Pushpak Bhattacharyya

Anupama Ray, Shubham Mishra, Apoorva Nunna et al.

While sentiment and emotion analysis have been studied extensively, the relationship between sarcasm and emotion has largely remained unexplored. A sarcastic expression may have a variety of underlying emotions. For example, "I love being ignored" belies sadness, while "my mobile is fabulous with a battery backup of only 15 minutes!" expresses frustration. Detecting the emotion behind a sarcastic expression is non-trivial yet an important task. We undertake the task of detecting the emotion in a sarcastic statement, which to the best of our knowledge, is hitherto unexplored. We start with the recently released multimodal sarcasm detection dataset (MUStARD) pre-annotated with 9 emotions. We identify and correct 343 incorrect emotion labels (out of 690). We double the size of the dataset, label it with emotions along with valence and arousal which are important indicators of emotional intensity. Finally, we label each sarcastic utterance with one of the four sarcasm types-Propositional, Embedded, Likeprefixed and Illocutionary, with the goal of advancing sarcasm detection research. Exhaustive experimentation with multimodal (text, audio, and video) fusion models establishes a benchmark for exact emotion recognition in sarcasm and outperforms the state-of-art sarcasm detection. We release the dataset enriched with various annotations and the code for research purposes: https://github.com/apoorva-nunna/MUStARD_Plus_Plus

Rudra Murthy, Pallab Bhattacharjee, Rahul Sharnagat et al.

Named Entity Recognition (NER) is a foundational NLP task that aims to provide class labels like Person, Location, Organisation, Time, and Number to words in free text. Named Entities can also be multi-word expressions where the additional I-O-B annotation information helps label them during the NER annotation process. While English and European languages have considerable annotated data for the NER task, Indian languages lack on that front -- both in terms of quantity and following annotation standards. This paper releases a significantly sized standard-abiding Hindi NER dataset containing 109,146 sentences and 2,220,856 tokens, annotated with 11 tags. We discuss the dataset statistics in all their essential detail and provide an in-depth analysis of the NER tag-set used with our data. The statistics of tag-set in our dataset show a healthy per-tag distribution, especially for prominent classes like Person, Location and Organisation. Since the proof of resource-effectiveness is in building models with the resource and testing the model on benchmark data and against the leader-board entries in shared tasks, we do the same with the aforesaid data. We use different language models to perform the sequence labelling task for NER and show the efficacy of our data by performing a comparative evaluation with models trained on another dataset available for the Hindi NER task. Our dataset helps achieve a weighted F1 score of 88.78 with all the tags and 92.22 when we collapse the tag-set, as discussed in the paper. To the best of our knowledge, no available dataset meets the standards of volume (amount) and variability (diversity), as far as Hindi NER is concerned. We fill this gap through this work, which we hope will significantly help NLP for Hindi. We release this dataset with our code and models at https://github.com/cfiltnlp/HiNER

Abhisek Tiwari, Anisha Saha, Sriparna Saha et al.

With the advancement of telemedicine, both researchers and medical practitioners are working hand-in-hand to develop various techniques to automate various medical operations, such as diagnosis report generation. In this paper, we first present a multi-modal clinical conversation summary generation task that takes a clinician-patient interaction (both textual and visual information) and generates a succinct synopsis of the conversation. We propose a knowledge-infused, multi-modal, multi-tasking medical domain identification and clinical conversation summary generation (MM-CliConSummation) framework. It leverages an adapter to infuse knowledge and visual features and unify the fused feature vector using a gated mechanism. Furthermore, we developed a multi-modal, multi-intent clinical conversation summarization corpus annotated with intent, symptom, and summary. The extensive set of experiments, both quantitatively and qualitatively, led to the following findings: (a) critical significance of visuals, (b) more precise and medical entity preserving summary with additional knowledge infusion, and (c) a correlation between medical department identification and clinical synopsis generation. Furthermore, the dataset and source code are available at https://github.com/NLP-RL/MM-CliConSummation.

Nihar Sahoo, Himanshu Gupta, Pushpak Bhattacharyya

With the rise of online hate speech, automatic detection of Hate Speech, Offensive texts as a natural language processing task is getting popular. However, very little research has been done to detect unintended social bias from these toxic language datasets. This paper introduces a new dataset ToxicBias curated from the existing dataset of Kaggle competition named "Jigsaw Unintended Bias in Toxicity Classification". We aim to detect social biases, their categories, and targeted groups. The dataset contains instances annotated for five different bias categories, viz., gender, race/ethnicity, religion, political, and LGBTQ. We train transformer-based models using our curated datasets and report baseline performance for bias identification, target generation, and bias implications. Model biases and their mitigation are also discussed in detail. Our study motivates a systematic extraction of social bias data from toxic language datasets. All the codes and dataset used for experiments in this work are publicly available

Kaveri Kale, pushpak Bhattacharyya, Kshitij Jadhav

Clinical practice frequently uses medical imaging for diagnosis and treatment. A significant challenge for automatic radiology report generation is that the radiology reports are long narratives consisting of multiple sentences for both abnormal and normal findings. Therefore, applying conventional image captioning approaches to generate the whole report proves to be insufficient, as these are designed to briefly describe images with short sentences. We propose a template-based approach to generate radiology reports from radiographs. Our approach involves the following: i) using a multilabel image classifier, produce the tags for the input radiograph; ii) using a transformer-based model, generate pathological descriptions (a description of abnormal findings seen on radiographs) from the tags generated in step (i); iii) using a BERT-based multi-label text classifier, find the spans in the normal report template to replace with the generated pathological descriptions; and iv) using a rule-based system, replace the identified span with the generated pathological description. We performed experiments with the two most popular radiology report datasets, IU Chest X-ray and MIMIC-CXR and demonstrated that the BLEU-1, ROUGE-L, METEOR, and CIDEr scores are better than the State-of-the-Art models by 25%, 36%, 44% and 48% respectively, on the IU X-RAY dataset. To the best of our knowledge, this is the first attempt to generate chest X-ray radiology reports by first creating small sentences for abnormal findings and then replacing them in the normal report template.

Vineet Bhat, Preethi Jyothi, Pushpak Bhattacharyya

Disfluency correction (DC) is the process of removing disfluent elements like fillers, repetitions and corrections from spoken utterances to create readable and interpretable text. DC is a vital post-processing step applied to Automatic Speech Recognition (ASR) outputs, before subsequent processing by downstream language understanding tasks. Existing DC research has primarily focused on English due to the unavailability of large-scale open-source datasets. Towards the goal of multilingual disfluency correction, we present a high-quality human-annotated DC corpus covering four important Indo-European languages: English, Hindi, German and French. We provide extensive analysis of results of state-of-the-art DC models across all four languages obtaining F1 scores of 97.55 (English), 94.29 (Hindi), 95.89 (German) and 92.97 (French). To demonstrate the benefits of DC on downstream tasks, we show that DC leads to 5.65 points increase in BLEU scores on average when used in conjunction with a state-of-the-art Machine Translation (MT) system. We release code to run our experiments along with our annotated dataset here.

