness, conscientiousness, extraversion, agreeableness, neuroticism) and analyze various augmentation strategies regarding their potential to address the challenges coming with such a small corpus. Our evaluation on a publicly available Twitter corpus shows a comparable performance to in-domain training for 4/5 personality traits with T5-based data augmentation.
Within textual emotion classification, the set of relevant labels depends on the domain and application scenario and might not be known at the time of model development. This conflicts with the classical paradigm of supervised learning in which the labels need to be predefined. A solution to obtain a model with a flexible set of labels is to use the paradigm of zero-shot learning as a natural language inference task, which in addition adds the advantage of not needing any labeled training data. This raises the question how to prompt a natural language inference model for zero-shot learning emotion classification. Options for prompt formulations include the emotion name anger alone or the statement “This text expresses anger”. With this paper, we analyze how sensitive a natural language inference-based zero-shot-learning classifier is to such changes to the prompt under consideration of the corpus: How carefully does the prompt need to be selected? We perform experiments on an established set of emotion datasets presenting different language registers according to different sources (tweets, events, blogs) with three natural language inference models and show that indeed the choice of a particular prompt formulation needs to fit to the corpus. We show that this challenge can be tackled with combinations of multiple prompts. Such ensemble is more robust across corpora than individual prompts and shows nearly the same performance as the individual best prompt for a particular corpus.
Social media contains unfiltered and unique information, which is potentially of great value, but, in the case of misinformation, can also do great harm. With regards to biomedical topics, false information can be particularly dangerous. Methods of automatic fact-checking and fake news detection address this problem, but have not been applied to the biomedical domain in social media yet. We aim to fill this research gap and annotate a corpus of 1200 tweets for implicit and explicit biomedical claims (the latter also with span annotations for the claim phrase). With this corpus, which we sample to be related to COVID-19, measles, cystic fibrosis, and depression, we develop baseline models which detect tweets that contain a claim automatically. Our analyses reveal that biomedical tweets are densely populated with claims (45 % in a corpus sampled to contain 1200 tweets focused on the domains mentioned above). Baseline classification experiments with embedding-based classifiers and BERT-based transfer learning demonstrate that the detection is challenging, however, shows acceptable performance for the identification of explicit expressions of claims. Implicit claim tweets are more challenging to detect.
When humans judge the affective content of texts, they also implicitly assess the correctness of such judgment, that is, their confidence. We hypothesize that people’s (in)confidence that they performed well in an annotation task leads to (dis)agreements among each other. If this is true, confidence may serve as a diagnostic tool for systematic differences in annotations. To probe our assumption, we conduct a study on a subset of the Corpus of Contemporary American English, in which we ask raters to distinguish neutral sentences from emotion-bearing ones, while scoring the confidence of their answers. Confidence turns out to approximate inter-annotator disagreements. Further, we find that confidence is correlated to emotion intensity: perceiving stronger affect in text prompts annotators to more certain classification performances. This insight is relevant for modelling studies of intensity, as it opens the question wether automatic regressors or classifiers actually predict intensity, or rather human’s self-perceived confidence.
Appraisal theories explain how the cognitive evaluation of an event leads to a particular emotion. In contrast to theories of basic emotions or affect (valence/arousal), this theory has not received a lot of attention in natural language processing. Yet, in psychology it has been proven powerful: Smith and Ellsworth (1985) showed that the appraisal dimensions attention, certainty, anticipated effort, pleasantness, responsibility/control and situational control discriminate between (at least) 15 emotion classes. We study different annotation strategies for these dimensions, based on the event-focused enISEAR corpus (Troiano et al., 2019). We analyze two manual annotation settings: (1) showing the text to annotate while masking the experienced emotion label; (2) revealing the emotion associated with the text. Setting 2 enables the annotators to develop a more realistic intuition of the described event, while Setting 1 is a more standard annotation procedure, purely relying on text. We evaluate these strategies in two ways: by measuring inter-annotator agreement and by fine- tuning RoBERTa to predict appraisal variables. Our results show that knowledge of the emotion increases annotators’ reliability. Further, we evaluate a purely automatic rule-based labeling strategy (inferring appraisal from annotated emotion classes). Training on automatically assigned labels leads to a competitive performance of our classifier, even when tested on manual annotations. This is an indicator that it might be possible to automatically create appraisal corpora for every domain for which emotion corpora already exist.
