Canada

Emotions are experienced and expressed differently across the world. In order to use Large Language Models (LMs) for multilingual tasks that require emotional sensitivity, LMs must reflect this cultural variation in emotion. We present our investigation on whether the widely-used multilingual LMs in 2023 reflect differences in emotional expressions across cultures and languages. We find that embeddings obtained from LMs (e.g., XLM-RoBERTa) are Anglocentric, and generative LMs (e.g., ChatGPT) reflect Western norms, even when responding to prompts in other languages. Our results show that multilingual LMs do not successfully learn the culturally appropriate nuances of emotion and we highlight possible research directions towards correcting this.

ACL 2023

Multilingual Language Models are not Multicultural: A Case Study in Emotion

cultural variation

emotion

computational social science

language models

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

workshop paper

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
 The conference will be held in Toronto, Canada, July 9-14, 2023. 
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience. 
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

As the e-commerce market continues to expand and online transactions proliferate, customer reviews have emerged as a critical element in shaping the purchasing decisions of prospective buyers. Previous studies have endeavored to identify key aspects of customer reviews through the development of sentiment analysis models and topic models. However, extracting specific dissatisfaction factors remains a challenging task. In this study, we delineate the pain point detection problem and propose Painsight, an unsupervised framework for automatically extracting distinct dissatisfaction factors from customer reviews without relying on ground truth labels. Painsight employs pre-trained language models to construct sentiment analysis and topic models, leveraging attribution scores derived from model gradients to extract dissatisfaction factors. Upon application of the proposed methodology to customer review data spanning five product categories, we successfully identified and categorized dissatisfaction factors within each group, as well as isolated factors for each type. Notably, Painsight outperformed benchmark methods, achieving substantial performance enhancements and exceptional results in human evaluations.

Painsight: An Extendable Opinion Mining Framework for Detecting Pain Points Based on Online Customer Reviews

Emotion Recognition in Conversations (ERC) has been gaining increasing
importance as conversational agents become more and more common. Recognizing
emotions is key for effective communication, being a crucial component in the
development of effective and empathetic conversational agents. Knowledge and
understanding of the conversational context are extremely valuable for
identifying the emotions of the interlocutor. We thus approach Emotion
Recognition in Conversations leveraging the conversational context, i.e.,
taking into attention previous conversational turns. The usual approach to
model the conversational context has been to produce context-independent
representations of each utterance and subsequently perform contextual modeling
of these. Here we propose context-dependent embedding representations of each
utterance by leveraging the contextual representational power of pre-trained
transformer language models. In our approach, we feed the conversational
context appended to the utterance to be classified as input to the RoBERTa
encoder, to which we append a simple classification module, thus discarding the
need to deal with context after obtaining the embeddings since these constitute
already an efficient representation of such context. We also investigate how
the number of introduced conversational turns influences our model performance.
The effectiveness of our approach is validated on the open-domain DailyDialog
dataset and on the task-oriented EmoWOZ dataset.

Context-Dependent Embedding Utterance Representations for Emotion Recognition in Conversations

Social media provide a rich source of data that can be mined and used for a wide variety of research purposes. However, annotating this data can be expensive, yet necessary for state-of-the-art pre-trained language models to achieve high prediction performance. Therefore, we combine pool-based active learning based on prediction uncertainty (an established method for reducing annotation costs) with unsupervised task adaptation through Masked Language Modeling (MLM). The results on three different datasets (two social media corpora, one benchmark dataset) show that task adaptation significantly improves results and that with only a fraction of the available training data, this approach reaches similar F1-scores as those achieved by an upper-bound baseline model fine-tuned on all training data. We hereby contribute to the scarce corpus of research on active learning with pre-trained language models and propose a cost-efficient annotation sampling and fine-tuning approach that can be applied to a wide variety of tasks and datasets.

Combining Active Learning and Task Adaptation with BERT for Cost-Effective Annotation of Social Media Datasets

In this paper, we leverage the GPT-3.5 language model both using the Chat-GPT API interface and the GPT-3.5 API interface to generate realistic examples of anti-vaccination tweets in Dutch with the aim of augmenting an imbalanced multi-label vaccine hesitancy argumentation classification dataset. In line with previous research, we devise a prompt that, on the one hand, instructs the model to generate realistic examples based on the gold standard dataset and, on the other hand, to assign multiple pseudo-labels (or a single pseudo-label) to the generated instances. We then augment our gold standard data with the generated examples and evaluate the impact thereof in a cross-validation setting with several state-of-the-art Dutch large language models. This augmentation technique predominantly shows improvements in F1 for classifying underrepresented classes while increasing the overall recall, paired with a slight decrease in precision for more common classes. Furthermore, we examine how well the synthetic data generalises to human data in the classification task. To our knowledge, we are the first to utilise Chat-GPT and GPT-3.5 for augmenting a Dutch multi-label dataset classification task.

Improving Dutch Vaccine Hesitancy Monitoring via Multi-Label Data Augmentation with GPT-3.5

This paper introduces the first emotion-annotated dataset for the Dari variant of Persian spoken in Afghanistan. The LetHerLearn dataset contains 7,600 tweets posted in reaction to the Taliban’s ban of women’s rights to education in 2022 and has been manually annotated according to Ekman’s emotion categories. We here detail the data collection and annotation process, present relevant dataset statistics as well as initial experiments on the resulting dataset, benchmarking a number of different neural architectures for the task of Dari emotion classification.