Sandhya Singh, Prapti Roy, Nihar Sahoo et al.

Movies reflect society and also hold power to transform opinions. Social biases and stereotypes present in movies can cause extensive damage due to their reach. These biases are not always found to be the need of storyline but can creep in as the author's bias. Movie production houses would prefer to ascertain that the bias present in a script is the story's demand. Today, when deep learning models can give human-level accuracy in multiple tasks, having an AI solution to identify the biases present in the script at the writing stage can help them avoid the inconvenience of stalled release, lawsuits, etc. Since AI solutions are data intensive and there exists no domain specific data to address the problem of biases in scripts, we introduce a new dataset of movie scripts that are annotated for identity bias. The dataset contains dialogue turns annotated for (i) bias labels for seven categories, viz., gender, race/ethnicity, religion, age, occupation, LGBTQ, and other, which contains biases like body shaming, personality bias, etc. (ii) labels for sensitivity, stereotype, sentiment, emotion, emotion intensity, (iii) all labels annotated with context awareness, (iv) target groups and reason for bias labels and (v) expert-driven group-validation process for high quality annotations. We also report various baseline performances for bias identification and category detection on our dataset.

Gopendra Vikram Singh, Priyanshu Priya, Mauajama Firdaus et al.

The long-standing goal of Artificial Intelligence (AI) has been to create human-like conversational systems. Such systems should have the ability to develop an emotional connection with the users, hence emotion recognition in dialogues is an important task. Emotion detection in dialogues is a challenging task because humans usually convey multiple emotions with varying degrees of intensities in a single utterance. Moreover, emotion in an utterance of a dialogue may be dependent on previous utterances making the task more complex. Emotion recognition has always been in great demand. However, most of the existing datasets for multi-label emotion and intensity detection in conversations are in English. To this end, we create a large conversational dataset in Hindi named EmoInHindi for multi-label emotion and intensity recognition in conversations containing 1,814 dialogues with a total of 44,247 utterances. We prepare our dataset in a Wizard-of-Oz manner for mental health and legal counselling of crime victims. Each utterance of the dialogue is annotated with one or more emotion categories from the 16 emotion classes including neutral, and their corresponding intensity values. We further propose strong contextual baselines that can detect emotion(s) and the corresponding intensity of an utterance given the conversational context.

Akshay Batheja, Pushpak Bhattacharyya

In this paper, we show that the combination of Phrase Pair Injection and Corpus Filtering boosts the performance of Neural Machine Translation (NMT) systems. We extract parallel phrases and sentences from the pseudo-parallel corpus and augment it with the parallel corpus to train the NMT models. With the proposed approach, we observe an improvement in the Machine Translation (MT) system for 3 low-resource language pairs, Hindi-Marathi, English-Marathi, and English-Pashto, and 6 translation directions by up to 2.7 BLEU points, on the FLORES test data. These BLEU score improvements are over the models trained using the whole pseudo-parallel corpus augmented with the parallel corpus.

Ankush Agarwal, Raj Gite, Shreya Laddha et al.

In the commercial aviation domain, there are a large number of documents, like, accident reports (NTSB, ASRS) and regulatory directives (ADs). There is a need for a system to access these diverse repositories efficiently in order to service needs in the aviation industry, like maintenance, compliance, and safety. In this paper, we propose a Knowledge Graph (KG) guided Deep Learning (DL) based Question Answering (QA) system for aviation safety. We construct a Knowledge Graph from Aircraft Accident reports and contribute this resource to the community of researchers. The efficacy of this resource is tested and proved by the aforesaid QA system. Natural Language Queries constructed from the documents mentioned above are converted into SPARQL (the interface language of the RDF graph database) queries and answered. On the DL side, we have two different QA models: (i) BERT QA which is a pipeline of Passage Retrieval (Sentence-BERT based) and Question Answering (BERT based), and (ii) the recently released GPT-3. We evaluate our system on a set of queries created from the accident reports. Our combined QA system achieves 9.3% increase in accuracy over GPT-3 and 40.3% increase over BERT QA. Thus, we infer that KG-DL performs better than either singly.

Abisek Rajakumar Kalarani, Pushpak Bhattacharyya, Niyati Chhaya et al.

Well-formed context aware image captions and tags in enterprise content such as marketing material are critical to ensure their brand presence and content recall. Manual creation and updates to ensure the same is non trivial given the scale and the tedium towards this task. We propose a new unified Vision-Language (VL) model based on the One For All (OFA) model, with a focus on context-assisted image captioning where the caption is generated based on both the image and its context. Our approach aims to overcome the context-independent (image and text are treated independently) nature of the existing approaches. We exploit context by pretraining our model with datasets of three tasks: news image captioning where the news article is the context, contextual visual entailment, and keyword extraction from the context. The second pretraining task is a new VL task, and we construct and release two datasets for the task with 1.1M and 2.2K data instances. Our system achieves state-of-the-art results with an improvement of up to 8.34 CIDEr score on the benchmark news image captioning datasets. To the best of our knowledge, ours is the first effort at incorporating contextual information in pretraining the models for the VL tasks.

Kaveri Kale, Pushpak Bhattacharyya, Aditya Shetty et al.

Conventionally, the radiologist prepares the diagnosis notes and shares them with the transcriptionist. Then the transcriptionist prepares a preliminary formatted report referring to the notes, and finally, the radiologist reviews the report, corrects the errors, and signs off. This workflow causes significant delays and errors in the report. In current research work, we focus on applications of NLP techniques like Information Extraction (IE) and domain-specific Knowledge Graph (KG) to automatically generate radiology reports from radiologist's dictation. This paper focuses on KG construction for each organ by extracting information from an existing large corpus of free-text radiology reports. We develop an information extraction pipeline that combines rule-based, pattern-based, and dictionary-based techniques with lexical-semantic features to extract entities and relations. Missing information in short dictation can be accessed from the KGs to generate pathological descriptions and hence the radiology report. Generated pathological descriptions evaluated using semantic similarity metrics, which shows 97% similarity with gold standard pathological descriptions. Also, our analysis shows that our IE module is performing better than the OpenIE tool for the radiology domain. Furthermore, we include a manual qualitative analysis from radiologists, which shows that 80-85% of the generated reports are correctly written, and the remaining are partially correct.

Soumitra Ghosh, Asif Ekbal, Pushpak Bhattacharyya

The World Health Organization (WHO) has emphasized the importance of significantly accelerating suicide prevention efforts to fulfill the United Nations' Sustainable Development Goal (SDG) objective of 2030. In this paper, we present an end-to-end multitask system to address a novel task of detection of two interpersonal risk factors of suicide, Perceived Burdensomeness (PB) and Thwarted Belongingness (TB) from suicide notes. We also introduce a manually translated code-mixed suicide notes corpus, CoMCEASE-v2.0, based on the benchmark CEASE-v2.0 dataset, annotated with temporal orientation, PB and TB labels. We exploit the temporal orientation and emotion information in the suicide notes to boost overall performance. For comprehensive evaluation of our proposed method, we compare it to several state-of-the-art approaches on the existing CEASE-v2.0 dataset and the newly announced CoMCEASE-v2.0 dataset. Empirical evaluation suggests that temporal and emotional information can substantially improve the detection of PB and TB.