The 2020 US Elections have been, more than ever before, characterized by social media campaigns and mutual accusations. We investigate in this paper if this manifests also in online communication of the supporters of the candidates Biden and Trump, by uttering hateful and offensive communication. We formulate an annotation task, in which we join the tasks of hateful/offensive speech detection and stance detection, and annotate 3000 Tweets from the campaign period, if they express a particular stance towards a candidate. Next to the established classes of favorable and against, we add mixed and neutral stances and also annotate if a candidate is mentioned with- out an opinion expression. Further, we an- notate if the tweet is written in an offensive style. This enables us to analyze if supporters of Joe Biden and the Democratic Party communicate differently than supporters of Donald Trump and the Republican Party. A BERT baseline classifier shows that the detection if somebody is a supporter of a candidate can be performed with high quality (.89 F1 for Trump and .91 F1 for Biden), while the detection that somebody expresses to be against a candidate is more challenging (.79 F1 and .64 F1, respectively). The automatic detection of hate/offensive speech remains challenging (with .53 F1). Our corpus is publicly available and constitutes a novel resource for computational modelling of offensive language under consideration of stances.
Emotion stimulus detection is the task of finding the cause of an emotion in a textual description, similar to target or aspect detection for sentiment analysis. Previous work approached this in three ways, namely (1) as text classification into an inventory of predefined possible stimuli (“Is the stimulus category A or B?”), (2) as sequence labeling of tokens (“Which tokens describe the stimulus?”), and (3) as clause classification (“Does this clause contain the emotion stimulus?”). So far, setting (3) has been evaluated broadly on Mandarin and (2) on English, but no comparison has been performed. Therefore, we analyze whether clause classification or token sequence labeling is better suited for emotion stimulus detection in English. We propose an integrated framework which enables us to evaluate the two different approaches comparably, implement models inspired by state-of-the-art approaches in Mandarin, and test them on four English data sets from different domains. Our results show that token sequence labeling is superior on three out of four datasets, in both clause-based and token sequence-based evaluation. The only case in which clause classification performs better is one data set with a high density of clause annotations. Our error analysis further confirms quantitatively and qualitatively that clauses are not the appropriate stimulus unit in English.
Automatic emotion categorization has been predominantly formulated as text classification in which textual units are assigned to an emotion from a predefined inventory, for instance following the fundamental emotion classes proposed by Paul Ekman (fear, joy, anger, disgust, sadness, surprise) or Robert Plutchik (adding trust, anticipation). This approach ignores existing psychological theories to some degree, which provide explanations regarding the perception of events. For instance, the description that somebody discovers a snake is associated with fear, based on the appraisal as being an unpleasant and non-controllable situation. This emotion reconstruction is even possible without having access to explicit reports of a subjective feeling (for instance expressing this with the words “I am afraid.”). Automatic classification approaches therefore need to learn properties of events as latent variables (for instance that the uncertainty and the mental or physical effort associated with the encounter of a snake leads to fear). With this paper, we propose to make such interpretations of events explicit, following theories of cognitive appraisal of events, and show their potential for emotion classification when being encoded in classification models. Our results show that high quality appraisal dimension assignments in event descriptions lead to an improvement in the classification of discrete emotion categories. We make our corpus of appraisal-annotated emotion-associated event descriptions publicly available.
