Accepted Publications

Ramit Sawhney*, Megh Thakkar*, Shrey Pandit*, Ritesh Singh Soun, Di Jin, Diyi Yang, Lucie Flek (2022).

DMix: Adaptive Distance-aware Interpolative Mixup [ACL-2022 Main Conference]

Victor Sanh, Albert Webson, Colin Raffel, Stephen H Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal Nayak, Debajyoti Datta, Jonathan Chang, Mike Tian-Jian Jiang, Han Wang, Matteo Manica, Sheng Shen, Zheng Xin Yong, Harshit Pandey, Rachel Bawden, Thomas Wang, Trishala Neeraj, Jos Rozen, Abheesht Sharma, Andrea Santilli, Thibault Fevry, Jason Alan Fries, Ryan Teehan, Stella Biderman, Leo Gao, Tali Bers, Thomas Wolf, Alexander M Rush (2021). Multitask Prompted Training Enables Zero-Shot Task Generalization [ICLR-2022]

Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable, prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely unseen tasks. We finetune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All prompts and trained models are available at this https URL and this https URL.

Swapnil Parekh*, Yaman Singla Kumar*, Somesh Singh*, Changyou Chen, Balaji Krishnamurthy, Rajiv Ratn Shah (2021). MINIMAL: Mining Models for Data Free Universal Adversarial Triggers [AAAI-2022]

It is well known that natural language models are vulnerable to adversarial attacks, which are mostly input-specific in nature. Recently, it has been shown that there also exist input-agnostic attacks in NLP models, called universal adversarial triggers. However, existing methods to craft universal triggers are data intensive. They require large amounts of data samples to generate adversarial triggers, which are typically inaccessible by attackers. For instance, previous works take 3000 data samples per class for the SNLI dataset to generate adversarial triggers. In this paper, we present a novel data-free approach, MINIMAL, to mine input-agnostic adversarial triggers from models. Using the triggers produced with our data-free algorithm, we reduce the accuracy of Stanford Sentiment Treebank's positive class from 93.6% to 9.6%. Similarly, for the Stanford Natural Language Inference (SNLI), our single-word trigger reduces the accuracy of the entailment class from 90.95% to less than 0.6\%. Despite being completely data-free, we get equivalent accuracy drops as data-dependent methods.

Abheesht Sharma*, Gunjan Chhablani*, Harshit Pandey*, Rajaswa Patil; (2021). DRIFT: A Toolkit for Diachronic Analysis of Scientific Literature [EMNLP 2021]

In this work, we present to the NLP community, and to the wider research community as a whole, an application for the diachronic analysis of research corpora. We open source an easy-to-use tool coined DRIFT, which allows researchers to track research trends and development over the years. The analysis methods are collated from well-cited research works, with a few of our own methods added for good measure. Succinctly put, some of the analysis methods are: keyword extraction, word clouds, predicting declining/stagnant/growing trends using Productivity, tracking bi-grams using Acceleration plots, finding the Semantic Drift of words, tracking trends using similarity, etc. To demonstrate the utility and efficacy of our tool, we perform a case study on the cs.CL corpus of the arXiv repository and draw inferences from the analysis methods. The toolkit and the associated code are available here:

Rajaswa Patil, Jasleen Dhillon, Siddhant Mahurkar, Saumitra Kulkarni, Manav Malhotra, Veeky Baths; (2021). Vyākarana: A Colorless Green Benchmark for Syntactic Evaluation in Indic Languages [MRL Workshop, EMNLP 2021]

While there has been significant progress towards developing NLU resources for Indic languages, syntactic evaluation has been relatively less explored. Unlike English, Indic languages have rich morphosyntax, grammatical genders, free linear word-order, and highly inflectional morphology. In this paper, we introduce Vyākarana: a benchmark of Colorless Green sentences in Indic languages for syntactic evaluation of multilingual language models. The benchmark comprises four syntax-related tasks: PoS Tagging, Syntax Tree-depth Prediction, Grammatical Case Marking, and Subject-Verb Agreement. We use the datasets from the evaluation tasks to probe five multilingual language models of varying architectures for syntax in Indic languages. Due to its prevalence, we also include a code-switching setting in our experiments. Our results show that the token-level and sentence-level representations from the Indic language models (IndicBERT and MuRIL) do not capture the syntax in Indic languages as efficiently as the other highly multilingual language models. Further, our layer-wise probing experiments reveal that while mBERT, DistilmBERT, and XLM-R localize the syntax in middle layers, the Indic language models do not show such syntactic localization.

Ramit Sawhney, Megh Thakkar, Shrey Pandit, Debdoot Mukherjee, Lucie Flek; (2021). DMix: Distance Constrained Interpolative Mixup [MRL Workshop, EMNLP 2021]

Interpolation-based regularisation methods have proven to be effective for various tasks and modalities. Mixup is a data augmentation method that generates virtual training samples from convex combinations of individual inputs and labels. We extend Mixup and propose DMix, distance-constrained interpolative Mixup for sentence classification leveraging the hyperbolic space. DMix achieves state-of-the-art results on sentence classification over existing data augmentation methods across datasets in four languages.