Ankush Agarwal, Sakharam Gawade, Amar Prakash Azad et al.

Large language models (LLMs) have demonstrated remarkable performance in a wide range of natural language tasks. However, as these models continue to grow in size, they face significant challenges in terms of computational costs. Additionally, LLMs often lack efficient domain-specific understanding, which is particularly crucial in specialized fields such as aviation and healthcare. To boost the domain-specific understanding, we propose, KITLM, a novel knowledge base integration approach into language model through relevant information infusion. By integrating pertinent knowledge, not only the performance of the language model is greatly enhanced, but the model size requirement is also significantly reduced while achieving comparable performance. Our proposed knowledge-infused model surpasses the performance of both GPT-3.5-turbo and the state-of-the-art knowledge infusion method, SKILL, achieving over 1.5 times improvement in exact match scores on the MetaQA. KITLM showed a similar performance boost in the aviation domain with AeroQA. The drastic performance improvement of KITLM over the existing methods can be attributed to the infusion of relevant knowledge while mitigating noise. In addition, we release two curated datasets to accelerate knowledge infusion research in specialized fields: a) AeroQA, a new benchmark dataset designed for multi-hop question-answering within the aviation domain, and b) Aviation Corpus, a dataset constructed from unstructured text extracted from the National Transportation Safety Board reports. Our research contributes to advancing the field of domain-specific language understanding and showcases the potential of knowledge infusion techniques in improving the performance of language models on question-answering.

Ankush Agarwal, Sakharam Gawade, Sachin Channabasavarajendra et al.

The integration of knowledge graphs with deep learning is thriving in improving the performance of various natural language processing (NLP) tasks. In this paper, we focus on knowledge-infused link prediction and question answering using language models, T5, and BLOOM across three domains: Aviation, Movie, and Web. In this context, we infuse knowledge in large and small language models and study their performance, and find the performance to be similar. For the link prediction task on the Aviation Knowledge Graph, we obtain a 0.2 hits@1 score using T5-small, T5-base, T5-large, and BLOOM. Using template-based scripts, we create a set of 1 million synthetic factoid QA pairs in the aviation domain from National Transportation Safety Board (NTSB) reports. On our curated QA pairs, the three models of T5 achieve a 0.7 hits@1 score. We validate out findings with the paired student t-test and Cohen's kappa scores. For link prediction on Aviation Knowledge Graph using T5-small and T5-large, we obtain a Cohen's kappa score of 0.76, showing substantial agreement between the models. Thus, we infer that small language models perform similar to large language models with the infusion of knowledge.

Abhisek Tiwari, Muhammed Sinan, Kaushik Roy et al.

Over the past two decades, dialogue modeling has made significant strides, moving from simple rule-based responses to personalized and persuasive response generation. However, despite these advancements, the objective functions and evaluation metrics for dialogue generation have remained stagnant. These lexical-based metrics, e.g., cross-entropy and BLEU, have two key limitations: (a) word-to-word matching without semantic consideration: It assigns the same credit for failure to generate "nice" and "rice" for "good", (b) missing context attribute for evaluating the generated response: Even if a generated response is relevant to the ongoing dialogue context, it may still be penalized for not matching the gold utterance provided in the corpus. In this paper, we first investigate these limitations comprehensively and propose a new loss function called Semantic Infused Contextualized diaLogue (SemTextualLogue) loss function. We also formulate an evaluation metric called Dialuation, incorporating both context and semantic relevance. We experimented with both non-pretrained and pre-trained models on two dialogue corpora, encompassing task-oriented and open-domain scenarios. We found that the dialogue generation models trained with SemTextualLogueloss attained superior performance compared to the traditional cross-entropy loss function. The findings establish that the effective training of a dialogue generation model hinges significantly on incorporating semantics and context. This pattern is also mirrored in the introduced Dialuation metric, where the consideration of both context and semantics correlates more strongly with human evaluation compared to traditional metrics.

Vineet Bhat, Preethi Jyothi, Pushpak Bhattacharyya

Disfluencies commonly occur in conversational speech. Speech with disfluencies can result in noisy Automatic Speech Recognition (ASR) transcripts, which affects downstream tasks like machine translation. In this paper, we propose an adversarially-trained sequence-tagging model for Disfluency Correction (DC) that utilizes a small amount of labeled real disfluent data in conjunction with a large amount of unlabeled data. We show the benefit of our proposed technique, which crucially depends on synthetically generated disfluent data, by evaluating it for DC in three Indian languages- Bengali, Hindi, and Marathi (all from the Indo-Aryan family). Our technique also performs well in removing stuttering disfluencies in ASR transcripts introduced by speech impairments. We achieve an average 6.15 points improvement in F1-score over competitive baselines across all three languages mentioned. To the best of our knowledge, we are the first to utilize adversarial training for DC and use it to correct stuttering disfluencies in English, establishing a new benchmark for this task.

Akshay Batheja, Pushpak Bhattacharyya

Quality Estimation (QE) is the task of evaluating the quality of a translation when reference translation is not available. The goal of QE aligns with the task of corpus filtering, where we assign the quality score to the sentence pairs present in the pseudo-parallel corpus. We propose a Quality Estimation based Filtering approach to extract high-quality parallel data from the pseudo-parallel corpus. To the best of our knowledge, this is a novel adaptation of the QE framework to extract quality parallel corpus from the pseudo-parallel corpus. By training with this filtered corpus, we observe an improvement in the Machine Translation (MT) system's performance by up to 1.8 BLEU points, for English-Marathi, Chinese-English, and Hindi-Bengali language pairs, over the baseline model. The baseline model is the one that is trained on the whole pseudo-parallel corpus. Our Few-shot QE model transfer learned from the English-Marathi QE model and fine-tuned on only 500 Hindi-Bengali training instances, shows an improvement of up to 0.6 BLEU points for Hindi-Bengali language pair, compared to the baseline model. This demonstrates the promise of transfer learning in the setting under discussion. QE systems typically require in the order of (7K-25K) of training data. Our Hindi-Bengali QE is trained on only 500 instances of training that is 1/40th of the normal requirement and achieves comparable performance. All the scripts and datasets utilized in this study will be publicly available.