Machine translation provides powerful methods to convert text between languages, and is therefore a technology enabling a multilingual world. An important part of communication, however, takes place at the non-propositional level (e.g., politeness, formality, emotions), and it is far from clear whether current MT methods properly translate this information. This paper investigates the specific hypothesis that the non-propositional level of emotions is at least partially lost in MT. We carry out a number of experiments in a back-translation setup and establish that (1) emotions are indeed partially lost during translation; (2) this tendency can be reversed almost completely with a simple re-ranking approach informed by an emotion classifier, taking advantage of diversity in the n-best list; (3) the re-ranking approach can also be applied to change emotions, obtaining a model for emotion style transfer. An in-depth qualitative analysis reveals that there are recurring linguistic changes through which emotions are toned down or amplified, such as change of modality.
Emotion recognition is predominantly formulated as text classification in which textual units are assigned to an emotion from a predefined inventory (e.g., fear, joy, anger, disgust, sadness, surprise, trust, anticipation). More recently, semantic role labeling approaches have been developed to extract structures from the text to answer questions like: “who is described to feel the emotion?” (experiencer), “what causes this emotion?” (stimulus), and at which entity is it directed?” (target). Though it has been shown that jointly modeling stimulus and emotion category prediction is beneficial for both subtasks, it remains unclear which of these semantic roles enables a classifier to infer the emotion. Is it the experiencer, because the identity of a person is biased towards a particular emotion (X is always happy)? Is it a particular target (everybody loves X) or a stimulus (doing X makes everybody sad)? We answer these questions by training emotion classification models on five available datasets annotated with at least one semantic role by masking the fillers of these roles in the text in a controlled manner and find that across multiple corpora, stimuli and targets carry emotion information, while the experiencer might be considered a confounder. Further, we analyze if informing the model about the position of the role improves the classification decision. Particularly on literature corpora we find that the role information improves the emotion classification.
We propose the task of emotion style transfer, which is particularly challenging, as emotions (here: anger, disgust, fear, joy, sadness, surprise) are on the fence between content and style. To understand the particular difficulties of this task, we design a transparent emotion style transfer pipeline based on three steps: (1) select the words that are promising to be substituted to change the emotion (with a brute-force approach and selection based on the attention mechanism of an emotion classifier), (2) find sets of words as candidates for substituting the words (based on lexical and distributional semantics), and (3) select the most promising combination of substitutions with an objective function which consists of components for content (based on BERT sentence embeddings), emotion (based on an emotion classifier), and fluency (based on a neural language model). This comparably straight-forward setup enables us to explore the task and understand in what cases lexical substitution can vary the emotional load of texts, how changes in content and style interact and if they are at odds. We further evaluate our pipeline quantitatively in an automated and an annotation study based on Tweets and find, indeed, that simultaneous adjustments of content and emotion are conflicting objectives: as we show in a qualitative analysis motivated by Scherer’s emotion component model, this is particularly the case for implicit emotion expressions based on cognitive appraisal or descriptions of bodily reactions.
Obituaries contain information about people’s values across times and cultures, which makes them a useful resource for exploring cultural history. They are typically structured similarly, with sections corresponding to Personal Information, Biographical Sketch, Characteristics, Family, Gratitude, Tribute, Funeral Information and Other aspects of the person. To make this information available for further studies, we propose a statistical model which recognizes these sections. To achieve that, we collect a corpus of 20058 English obituaries from TheDaily Item, Remembering.CA and The London Free Press. The evaluation of our annotation guidelines with three annotators on 1008 obituaries shows a substantial agreement of Fleiss κ = 0.87. Formulated as an automatic segmentation task, a convolutional neural network outperforms bag-of-words and embedding-based BiLSTMs and BiLSTM-CRFs with a micro F1 = 0.81.
Most research on emotion analysis from text focuses on the task of emotion classification or emotion intensity regression. Fewer works address emotions as a phenomenon to be tackled with structured learning, which can be explained by the lack of relevant datasets. We fill this gap by releasing a dataset of 5000 English news headlines annotated via crowdsourcing with their associated emotions, the corresponding emotion experiencers and textual cues, related emotion causes and targets, as well as the reader’s perception of the emotion of the headline. This annotation task is comparably challenging, given the large number of classes and roles to be identified. We therefore propose a multiphase annotation procedure in which we first find relevant instances with emotional content and then annotate the more fine-grained aspects. Finally, we develop a baseline for the task of automatic prediction of semantic role structures and discuss the results. The corpus we release enables further research on emotion classification, emotion intensity prediction, emotion cause detection, and supports further qualitative studies.