Megh Thakkar, Vishwa Shah, Ramit Sawhney, Debdoot Mukherjee; (2021). Sequence Mixup for Zero-Shot Cross-Lingual Part-Of-Speech Tagging [MRL Workshop, EMNLP 2021]

There have been efforts in cross-lingual transfer learning for various tasks. We present an approach utilizing an interpolative data augmentation method, Mixup, to improve the generalizability of models for part-of-speech tagging trained on a source language, improving its performance on unseen target languages. Through experiments on ten languages with diverse structures and language roots, we put forward its applicability for downstream zero-shot cross-lingual tasks.

Abheesht Sharma*, Harshit Pandey*, Gunjan Chhablani*, Yash Bhartia*, Tirtharaj Dash; (2021). LRG at SemEval-2021 Task 4: Improving Reading Comprehension with Abstract Words using Augmentation, Linguistic Features and Voting [SemEval Workshop, ACL-IJCNLP 2021]

In this article, we present our methodologies for SemEval-2021 Task-4: Reading Comprehension of Abstract Meaning. Given a fill-in-the-blank-type question and a corresponding context, the task is to predict the most suitable word from a list of 5 options. There are three sub-tasks within this task: Imperceptibility (subtask-I), Non-Specificity (subtask-II), and Intersection (subtask-III). We use encoders of transformers-based models pre-trained on the masked language modelling (MLM) task to build our Fill-in-the-blank (FitB) models. Moreover, to model imperceptibility, we define certain linguistic features, and to model non-specificity, we leverage information from hypernyms and hyponyms provided by a lexical database. Specifically, for non-specificity, we try out augmentation techniques, and other statistical techniques. We also propose variants, namely Chunk Voting and Max Context, to take care of input length restrictions for BERT, etc. Additionally, we perform a thorough ablation study, and use Integrated Gradients to explain our predictions on a few samples. Our best submissions achieve accuracies of 75.31% and 77.84%, on the test sets for subtask-I and subtask-II, respectively. For subtask-III, we achieve accuracies of 65.64% and 62.27%.

Gunjan Chhablani*, Yash Bhartia*, Abheesht Sharma*, Harshit Pandey*, Shan Suthaharan; (2021). NLRG at SemEval-2021 Task 5: Toxic Spans Detection Leveraging BERT-based Token Classification and Span Prediction and Span Prediction Techniques [SemEval Workshop, ACL-IJCNLP 2021]

Toxicity detection of text has been a popular NLP task in the recent years. In SemEval-2021 Task-5 Toxic Spans Detection, the focus is on detecting toxic spans within passages. Most state-of-the-art span detection approaches employ various techniques, each of which can be broadly classified into Token Classification or Span Prediction approaches. In our paper, we explore simple versions of both of these approaches and their performance on the task. Specifically, we use BERT-based models -- BERT, RoBERTa, and SpanBERT for both approaches. We also combine these approaches and modify them to bring improvements for Toxic Spans prediction. To this end, we investigate results on four hybrid approaches -- Multi-Span, Span+Token, LSTM-CRF, and a combination of predicted offsets using union/intersection. Additionally, we perform a thorough ablative analysis and analyze our observed results. Our best submission -- a combination of SpanBERT Span Predictor and RoBERTa Token Classifier predictions -- achieves an F1 score of 0.6753 on the test set. Our best post-eval F1 score is 0.6895 on intersection of predicted offsets from top-3 RoBERTa Token Classification checkpoints. These approaches improve the performance by 3% on average than those of the shared baseline models -- RNNSL and SpaCy NER.

Abheesht Sharma*, Harshit Pandey*; (2020). LRG at TREC 2020: Document Ranking with XLNet-Based Models [TREC 2020]

Establishing a good information retrieval system in popular mediums of entertainment is a quickly growing area of investigation for companies and researchers alike. We delve into the domain of information retrieval for podcasts. In Spotify’s Podcast Challenge, we are given a user’s query with a description to find the most relevant short segment from the given dataset having all the podcasts. Previous techniques that include solely classical Information Retrieval (IR) techniques, perform poorly when descriptive queries are presented. On the other hand, models which exclusively rely on large neural networks tend to perform better. The downside to this technique is that a considerable amount of time and computing power are required to infer the result. We experiment with two hybrid models which first filter out the best podcasts based on user’s query with a classical IR technique, and then perform re-ranking on the shortlisted documents based on the detailed description using a transformer-based model.

Propaganda spreads the ideology and beliefs of like-minded people, brainwashing their audiences, and sometimes leading to violence. SemEval 2020 Task-11 aims to design automated systems for news propaganda detection. Task-11 consists of two sub-tasks, namely, Span Identification - given any news article, the system tags those specific fragments which contain at least one propaganda technique; and Technique Classification - correctly classify a given propagandist statement amongst 14 propaganda techniques. For sub-task 1, we use contextual embeddings extracted from pre-trained transformer models to represent the text data at various granularities and propose a multi-granularity knowledge sharing approach. For sub-task 2, we use an ensemble of BERT and logistic regression classifiers with linguistic features. Our results reveal that the linguistic features are the strong indicators for covering minority classes in a highly imbalanced dataset.