N V S Abhishek, Pushpak Bhattacharyya

Speech Emotion Recognition (SER) is the task of identifying the emotion expressed in a spoken utterance. Emotion recognition is essential in building robust conversational agents in domains such as law, healthcare, education, and customer support. Most of the studies published on SER use datasets created by employing professional actors in a noise-free environment. In natural settings such as a customer care conversation, the audio is often noisy with speakers regularly switching between different languages as they see fit. We have worked in collaboration with a leading unicorn in the Conversational AI sector to develop Natural Speech Emotion Dataset (NSED). NSED is a natural code-mixed speech emotion dataset where each utterance in a conversation is annotated with emotion, sentiment, valence, arousal, and dominance (VAD) values. In this paper, we show that by incorporating word-level VAD value we improve on the task of SER by 2%, for negative emotions, over the baseline value for NSED. High accuracy for negative emotion recognition is essential because customers expressing negative opinions/views need to be pacified with urgency, lest complaints and dissatisfaction snowball and get out of hand. Escalation of negative opinions speedily is crucial for business interests. Our study then can be utilized to develop conversational agents which are more polite and empathetic in such situations.

Sandeep Singamsetty, Nishtha Madaan, Sameep Mehta et al.

The prevalence of half-truths, which are statements containing some truth but that are ultimately deceptive, has risen with the increasing use of the internet. To help combat this problem, we have created a comprehensive pipeline consisting of a half-truth detection model and a claim editing model. Our approach utilizes the T5 model for controlled claim editing; "controlled" here means precise adjustments to select parts of a claim. Our methodology achieves an average BLEU score of 0.88 (on a scale of 0-1) and a disinfo-debunk score of 85% on edited claims. Significantly, our T5-based approach outperforms other Language Models such as GPT2, RoBERTa, PEGASUS, and Tailor, with average improvements of 82%, 57%, 42%, and 23% in disinfo-debunk scores, respectively. By extending the LIAR PLUS dataset, we achieve an F1 score of 82% for the half-truth detection model, setting a new benchmark in the field. While previous attempts have been made at half-truth detection, our approach is, to the best of our knowledge, the first to attempt to debunk half-truths.

Prerak Gandhi, Vishal Pramanik, Pushpak Bhattacharyya

Storytelling is the lifeline of the entertainment industry -- movies, TV shows, and stand-up comedies, all need stories. A good and gripping script is the lifeline of storytelling and demands creativity and resource investment. Good scriptwriters are rare to find and often work under severe time pressure. Consequently, entertainment media are actively looking for automation. In this paper, we present an AI-based script-writing workbench called KUROSAWA which addresses the tasks of plot generation and script generation. Plot generation aims to generate a coherent and creative plot (600-800 words) given a prompt (15-40 words). Script generation, on the other hand, generates a scene (200-500 words) in a screenplay format from a brief description (15-40 words). Kurosawa needs data to train. We use a 4-act structure of storytelling to annotate the plot dataset manually. We create a dataset of 1000 manually annotated plots and their corresponding prompts/storylines and a gold-standard dataset of 1000 scenes with four main elements -- scene headings, action lines, dialogues, and character names -- tagged individually. We fine-tune GPT-3 with the above datasets to generate plots and scenes. These plots and scenes are first evaluated and then used by the scriptwriters of a large and famous media platform ErosNow. We release the annotated datasets and the models trained on these datasets as a working benchmark for automatic movie plot and script generation.

Nithish Kannen, Arif Ahmad, Marco Andreetto et al.

Text-to-Image (T2I) models are being increasingly adopted in diverse global communities where they create visual representations of their unique cultures. Current T2I benchmarks primarily focus on faithfulness, aesthetics, and realism of generated images, overlooking the critical dimension of cultural competence. In this work, we introduce a framework to evaluate cultural competence of T2I models along two crucial dimensions: cultural awareness and cultural diversity, and present a scalable approach using a combination of structured knowledge bases and large language models to build a large dataset of cultural artifacts to enable this evaluation. In particular, we apply this approach to build CUBE (CUltural BEnchmark for Text-to-Image models), a first-of-its-kind benchmark to evaluate cultural competence of T2I models. CUBE covers cultural artifacts associated with 8 countries across different geo-cultural regions and along 3 concepts: cuisine, landmarks, and art. CUBE consists of 1) CUBE-1K, a set of high-quality prompts that enable the evaluation of cultural awareness, and 2) CUBE-CSpace, a larger dataset of cultural artifacts that serves as grounding to evaluate cultural diversity. We also introduce cultural diversity as a novel T2I evaluation component, leveraging quality-weighted Vendi score. Our evaluations reveal significant gaps in the cultural awareness of existing models across countries and provide valuable insights into the cultural diversity of T2I outputs for under-specified prompts. Our methodology is extendable to other cultural regions and concepts, and can facilitate the development of T2I models that better cater to the global population.

Tamali Banerjee, Rudra Murthy, Pushpak Bhattacharyya

We aim to investigate whether UNMT approaches with self-supervised pre-training are robust to word-order divergence between language pairs. We achieve this by comparing two models pre-trained with the same self-supervised pre-training objective. The first model is trained on language pairs with different word-orders, and the second model is trained on the same language pairs with source language re-ordered to match the word-order of the target language. Ideally, UNMT approaches which are robust to word-order divergence should exhibit no visible performance difference between the two configurations. In this paper, we investigate two such self-supervised pre-training based UNMT approaches, namely Masked Sequence-to-Sequence Pre-Training, (MASS) (which does not have shuffling noise) and Denoising AutoEncoder (DAE), (which has shuffling noise). We experiment with five English$\rightarrow$Indic language pairs, i.e., en-hi, en-bn, en-gu, en-kn, and en-ta) where word-order of the source language is SVO (Subject-Verb-Object), and the word-order of the target languages is SOV (Subject-Object-Verb). We observed that for these language pairs, DAE-based UNMT approach consistently outperforms MASS in terms of translation accuracies. Moreover, bridging the word-order gap using reordering improves the translation accuracy of MASS-based UNMT models, while it cannot improve the translation accuracy of DAE-based UNMT models. This observation indicates that DAE-based UNMT is more robust to word-order divergence than MASS-based UNMT. Word-shuffling noise in DAE approach could be the possible reason for the approach being robust to word-order divergence.

Swapnil Bhosale, Abhra Chaudhuri, Alex Lee Robert Williams et al.

The introduction of the MUStARD dataset, and its emotion recognition extension MUStARD++, have identified sarcasm to be a multi-modal phenomenon -- expressed not only in natural language text, but also through manners of speech (like tonality and intonation) and visual cues (facial expression). With this work, we aim to perform a rigorous benchmarking of the MUStARD++ dataset by considering state-of-the-art language, speech, and visual encoders, for fully utilizing the totality of the multi-modal richness that it has to offer, achieving a 2\% improvement in macro-F1 over the existing benchmark. Additionally, to cure the imbalance in the `sarcasm type' category in MUStARD++, we propose an extension, which we call \emph{MUStARD++ Balanced}, benchmarking the same with instances from the extension split across both train and test sets, achieving a further 2.4\% macro-F1 boost. The new clips were taken from a novel source -- the TV show, House MD, which adds to the diversity of the dataset, and were manually annotated by multiple annotators with substantial inter-annotator agreement in terms of Cohen's kappa and Krippendorf's alpha. Our code, extended data, and SOTA benchmark models are made public.