Most approaches to emotion analysis of social media, literature, news, and other domains focus exclusively on basic emotion categories as defined by Ekman or Plutchik. However, art (such as literature) enables engagement in a broader range of more complex and subtle emotions. These have been shown to also include mixed emotional responses. We consider emotions in poetry as they are elicited in the reader, rather than what is expressed in the text or intended by the author. Thus, we conceptualize a set of aesthetic emotions that are predictive of aesthetic appreciation in the reader, and allow the annotation of multiple labels per line to capture mixed emotions within their context. We evaluate this novel setting in an annotation experiment both with carefully trained experts and via crowdsourcing. Our annotation with experts leads to an acceptable agreement of k = .70, resulting in a consistent dataset for future large scale analysis. Finally, we conduct first emotion classification experiments based on BERT, showing that identifying aesthetic emotions is challenging in our data, with up to .52 F1-micro on the German subset. Data and resources are available at https://github.com/tnhaider/poetry-emotion.
Public companies are obliged to include financial and non-financial information within their cor- porate filings under Regulation S-K, in the United States (SEC, 2010). However, the requirements still allow for manager’s discretion. This raises the question to which extent the information is actually included and if this information is at all relevant for investors. We answer this question by training and evaluating an end-to-end deep learning approach (based on BERT and GloVe embeddings) to predict the financial and environmental performance of the company from the “Management’s Discussion and Analysis of Financial Conditions and Results of Operations” (MD&A) section of 10-K (yearly) and 10-Q (quarterly) filings. We further analyse the mediating effect of the environmental performance on the relationship between the company’s disclosures and financial performance. Hereby, we address the results of previous studies regarding environ- mental performance. We find that the textual information contained within the MD&A section does not allow for conclusions about the future (corporate) financial performance. However, there is evidence that the environmental performance can be extracted by natural language processing methods.
Span identification (in short, span ID) tasks such as chunking, NER, or code-switching detection, ask models to identify and classify relevant spans in a text. Despite being a staple of NLP, and sharing a common structure, there is little insight on how these tasks’ properties influence their difficulty, and thus little guidance on what model families work well on span ID tasks, and why. We analyze span ID tasks via performance prediction, estimating how well neural architectures do on different tasks. Our contributions are: (a) we identify key properties of span ID tasks that can inform performance prediction; (b) we carry out a large-scale experiment on English data, building a model to predict performance for unseen span ID tasks that can support architecture choices; (c), we investigate the parameters of the meta model, yielding new insights on how model and task properties interact to affect span ID performance. We find, e.g., that span frequency is especially important for LSTMs, and that CRFs help when spans are infrequent and boundaries non-distinctive.
Sentiment analysis has a range of corpora available across multiple languages. For emotion analysis, the situation is more limited, which hinders potential research on crosslingual modeling and the development of predictive models for other languages. In this paper, we fill this gap for German by constructing deISEAR, a corpus designed in analogy to the well-established English ISEAR emotion dataset. Motivated by Scherer’s appraisal theory, we implement a crowdsourcing experiment which consists of two steps. In step 1, participants create descriptions of emotional events for a given emotion. In step 2, five annotators assess the emotion expressed by the texts. We show that transferring an emotion classification model from the original English ISEAR to the German crowdsourced deISEAR via machine translation does not, on average, cause a performance drop.