Emotion Analysis of Tweets Banning Education in Afghanistan

The prevalence of hate speech on online platforms has become a pressing concern for society, leading to increased attention towards detecting hate speech. Prior work in this area has primarily focused on identifying hate speech at the utterance level that reflects the complex nature of hate speech. In this paper, we propose a targeted and efficient approach to identifying hate speech by detecting slurs at the lexical level using contextualized word embeddings. We hypothesize that slurs have a systematically different representation than their neutral counterparts, making them identifiable through existing methods for discovering semantic dimensions in word embeddings. The results demonstrate the effectiveness of our approach in predicting slurs, confirming linguistic theory that the meaning of slurs is stable across contexts. Our robust hate dimension approach for slur identification offers a promising solution to tackle a smaller yet crucial piece of the complex puzzle of hate speech detection.

Identifying Slurs and Lexical Hate Speech via Light-Weight Dimension Projection in Embedding Space

The problem of subjectivity detection is often approached as a preparatory binary task for sentiment analysis, despite the fact that theoretically subjectivity is often defined as a matter of degree. In this work, we approach subjectivity analysis as a regression task and test the efficiency of a transformer RoBERTa model in annotating subjectivity of online news, including news from social media, based on a small subset of human-labeled training data. The results of experiments comparing our model to an existing rule-based subjectivity regressor and a state-of-the-art binary classifier reveal that: 1) our model highly correlates with the human subjectivity ratings and outperforms the widely used rule-based "pattern" subjectivity regressor (De Smedt and Daelemans, 2012); 2) our model performs well as a binary classifier and generalizes to the benchmark subjectivity dataset (Pang and Lee, 2004); 3) in contrast, state-of-the-art classifiers trained on the benchmark dataset show catastrophic performance on our human-labeled data. The results bring to light the issues of the gold standard subjectivity dataset, and the models trained on it, which seem to distinguish between the origin/style of the texts rather than subjectivity as perceived by human English speakers.

Analyzing Subjectivity Using a Transformer-Based Regressor Trained on Naïve Speakers’ Judgements

In this presentation, we present the result of our research, where we investigate potential bias in fine-tuned transformer models for irony detection. Bias is defined in this research as spurious associations between word n-grams and class labels, that can cause the system to rely too much on superficial cues and miss the essence of the irony. For this purpose, we looked for correlations between class labels and words that are prone to trigger irony, such as positive adjectives, intensifiers and topical nouns. Additionally, we investigate our irony model’s predictions before and after manipulating the data set through irony trigger replacements. We further support these insights with state-of-the-art explainability techniques (Layer Integrated Gradients, Discretized Integrated Gradients and Layer-wise Relevance Propagation). Both approaches confirm the hypothesis that transformer models generally encode correlations between positive sentiments and ironic texts, with even higher correlations between vividly expressed sentiment and irony. Based on these insights, we implemented a number of modification strategies to enhance the robustness of our irony classifier.

A Fine Line Between Irony and Sincerity: Identifying Bias in Transformer Models for Irony Detection

Humor is a central aspect of human communication that has not been solved for artificial agents so far. Large language models (LLMs) are increasingly able to capture implicit and contextual information. Especially, OpenAI’s ChatGPT recently gained immense public attention. The GPT3-based model almost seems to communicate on a human level and can even tell jokes. Humor is an essential component of human communication. But is ChatGPT really funny?
We put ChatGPT’s sense of humor to the test. In a series of exploratory experiments around jokes, i.e., generation, explanation, and detection, we seek to understand ChatGPT’s capability to grasp and reproduce human humor. Since the model itself is not accessible, we applied prompt-based experiments. 
Our empirical evidence indicates that jokes are not hard-coded but mostly also not newly generated by the model. Over 90% of 1008 generated jokes were the same 25 Jokes. The system accurately explains valid jokes but also comes up with fictional explanations for invalid jokes. Joke-typical characteristics can mislead ChatGPT in the classification of jokes. ChatGPT has not solved computational humor yet but it
can be a big leap toward “funny” machines.

ChatGPT is fun, but it is not funny! Humor is still challenging Large Language Models

Recent advances in the development of large Pretrained Language Models, such as GPT, BERT and Bloom, have achieved remarkable performance on a wide range of different NLP tasks. However, when used for text generation tasks, these models still have limitations when it comes to controlling the content and style of the generated text, often producing content that is incorrect, irrelevant, or inappropriate in the context of a given task. In this survey paper, we explore methods for controllable text generation with a focus on sentiment control. We systematically collect papers from the ACL Anthology, create a categorisation scheme based on different control techniques and controlled attributes, and use the scheme to categorise and compare methods. The result is a detailed and comprehensive overview of state-of-the-art techniques for sentiment-controlled text generation categorised on the basis of how the control is implemented and what attributes are controlled and providing a clear idea of their relative strengths and weaknesses.

Downloads

Next from ACL 2023

Painsight: An Extendable Opinion Mining Framework for Detecting Pain Points Based on Online Customer Reviews

Similar lecture

A Survey of Multilingual Reasoning in Language Models

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from ACL 2023

Painsight: An Extendable Opinion Mining Framework for Detecting Pain Points Based on Online Customer Reviews

Similar lecture

A Survey of Multilingual Reasoning in Language Models

Downloads