Kaustubh Shivshankar Shejole, Sourabh Deoghare, Pushpak Bhattacharyya

Neural Machine Translation (NMT) systems rely heavily on explicit punctuation cues to resolve semantic ambiguities in a source sentence. Inputting user-generated sentences, which are likely to contain missing or incorrect punctuation, results in fluent but semantically disastrous translations. This work attempts to highlight and address the problem of punctuation robustness of NMT systems through an English-to-Marathi translation. First, we introduce \textbf{\textit{Viram}}, a human-curated diagnostic benchmark of 54 punctuation-ambiguous English-Marathi sentence pairs to stress-test existing NMT systems. Second, we evaluate two simple remediation strategies: cascade-based \textit{restore-then-translate} and \textit{direct fine-tuning}. Our experimental results and analysis demonstrate that both strategies yield substantial NMT performance improvements. Furthermore, we find that current Large Language Models (LLMs) exhibit relatively poorer robustness in translating such sentences than these task-specific strategies, thus necessitating further research in this area. The code and dataset are available at https://github.com/KaustubhShejole/Viram_Marathi.

Sanjeev Kumar, Preethi Jyothi, Pushpak Bhattacharyya

Evaluating machine translation (MT) quality in extremely low-resource language (ELRL) scenarios poses unique challenges, as widely used metrics such as BLEU, effective in high-resource settings, often misrepresent quality in data-scarce contexts. This work presents a comparative analysis of BLEU, an n-gram-based metric, and ChrF++, a character-based metric, for MT evaluation in ELRL settings. We examine how each metric responds to translation artifacts, including hallucinations, repetition, source-text copying, and diacritic (\textit{matra}) variations across three ELRLs: Magahi, Bhojpuri, and Chhattisgarhi, with a focus on outputs from large language models (LLMs) and neural MT (NMT) systems. While recent work often relies solely on ChrF++, our findings show that BLEU, despite its lower absolute scores, provides complementary lexical-precision insights that improve interpretability.

Tejpalsingh Siledar, Swaroop Nath, Sankara Sri Raghava Ravindra Muddu et al.

Evaluation of opinion summaries using conventional reference-based metrics rarely provides a holistic evaluation and has been shown to have a relatively low correlation with human judgments. Recent studies suggest using Large Language Models (LLMs) as reference-free metrics for NLG evaluation, however, they remain unexplored for opinion summary evaluation. Moreover, limited opinion summary evaluation datasets inhibit progress. To address this, we release the SUMMEVAL-OP dataset covering 7 dimensions related to the evaluation of opinion summaries: fluency, coherence, relevance, faithfulness, aspect coverage, sentiment consistency, and specificity. We investigate Op-I-Prompt a dimension-independent prompt, and Op-Prompts, a dimension-dependent set of prompts for opinion summary evaluation. Experiments indicate that Op-I-Prompt emerges as a good alternative for evaluating opinion summaries achieving an average Spearman correlation of 0.70 with humans, outperforming all previous approaches. To the best of our knowledge, we are the first to investigate LLMs as evaluators on both closed-source and open-source models in the opinion summarization domain.

Swaroop Nath, Harshad Khadilkar, Pushpak Bhattacharyya

Query-focused Summarization (QfS) deals with systems that generate summaries from document(s) based on a query. Motivated by the insight that Reinforcement Learning (RL) provides a generalization to Supervised Learning (SL) for Natural Language Generation, and thereby performs better (empirically) than SL, we use an RL-based approach for this task of QfS. Additionally, we also resolve the conflict of employing RL in Transformers with Teacher Forcing. We develop multiple Policy Gradient networks, trained on various reward signals: ROUGE, BLEU, and Semantic Similarity, which lead to a 10-point improvement over the State-of-the-Art approach on the ROUGE-L metric for a benchmark dataset (ELI5). We also show performance of our approach in zero-shot setting for another benchmark dataset (DebatePedia) -- our approach leads to results comparable to baselines, which were specifically trained on DebatePedia. To aid the RL training, we propose a better semantic similarity reward, enabled by a novel Passage Embedding scheme developed using Cluster Hypothesis. Lastly, we contribute a gold-standard test dataset to further research in QfS and Long-form Question Answering (LfQA).

Mehant Kammakomati, Sameer Pimparkhede, Srikanth Tamilselvam et al.

Recent work shows Large Language Models (LLMs) struggle to understand natural language constraints for various text generation tasks in zero- and few-shot settings. While, in the code domain, there is wide usage of constraints in code format to maintain the integrity of code written in Domain-Specific Languages (DSLs) like JSON and YAML which are widely used for system-level programming tasks in enterprises. Given that LLMs are increasingly used for system-level code tasks, evaluating if they can comprehend these code constraints is crucial. However, no work has been done to evaluate their controllability over code constraints. Hence, we introduce ConCodeEval, a first-of-its-kind benchmark having two novel tasks for code constraints across five representations. Our findings suggest that language models struggle with code constraints. Code languages that perform excellently for normal code tasks do not perform well when the same languages represent fine-grained constraints.

Ramakrishna Appicharla, Baban Gain, Santanu Pal et al.

In document-level neural machine translation (DocNMT), multi-encoder approaches are common in encoding context and source sentences. Recent studies \cite{li-etal-2020-multi-encoder} have shown that the context encoder generates noise and makes the model robust to the choice of context. This paper further investigates this observation by explicitly modelling context encoding through multi-task learning (MTL) to make the model sensitive to the choice of context. We conduct experiments on cascade MTL architecture, which consists of one encoder and two decoders. Generation of the source from the context is considered an auxiliary task, and generation of the target from the source is the main task. We experimented with German--English language pairs on News, TED, and Europarl corpora. Evaluation results show that the proposed MTL approach performs better than concatenation-based and multi-encoder DocNMT models in low-resource settings and is sensitive to the choice of context. However, we observe that the MTL models are failing to generate the source from the context. These observations align with the previous studies, and this might suggest that the available document-level parallel corpora are not context-aware, and a robust sentence-level model can outperform the context-aware models.

Sapan Shah, Sreedhar Reddy, Pushpak Bhattacharyya

We present a novel retrofitting method to induce emotion aspects into pre-trained language models (PLMs) such as BERT and RoBERTa. Our method updates pre-trained network weights using contrastive learning so that the text fragments exhibiting similar emotions are encoded nearby in the representation space, and the fragments with different emotion content are pushed apart. While doing so, it also ensures that the linguistic knowledge already present in PLMs is not inadvertently perturbed. The language models retrofitted by our method, i.e., BERTEmo and RoBERTaEmo, produce emotion-aware text representations, as evaluated through different clustering and retrieval metrics. For the downstream tasks on sentiment analysis and sarcasm detection, they perform better than their pre-trained counterparts (about 1% improvement in F1-score) and other existing approaches. Additionally, a more significant boost in performance is observed for the retrofitted models over pre-trained ones in few-shot learning setting.