The development of a fictional plot is centered around characters who closely interact with each other forming dynamic social networks. In literature analysis, such networks have mostly been analyzed without particular relation types or focusing on roles which the characters take with respect to each other. We argue that an important aspect for the analysis of stories and their development is the emotion between characters. In this paper, we combine these aspects into a unified framework to classify emotional relationships of fictional characters. We formalize it as a new task and describe the annotation of a corpus, based on fan-fiction short stories. The extraction pipeline which we propose consists of character identification (which we treat as given by an oracle here) and the relation classification. For the latter, we provide results using several approaches previously proposed for relation identification with neural methods. The best result of 0.45 F1 is achieved with a GRU with character position indicators on the task of predicting undirected emotion relations in the associated social network graph.
The automatic detection of satire vs. regular news is relevant for downstream applications (for instance, knowledge base population) and to improve the understanding of linguistic characteristics of satire. Recent approaches build upon corpora which have been labeled automatically based on article sources. We hypothesize that this encourages the models to learn characteristics for different publication sources (e.g., “The Onion” vs. “The Guardian”) rather than characteristics of satire, leading to poor generalization performance to unseen publication sources. We therefore propose a novel model for satire detection with an adversarial component to control for the confounding variable of publication source. On a large novel data set collected from German news (which we make available to the research community), we observe comparable satire classification performance and, as desired, a considerable drop in publication classification performance with adversarial training. Our analysis shows that the adversarial component is crucial for the model to learn to pay attention to linguistic properties of satire.
Adjective phrases like “a little bit surprised”, “completely shocked”, or “not stunned at all” are not handled properly by current state-of-the-art emotion classification and intensity prediction systems. Based on this finding, we analyze differences between embeddings used by these systems in regard to their capability of handling such cases and argue that intensifiers in context of emotion words need special treatment, as is established for sentiment polarity classification, but not for more fine-grained emotion prediction. To resolve this issue, we analyze different aspects of a post-processing pipeline which enriches the word representations of such phrases. This includes expansion of semantic spaces at the phrase level and sub-word level followed by retrofitting to emotion lexicons. We evaluate the impact of these steps with ‘A La Carte and Bag-of-Substrings extensions based on pretrained GloVe,Word2vec, and fastText embeddings against a crowd-sourced corpus of intensity annotations for tweets containing our focus phrases. We show that the fastText-based models do not gain from handling these specific phrases under inspection. For Word2vec embeddings, we show that our post-processing pipeline improves the results by up to 8% on a novel dataset densly populated with intensifiers while it does not decrease the performance on the established EmoInt dataset.
Centrality of emotion for the stories told by humans is underpinned by numerous studies in literature and psychology. The research in automatic storytelling has recently turned towards emotional storytelling, in which characters’ emotions play an important role in the plot development (Theune et al., 2004; y Perez, 2007; Mendez et al., 2016). However, these studies mainly use emotion to generate propositional statements in the form “A feels affection towards B” or “A confronts B”. At the same time, emotional behavior does not boil down to such propositional descriptions, as humans display complex and highly variable patterns in communicating their emotions, both verbally and non-verbally. In this paper, we analyze how emotions are expressed non-verbally in a corpus of fan fiction short stories. Our analysis shows that stories written by humans convey character emotions along various non-verbal channels. We find that some non-verbal channels, such as facial expressions and voice characteristics of the characters, are more strongly associated with joy, while gestures and body postures are more likely to occur with trust. Based on our analysis, we argue that automatic storytelling systems should take variability of emotion into account when generating descriptions of characters’ emotions.
Sentiment analysis in low-resource languages suffers from a lack of annotated corpora to estimate high-performing models. Machine translation and bilingual word embeddings provide some relief through cross-lingual sentiment approaches. However, they either require large amounts of parallel data or do not sufficiently capture sentiment information. We introduce Bilingual Sentiment Embeddings (BLSE), which jointly represent sentiment information in a source and target language. This model only requires a small bilingual lexicon, a source-language corpus annotated for sentiment, and monolingual word embeddings for each language. We perform experiments on three language combinations (Spanish, Catalan, Basque) for sentence-level cross-lingual sentiment classification and find that our model significantly outperforms state-of-the-art methods on four out of six experimental setups, as well as capturing complementary information to machine translation. Our analysis of the resulting embedding space provides evidence that it represents sentiment information in the resource-poor target language without any annotated data in that language.