Saiful Haq, Ashutosh Sharma, Pushpak Bhattacharyya

In this paper, we introduce Neural Information Retrieval resources for 11 widely spoken Indian Languages (Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Oriya, Punjabi, Tamil, and Telugu) from two major Indian language families (Indo-Aryan and Dravidian). These resources include (a) INDIC-MARCO, a multilingual version of the MSMARCO dataset in 11 Indian Languages created using Machine Translation, and (b) Indic-ColBERT, a collection of 11 distinct Monolingual Neural Information Retrieval models, each trained on one of the 11 languages in the INDIC-MARCO dataset. To the best of our knowledge, IndicIRSuite is the first attempt at building large-scale Neural Information Retrieval resources for a large number of Indian languages, and we hope that it will help accelerate research in Neural IR for Indian Languages. Experiments demonstrate that Indic-ColBERT achieves 47.47% improvement in the MRR@10 score averaged over the INDIC-MARCO baselines for all 11 Indian languages except Oriya, 12.26% improvement in the NDCG@10 score averaged over the MIRACL Bengali and Hindi Language baselines, and 20% improvement in the MRR@100 Score over the Mr.Tydi Bengali Language baseline. IndicIRSuite is available at https://github.com/saifulhaq95/IndicIRSuite

Swaroop Nath, Harshad Khadilkar, Pushpak Bhattacharyya

Transformers have become pivotal in Natural Language Processing, demonstrating remarkable success in applications like Machine Translation and Summarization. Given their widespread adoption, several works have attempted to analyze the expressivity of Transformers. Expressivity of a neural network is the class of functions it can approximate. A neural network is fully expressive if it can act as a universal function approximator. We attempt to analyze the same for Transformers. Contrary to existing claims, our findings reveal that Transformers struggle to reliably approximate smooth functions, relying on piecewise constant approximations with sizable intervals. The central question emerges as: ''Are Transformers truly Universal Function Approximators?'' To address this, we conduct a thorough investigation, providing theoretical insights and supporting evidence through experiments. Theoretically, we prove that Transformer Encoders cannot approximate smooth functions. Experimentally, we complement our theory and show that the full Transformer architecture cannot approximate smooth functions. By shedding light on these challenges, we advocate a refined understanding of Transformers' capabilities. Code Link: https://github.com/swaroop-nath/transformer-expressivity.

Kaustubh Shivshankar Shejole, Pushpak Bhattacharyya

Stereotypes are known to have very harmful effects, making their detection critically important. However, current research predominantly focuses on detecting and evaluating stereotypical biases, thereby leaving the study of stereotypes in its early stages. Our study revealed that many works have failed to clearly distinguish between stereotypes and stereotypical biases, which has significantly slowed progress in advancing research in this area. Stereotype and Anti-stereotype detection is a problem that requires social knowledge; hence, it is one of the most difficult areas in Responsible AI. This work investigates this task, where we propose a five-tuple definition and provide precise terminologies disentangling stereotypes, anti-stereotypes, stereotypical bias, and general bias. We provide a conceptual framework grounded in social psychology for reliable detection. We identify key shortcomings in existing benchmarks for this task of stereotype and anti-stereotype detection. To address these gaps, we developed StereoDetect, a well curated, definition-aligned benchmark dataset designed for this task. We show that sub-10B language models and GPT-4o frequently misclassify anti-stereotypes and fail to recognize neutral overgeneralizations. We demonstrate StereoDetect's effectiveness through multiple qualitative and quantitative comparisons with existing benchmarks and models fine-tuned on them. The dataset and code is available at https://github.com/KaustubhShejole/StereoDetect.

Abhisek Tiwari, Shreyangshu Bera, Sriparna Saha et al.

Over the past few years, the use of the Internet for healthcare-related tasks has grown by leaps and bounds, posing a challenge in effectively managing and processing information to ensure its efficient utilization. During moments of emotional turmoil and psychological challenges, we frequently turn to the internet as our initial source of support, choosing this over discussing our feelings with others due to the associated social stigma. In this paper, we propose a new task of multi-modal medical concern summary (MMCS) generation, which provides a short and precise summary of patients' major concerns brought up during the consultation. Nonverbal cues, such as patients' gestures and facial expressions, aid in accurately identifying patients' concerns. Doctors also consider patients' personal information, such as age and gender, in order to describe the medical condition appropriately. Motivated by the potential efficacy of patients' personal context and visual gestures, we propose a transformer-based multi-task, multi-modal intent-recognition, and medical concern summary generation (IR-MMCSG) system. Furthermore, we propose a multitasking framework for intent recognition and medical concern summary generation for doctor-patient consultations. We construct the first multi-modal medical concern summary generation (MM-MediConSummation) corpus, which includes patient-doctor consultations annotated with medical concern summaries, intents, patient personal information, doctor's recommendations, and keywords. Our experiments and analysis demonstrate (a) the significant role of patients' expressions/gestures and their personal information in intent identification and medical concern summary generation, and (b) the strong correlation between intent recognition and patients' medical concern summary generation The dataset and source code are available at https://github.com/NLP-RL/MMCSG.

Arif Ahmad, Mothika Gayathri Khyathi, Pushpak Bhattacharyya

Reduplication and repetition, though similar in form, serve distinct linguistic purposes. Reduplication is a deliberate morphological process used to express grammatical, semantic, or pragmatic nuances, while repetition is often unintentional and indicative of disfluency. This paper presents the first large-scale study of reduplication and repetition in speech using computational linguistics. We introduce IndicRedRep, a new publicly available dataset containing Hindi, Telugu, and Marathi text annotated with reduplication and repetition at the word level. We evaluate transformer-based models for multi-class reduplication and repetition token classification, utilizing the Reparandum-Interregnum-Repair structure to distinguish between the two phenomena. Our models achieve macro F1 scores of up to 85.62% in Hindi, 83.95% in Telugu, and 84.82% in Marathi for reduplication-repetition classification.

Shrey Ganatra, Swapnil Bhattacharyya, Harshvivek Kashid et al.

Access to consumer grievance redressal in India is often hindered by procedural complexity, legal jargon, and jurisdictional challenges. To address this, we present $\textbf{Grahak-Nyay}$ (Justice-to-Consumers), a chatbot that streamlines the process using open-source Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). Grahak-Nyay simplifies legal complexities through a concise and up-to-date knowledge base. We introduce three novel datasets: $\textit{GeneralQA}$ (general consumer law), $\textit{SectoralQA}$ (sector-specific knowledge) and $\textit{SyntheticQA}$ (for RAG evaluation), along with $\textit{NyayChat}$, a dataset of 300 annotated chatbot conversations. We also introduce $\textit{Judgments}$ data sourced from Indian Consumer Courts to aid the chatbot in decision making and to enhance user trust. We also propose $\textbf{HAB}$ metrics ($\textbf{Helpfulness, Accuracy, Brevity}$) to evaluate chatbot performance. Legal domain experts validated Grahak-Nyay's effectiveness. Code and datasets will be released.

Arnav Attri, Anuj Attri, Pushpak Bhattacharyya et al.