Supervised machine learning algorithms require training data whose generation for complex relation extraction tasks tends to be difficult. Being optimized for relation extraction at sentence level, many annotation tools lack in facilitating the annotation of relational structures that are widely spread across the text. This leads to non-intuitive and cumbersome visualizations, making the annotation process unnecessarily time-consuming. We propose SANTO, an easy-to-use, domain-adaptive annotation tool specialized for complex slot filling tasks which may involve problems of cardinality and referential grounding. The web-based architecture enables fast and clearly structured annotation for multiple users in parallel. Relational structures are formulated as templates following the conceptualization of an underlying ontology. Further, import and export procedures of standard formats enable interoperability with external sources and tools.
Domain adaptation for sentiment analysis is challenging due to the fact that supervised classifiers are very sensitive to changes in domain. The two most prominent approaches to this problem are structural correspondence learning and autoencoders. However, they either require long training times or suffer greatly on highly divergent domains. Inspired by recent advances in cross-lingual sentiment analysis, we provide a novel perspective and cast the domain adaptation problem as an embedding projection task. Our model takes as input two mono-domain embedding spaces and learns to project them to a bi-domain space, which is jointly optimized to (1) project across domains and to (2) predict sentiment. We perform domain adaptation experiments on 20 source-target domain pairs for sentiment classification and report novel state-of-the-art results on 11 domain pairs, including the Amazon domain adaptation datasets and SemEval 2013 and 2016 datasets. Our analysis shows that our model performs comparably to state-of-the-art approaches on domains that are similar, while performing significantly better on highly divergent domains. Our code is available at https://github.com/jbarnesspain/domain_blse
Most approaches to emotion analysis in fictional texts focus on detecting the emotion expressed in text. We argue that this is a simplification which leads to an overgeneralized interpretation of the results, as it does not take into account who experiences an emotion and why. Emotions play a crucial role in the interaction between characters and the events they are involved in. Until today, no specific corpora that capture such an interaction were available for literature. We aim at filling this gap and present a publicly available corpus based on Project Gutenberg, REMAN (Relational EMotion ANnotation), manually annotated for spans which correspond to emotion trigger phrases and entities/events in the roles of experiencers, targets, and causes of the emotion. We provide baseline results for the automatic prediction of these relational structures and show that emotion lexicons are not able to encompass the high variability of emotion expressions and demonstrate that statistical models benefit from joint modeling of emotions with its roles in all subtasks. The corpus that we provide enables future research on the recognition of emotions and associated entities in text. It supports qualitative literary studies and digital humanities. The corpus is available at http://www.ims.uni-stuttgart.de/data/reman .
Several datasets have been annotated and published for classification of emotions. They differ in several ways: (1) the use of different annotation schemata (e. g., discrete label sets, including joy, anger, fear, or sadness or continuous values including valence, or arousal), (2) the domain, and, (3) the file formats. This leads to several research gaps: supervised models often only use a limited set of available resources. Additionally, no previous work has compared emotion corpora in a systematic manner. We aim at contributing to this situation with a survey of the datasets, and aggregate them in a common file format with a common annotation schema. Based on this aggregation, we perform the first cross-corpus classification experiments in the spirit of future research enabled by this paper, in order to gain insight and a better understanding of differences of models inferred from the data. This work also simplifies the choice of the most appropriate resources for developing a model for a novel domain. One result from our analysis is that a subset of corpora is better classified with models trained on a different corpus. For none of the corpora, training on all data altogether is better than using a subselection of the resources. Our unified corpus is available at http://www.ims.uni-stuttgart.de/data/unifyemotion.