Product recommendations inherently involve comparisons, yet traditional opinion summarization often fails to provide holistic comparative insights. We propose the novel task of generating Query-Focused Comparative Explainable Summaries (QF-CES) using Multi-Source Opinion Summarization (M-OS). To address the lack of query-focused recommendation datasets, we introduce MS-Q2P, comprising 7,500 queries mapped to 22,500 recommended products with metadata. We leverage Large Language Models (LLMs) to generate tabular comparative summaries with query-specific explanations. Our approach is personalized, privacy-preserving, recommendation engine-agnostic, and category-agnostic. M-OS as an intermediate step reduces inference latency approximately by 40% compared to the direct input approach (DIA), which processes raw data directly. We evaluate open-source and proprietary LLMs for generating and assessing QF-CES. Extensive evaluations using QF-CES-PROMPT across 5 dimensions (clarity, faithfulness, informativeness, format adherence, and query relevance) showed an average Spearman correlation of 0.74 with human judgments, indicating its potential for QF-CES evaluation.

Sameer Pimparkhede, Mehant Kammakomati, Srikanth Tamilselvam et al.

Recent developments show that Large Language Models (LLMs) produce state-of-the-art performance on natural language (NL) to code generation for resource-rich general-purpose languages like C++, Java, and Python. However, their practical usage for structured domain-specific languages (DSLs) such as YAML, JSON is limited due to domain-specific schema, grammar, and customizations generally unseen by LLMs during pre-training. Efforts have been made to mitigate this challenge via in-context learning through relevant examples or by fine-tuning. However, it suffers from problems, such as limited DSL samples and prompt sensitivity but enterprises maintain good documentation of the DSLs. Therefore, we propose DocCGen, a framework that can leverage such rich knowledge by breaking the NL-to-Code generation task for structured code languages into a two-step process. First, it detects the correct libraries using the library documentation that best matches the NL query. Then, it utilizes schema rules extracted from the documentation of these libraries to constrain the decoding. We evaluate our framework for two complex structured languages, Ansible YAML and Bash command, consisting of two settings: Out-of-domain (OOD) and In-domain (ID). Our extensive experiments show that DocCGen consistently improves different-sized language models across all six evaluation metrics, reducing syntactic and semantic errors in structured code. We plan to open-source the datasets and code to motivate research in constrained code generation.

Shrey Ganatra, Spandan Anaokar, Pushpak Bhattacharyya

The proliferation of IoT devices generates vast interaction data, offering insights into user behaviour. While prior work predicts what actions users perform, the timing of these actions -- critical for enabling proactive and efficient smart systems -- remains relatively underexplored. Addressing this gap, we focus on predicting the time of the next user action in smart environments. Due to the lack of public datasets with fine-grained timestamps suitable for this task and associated privacy concerns, we contribute a dataset of 11.6k sequences synthesized based on human annotations of interaction patterns, pairing actions with precise timestamps. To this end, we introduce Timing-Matters, a Transformer-Encoder based method that predicts action timing, achieving 38.30% accuracy on the synthesized dataset, outperforming the best baseline by 6%, and showing 1--6% improvements on other open datasets. Our code and dataset will be publicly released.

Sourabh Deoghare, Diptesh Kanojia, Pushpak Bhattacharyya

This exploratory study investigates the potential of multilingual Automatic Post-Editing (APE) systems to enhance the quality of machine translations for low-resource Indo-Aryan languages. Focusing on two closely related language pairs, English-Marathi and English-Hindi, we exploit the linguistic similarities to develop a robust multilingual APE model. To facilitate cross-linguistic transfer, we generate synthetic Hindi-Marathi and Marathi-Hindi APE triplets. Additionally, we incorporate a Quality Estimation (QE)-APE multi-task learning framework. While the experimental results underline the complementary nature of APE and QE, we also observe that QE-APE multitask learning facilitates effective domain adaptation. Our experiments demonstrate that the multilingual APE models outperform their corresponding English-Hindi and English-Marathi single-pair models by $2.5$ and $2.39$ TER points, respectively, with further notable improvements over the multilingual APE model observed through multi-task learning ($+1.29$ and $+1.44$ TER points), data augmentation ($+0.53$ and $+0.45$ TER points) and domain adaptation ($+0.35$ and $+0.45$ TER points). We release the synthetic data, code, and models accrued during this study publicly at https://github.com/cfiltnlp/Multilingual-APE.

Dushyant Singh Chauhan, Gopendra Vikram Singh, Navonil Majumder et al.

Humor recognition in conversations is a challenging task that has recently gained popularity due to its importance in dialogue understanding, including in multimodal settings (i.e., text, acoustics, and visual). The few existing datasets for humor are mostly in English. However, due to the tremendous growth in multilingual content, there is a great demand to build models and systems that support multilingual information access. To this end, we propose a dataset for Multimodal Multiparty Hindi Humor (M2H2) recognition in conversations containing 6,191 utterances from 13 episodes of a very popular TV series "Shrimaan Shrimati Phir Se". Each utterance is annotated with humor/non-humor labels and encompasses acoustic, visual, and textual modalities. We propose several strong multimodal baselines and show the importance of contextual and multimodal information for humor recognition in conversations. The empirical results on M2H2 dataset demonstrate that multimodal information complements unimodal information for humor recognition. The dataset and the baselines are available at http://www.iitp.ac.in/~ai-nlp-ml/resources.html and https://github.com/declare-lab/M2H2-dataset.

Sneha Oram, Ojaswita Bhushan, Pushpak Bhattacharyya

In this work, we conduct an analysis to examine the consistency of Large Language Models (LLMs) with respect to their own generated responses in an emotionally-driven conversational context. Specifically, the text generated by LLM is framed as a query to the same model, and its responses are subsequently assessed. This is performed with three queries across two dimensions of extreme and moderate emotions. The three queries are, in particular, false claim queries that contain inherently wrong assumptions (false presuppositions) in increasing order of intensity. Two commercial models, Claude-3.5-haiku, GPT4o-mini, and a medium-sized model, Mistral-7B, are considered in the study. Our findings indicate that LLMs exhibit below-average performance and remain vulnerable to false beliefs embedded within queries. This susceptibility is especially pronounced for moderate emotional content. Furthermore, an extended attention-score-based analysis highlights a shift in models' priority from evaluative to generative. The results raise important considerations for LLMs' deployment in high-stakes, emotionally sensitive contexts.

Settaluri Lakshmi Sravanthi, Meet Doshi, Tankala Pavan Kalyan et al.

LLMs have demonstrated remarkable capability for understanding semantics, but they often struggle with understanding pragmatics. To demonstrate this fact, we release a Pragmatics Understanding Benchmark (PUB) dataset consisting of fourteen tasks in four pragmatics phenomena, namely, Implicature, Presupposition, Reference, and Deixis. We curated high-quality test sets for each task, consisting of Multiple Choice Question Answers (MCQA). PUB includes a total of 28k data points, 6.1k of which have been created by us, and the rest are adapted from existing datasets. We evaluated nine models varying in the number of parameters and type of training. Our study indicates that fine-tuning for instruction-following and chat significantly enhances the pragmatics capabilities of smaller language models. However, for larger models, the base versions perform comparably with their chat-adapted counterparts. Additionally, there is a noticeable performance gap between human capabilities and model capabilities. Furthermore, unlike the consistent performance of humans across various tasks, the models demonstrate variability in their proficiency, with performance levels fluctuating due to different hints and the complexities of tasks within the same dataset. Overall, the benchmark aims to provide a comprehensive evaluation of LLM's ability to handle real-world language tasks that require pragmatic reasoning.