Past shared tasks on emotions use data with both overt expressions of emotions (I am so happy to see you!) as well as subtle expressions where the emotions have to be inferred, for instance from event descriptions. Further, most datasets do not focus on the cause or the stimulus of the emotion. Here, for the first time, we propose a shared task where systems have to predict the emotions in a large automatically labeled dataset of tweets without access to words denoting emotions. Based on this intention, we call this the Implicit Emotion Shared Task (IEST) because the systems have to infer the emotion mostly from the context. Every tweet has an occurrence of an explicit emotion word that is masked. The tweets are collected in a manner such that they are likely to include a description of the cause of the emotion – the stimulus. Altogether, 30 teams submitted results which range from macro F1 scores of 21 % to 71 %. The baseline (Max-Ent bag of words and bigrams) obtains an F1 score of 60 % which was available to the participants during the development phase. A study with human annotators suggests that automatic methods outperform human predictions, possibly by honing into subtle textual clues not used by humans. Corpora, resources, and results are available at the shared task website at http://implicitemotions.wassa2018.com.
Most machine learning systems for natural language processing are tailored to specific tasks. As a result, comparability of models across tasks is missing and their applicability to new tasks is limited. This affects end users without machine learning experience as well as model developers. To address these limitations, we present DERE, a novel framework for declarative specification and compilation of template-based information extraction. It uses a generic specification language for the task and for data annotations in terms of spans and frames. This formalism enables the representation of a large variety of natural language processing challenges. The backend can be instantiated by different models, following different paradigms. The clear separation of frame specification and model backend will ease the implementation of new models and the evaluation of different models across different tasks. Furthermore, it simplifies transfer learning, joint learning across tasks and/or domains as well as the assessment of model generalizability. DERE is available as open-source software.
Literary genres are commonly viewed as being defined in terms of content and stylistic features. In this paper, we focus on one particular class of lexical features, namely emotion information, and investigate the hypothesis that emotion-related information correlates with particular genres. Using genre classification as a testbed, we compare a model that computes lexicon-based emotion scores globally for complete stories with a model that tracks emotion arcs through stories on a subset of Project Gutenberg with five genres. Our main findings are: (a), the global emotion model is competitive with a large-vocabulary bag-of-words genre classifier (80%F1); (b), the emotion arc model shows a lower performance (59 % F1) but shows complementary behavior to the global model, as indicated by a very good performance of an oracle model (94 % F1) and an improved performance of an ensemble model (84 % F1); (c), genres differ in the extent to which stories follow the same emotional arcs, with particularly uniform behavior for anger (mystery) and fear (adventures, romance, humor, science fiction).
There has been a good amount of progress in sentiment analysis over the past 10 years, including the proposal of new methods and the creation of benchmark datasets. In some papers, however, there is a tendency to compare models only on one or two datasets, either because of time restraints or because the model is tailored to a specific task. Accordingly, it is hard to understand how well a certain model generalizes across different tasks and datasets. In this paper, we contribute to this situation by comparing several models on six different benchmarks, which belong to different domains and additionally have different levels of granularity (binary, 3-class, 4-class and 5-class). We show that Bi-LSTMs perform well across datasets and that both LSTMs and Bi-LSTMs are particularly good at fine-grained sentiment tasks (i.e., with more than two classes). Incorporating sentiment information into word embeddings during training gives good results for datasets that are lexically similar to the training data. With our experiments, we contribute to a better understanding of the performance of different model architectures on different data sets. Consequently, we detect novel state-of-the-art results on the SenTube datasets.
There is a rich variety of data sets for sentiment analysis (viz., polarity and subjectivity classification). For the more challenging task of detecting discrete emotions following the definitions of Ekman and Plutchik, however, there are much fewer data sets, and notably no resources for the social media domain. This paper contributes to closing this gap by extending the SemEval 2016 stance and sentiment datasetwith emotion annotation. We (a) analyse annotation reliability and annotation merging; (b) investigate the relation between emotion annotation and the other annotation layers (stance, sentiment); (c) report modelling results as a baseline for future work.
Social media are used by an increasing number of political actors. A small subset of these is interested in pursuing extremist motives such as mobilization, recruiting or radicalization activities. In order to counteract these trends, online providers and state institutions reinforce their monitoring efforts, mostly relying on manual workflows. We propose a machine learning approach to support manual attempts towards identifying right-wing extremist content in German Twitter profiles. Based on a fine-grained conceptualization of right-wing extremism, we frame the task as ranking each individual profile on a continuum spanning different degrees of right-wing extremism, based on a nearest neighbour approach. A quantitative evaluation reveals that our ranking model yields robust performance (up to 0.81 F1 score) when being used for predicting discrete class labels. At the same time, the model provides plausible continuous ranking scores for a small sample of borderline cases at the division of right-wing extremism and New Right political movements.
Our submission to the WASSA-2017 shared task on the prediction of emotion intensity in tweets is a supervised learning method with extended lexicons of affective norms. We combine three main information sources in a random forrest regressor, namely (1), manually created resources, (2) automatically extended lexicons, and (3) the output of a neural network (CNN-LSTM) for sentence regression. All three feature sets perform similarly well in isolation (≈ .67 macro average Pearson correlation). The combination achieves .72 on the official test set (ranked 2nd out of 22 participants). Our analysis reveals that performance is increased by providing cross-emotional intensity predictions. The automatic extension of lexicon features benefit from domain specific embeddings. Complementary ratings for affective norms increase the impact of lexicon features. Our resources (ratings for 1.6 million twitter specific words) and our implementation is publicly available at http://www.ims.uni-stuttgart.de/data/ims_emoint.
Systems which build on top of information extraction are typically challenged to extract knowledge that, while correct, is not yet well-known. We hypothesize that a good confidence measure for relational information has the property that such interesting information is found between information extracted with very high confidence and very low confidence. We discuss confidence estimation for the domain of biomedical protein-protein relation discovery in biomedical literature. As facts reported in papers take some time to be validated and recorded in biomedical databases, such task gives rise to large quantities of unknown but potentially true candidate relations. It is thus important to rank them based on supporting evidence rather than discard them. In this paper, we discuss this task and propose different approaches for confidence estimation and a pipeline to evaluate such methods. We show that the most straight-forward approach, a combination of different confidence measures from pipeline modules seems not to work well. We discuss this negative result and pinpoint potential future research directions.
The automatic analysis of texts containing opinions of users about, e.g., products or political views has gained attention within the last decades. However, previous work on the task of analyzing user reviews about mobile applications in app stores is limited. Publicly available corpora do not exist, such that a comparison of different methods and models is difficult. We fill this gap by contributing the Sentiment Corpus of App Reviews (SCARE), which contains fine-grained annotations of application aspects, subjective (evaluative) phrases and relations between both. This corpus consists of 1,760 annotated application reviews from the Google Play Store with 2,487 aspects and 3,959 subjective phrases. We describe the process and methodology how the corpus was created. The Fleiss Kappa between four annotators reveals an agreement of 0.72. We provide a strong baseline with a linear-chain conditional random field and word-embedding features with a performance of 0.62 for aspect detection and 0.63 for the extraction of subjective phrases. The corpus is available to the research community to support the development of sentiment analysis methods on mobile application reviews.
Opinion mining has received wide attention in recent years. Models for this task are typically trained or evaluated with a manually annotated dataset. However, fine-grained annotation of sentiments including information about aspects and their evaluation is very labour-intensive. The data available so far is limited. Contributing to this situation, this paper describes the Bielefeld University Sentiment Analysis Corpus for German and English (USAGE), which we offer freely to the community and which contains the annotation of product reviews from Amazon with both aspects and subjective phrases. It provides information on segments in the text which denote an aspect or a subjective evaluative phrase which refers to the aspect. Relations and coreferences are explicitly annotated. This dataset contains 622 English and 611 German reviews, allowing to investigate how to port sentiment analysis systems across languages and domains. We describe the methodology how the corpus was created and provide statistics including inter-annotator agreement. We further provide figures for a baseline system and results for German and English as well as in a cross-domain setting. The results are encouraging in that they show that aspects and phrases can be extracted robustly without the need of tuning to a particular type of products.