Nihar Ranjan Sahoo, Pranamya Prashant Kulkarni, Narjis Asad et al.

The pervasive influence of social biases in language data has sparked the need for benchmark datasets that capture and evaluate these biases in Large Language Models (LLMs). Existing efforts predominantly focus on English language and the Western context, leaving a void for a reliable dataset that encapsulates India's unique socio-cultural nuances. To bridge this gap, we introduce IndiBias, a comprehensive benchmarking dataset designed specifically for evaluating social biases in the Indian context. We filter and translate the existing CrowS-Pairs dataset to create a benchmark dataset suited to the Indian context in Hindi language. Additionally, we leverage LLMs including ChatGPT and InstructGPT to augment our dataset with diverse societal biases and stereotypes prevalent in India. The included bias dimensions encompass gender, religion, caste, age, region, physical appearance, and occupation. We also build a resource to address intersectional biases along three intersectional dimensions. Our dataset contains 800 sentence pairs and 300 tuples for bias measurement across different demographics. The dataset is available in English and Hindi, providing a size comparable to existing benchmark datasets. Furthermore, using IndiBias we compare ten different language models on multiple bias measurement metrics. We observed that the language models exhibit more bias across a majority of the intersectional groups.

Tejpalsingh Siledar, Rupasai Rangaraju, Sankara Sri Raghava Ravindra Muddu et al.

In e-commerce, opinion summarization is the process of summarizing the consensus opinions found in product reviews. However, the potential of additional sources such as product description and question-answers (QA) has been considered less often. Moreover, the absence of any supervised training data makes this task challenging. To address this, we propose a novel synthetic dataset creation (SDC) strategy that leverages information from reviews as well as additional sources for selecting one of the reviews as a pseudo-summary to enable supervised training. Our Multi-Encoder Decoder framework for Opinion Summarization (MEDOS) employs a separate encoder for each source, enabling effective selection of information while generating the summary. For evaluation, due to the unavailability of test sets with additional sources, we extend the Amazon, Oposum+, and Flipkart test sets and leverage ChatGPT to annotate summaries. Experiments across nine test sets demonstrate that the combination of our SDC approach and MEDOS model achieves on average a 14.5% improvement in ROUGE-1 F1 over the SOTA. Moreover, comparative analysis underlines the significance of incorporating additional sources for generating more informative summaries. Human evaluations further indicate that MEDOS scores relatively higher in coherence and fluency with 0.41 and 0.5 (-1 to 1) respectively, compared to existing models. To the best of our knowledge, we are the first to generate opinion summaries leveraging additional sources in a self-supervised setting.

Poulami Ghosh, Shikhar Vashishth, Raj Dabre et al. · cmu

Contemporary deep learning models effectively handle languages with diverse morphology despite not being directly integrated into them. Morphology and word order are closely linked, with the latter incorporated into transformer-based models through positional encodings. This prompts a fundamental inquiry: Is there a correlation between the morphological complexity of a language and the utilization of positional encoding in pre-trained language models? In pursuit of an answer, we present the first study addressing this question, encompassing 22 languages and 5 downstream tasks. Our findings reveal that the importance of positional encoding diminishes with increasing morphological complexity in languages. Our study motivates the need for a deeper understanding of positional encoding, augmenting them to better reflect the different languages under consideration.

Meet Doshi, Raj Dabre, Pushpak Bhattacharyya

In this paper, we explore the utility of translationese as synthetic data created using machine translation for pre-training language models (LMs) for low-resource languages (LRLs). Our simple methodology consists of translating large amounts of web-crawled monolingual documents (clean) into the LRLs, followed by filtering the translated documents using tiny LMs trained on small but clean LRL data. Taking the case of Indian languages, we pre-train LMs from scratch with 28M and 85M parameters, and then fine-tune them for 5 downstream natural language understanding (NLU) and 4 generative (NLG) tasks. We observe that pre-training on filtered synthetic data leads to relative performance drops of only 0.87% for NLU and 2.35% for NLG, compared to pre-training on clean data, and this gap further diminishes upon the inclusion of a small amount of clean data. We also study the impact of synthetic data filtering and the choice of source language for synthetic data generation. Furthermore, evaluating continually pre-trained larger models like Gemma-2B and Llama-3-8B in few-shot settings, we observe that using synthetic data is competitive with using clean data. Our findings suggest that synthetic data shows promise for bridging the pre-training gap between English and LRLs.

Avadhut Sardeshmukh, Sreedhar Reddy, BP Gautham et al.

We propose a variational autoencoder (VAE)-based model for building forward and inverse structure-property linkages, a problem of paramount importance in computational materials science. Our model systematically combines VAE with regression, linking the two models through a two-level prior conditioned on the regression variables. The regression loss is optimized jointly with the reconstruction loss of the variational autoencoder, learning microstructure features relevant for property prediction and reconstruction. The resultant model can be used for both forward and inverse prediction i.e., for predicting the properties of a given microstructure as well as for predicting the microstructure required to obtain given properties. Since the inverse problem is ill-posed (one-to-many), we derive the objective function using a multi-modal Gaussian mixture prior enabling the model to infer multiple microstructures for a target set of properties. We show that for forward prediction, our model is as accurate as state-of-the-art forward-only models. Additionally, our method enables direct inverse inference. We show that the microstructures inferred using our model achieve desired properties reasonably accurately, avoiding the need for expensive optimization loops.

Krishanu Maity, Prince Jha, Raghav Jain et al.

Cyberbullying has become a big issue with the popularity of different social media networks and online communication apps. While plenty of research is going on to develop better models for cyberbullying detection in monolingual language, there is very little research on the code-mixed languages and explainability aspect of cyberbullying. Recent laws like "right to explanations" of General Data Protection Regulation, have spurred research in developing interpretable models rather than focusing on performance. Motivated by this we develop the first interpretable multi-task model called {\em mExCB} for automatic cyberbullying detection from code-mixed languages which can simultaneously solve several tasks, cyberbullying detection, explanation/rationale identification, target group detection and sentiment analysis. We have introduced {\em BullyExplain}, the first benchmark dataset for explainable cyberbullying detection in code-mixed language. Each post in {\em BullyExplain} dataset is annotated with four labels, i.e., {\em bully label, sentiment label, target and rationales (explainability)}, i.e., which phrases are being responsible for annotating the post as a bully. The proposed multitask framework (mExCB) based on CNN and GRU with word and sub-sentence (SS) level attention is able to outperform several baselines and state of the art models when applied on {\em BullyExplain} dataset.