In this paper, we assess the ability of BERT and its derivative models (RoBERTa, DistilBERT, and ALBERT) for short-edits based humor grading. We test these models for humor grading and classification tasks on the Humicroedit and the FunLines dataset. We perform extensive experiments with these models to test their language modeling and generalization abilities via zero-shot inference and cross-dataset inference based approaches. Further, we also inspect the role of self-attention layers in humor-grading by performing a qualitative analysis over the self-attention weights from the final layer of the trained BERT model. Our experiments show that all the pre-trained BERT derivative models show significant generalization capabilities for humor-grading related tasks.

In this paper, we describe an approach for modelling causal reasoning in natural language by detecting counterfactuals in text using multi-head self-attention weights. We use pre-trained transformer models to extract contextual embeddings and self-attention weights from the text. We show the use of convolutional layers to extract task-specific features from these self-attention weights. Further, we describe a fine-tuning approach with a common base model for knowledge sharing between the two closely related sub-tasks for counterfactual detection. We analyze and compare the performance of various transformer models in our experiments. Finally, we perform a qualitative analysis with the multi-head self-attention weights to interpret our models’ dynamics.

Shashank Madhusudhan*, Siddhant Mahurkar*, Suresh Kumar Nagarajan; (2020). Attributional analysis of Multi-Modal Fake News Detection Models (Grand Challenge) [IEEE BigMM 2020]

Fake news detection is a procedure for identifying a particular news article as counterfeit or real. In this paper, we propose and assess the ability of two approaches for the task of multi-modal fake news detection. For the first approach, we fuse the textual and image modalities. The textual features are obtained from the pre-trained language models such as BERT and SBERT and image features are extracted from ResNet-18 pre-trained on ImageNet dataset. In the second approach, we use Visual Attention for fake news detection. We test both the strategies on Gossipcop and Politifact dataset. Our experiments show that the complete text of the article and the BERT model setting provides the best result. Further, we use Integrated gradients to analyze our models by observing input attributions.


Stephen H Bach, Victor Sanh, Zheng-Xin Yong, Albert Webson, Colin Raffel, Nihal V Nayak, Abheesht Sharma, Taewoon Kim, M Saiful Bari, Thibault Fevry, Zaid Alyafeai, Manan Dey, Andrea Santilli, Zhiqing Sun, Srulik Ben-David, Canwen Xu, Gunjan Chhablani, Han Wang, Jason Alan Fries, Maged S Al-shaibani, Shanya Sharma, Urmish Thakker, Khalid Almubarak, Xiangru Tang, Mike Tian-Jian Jiang, Alexander M Rush (2022). PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts [arXiv]

PromptSource is a system for creating, sharing, and using natural language prompts. Prompts are functions that map an example from a dataset to a natural language input and target output. Using prompts to train and query language models is an emerging area in NLP that requires new tools that let users develop and refine these prompts collaboratively. PromptSource addresses the emergent challenges in this new setting with (1) a templating language for defining data-linked prompts, (2) an interface that lets users quickly iterate on prompt development by observing outputs of their prompts on many examples, and (3) a community-driven set of guidelines for contributing new prompts to a common pool. Over 2,000 prompts for roughly 170 datasets are already available in PromptSource. PromptSource is available at this https URL.

Yaman Singla Kumar*, Swapnil Parekh*, Somesh Singh*, Junyi Jessy Li, Rajiv Ratn Shah, Changyou Chen (2021). AES Systems Are Both Overstable And Oversensitive: Explaining Why And Proposing Defenses [arXiv]

Deep-learning based Automatic Essay Scoring (AES) systems are being actively used by states and language testing agencies alike to evaluate millions of candidates for life-changing decisions ranging from college applications to visa approvals. However, little research has been put to understand and interpret the black-box nature of deep-learning based scoring algorithms. Previous studies indicate that scoring models can be easily fooled. In this paper, we explore the reason behind their surprising adversarial brittleness. We utilize recent advances in interpretability to find the extent to which features such as coherence, content, vocabulary, and relevance are important for automated scoring mechanisms. We use this to investigate the oversensitivity i.e., large change in output score with a little change in input essay content) and overstability i.e., little change in output scores with large changes in input essay content) of AES. Our results indicate that autoscoring models, despite getting trained as "end-to-end" models with rich contextual embeddings such as BERT, behave like bag-of-words models. A few words determine the essay score without the requirement of any context making the model largely overstable. This is in stark contrast to recent probing studies on pre-trained representation learning models, which show that rich linguistic features such as parts-of-speech and morphology are encoded by them. Further, we also find that the models have learnt dataset biases, making them oversensitive. To deal with these issues, we propose detection-based protection models that can detect oversensitivity and overstability causing samples with high accuracies. We find that our proposed models are able to detect unusual attribution patterns and flag adversarial samples successfully.

Rajaswa Patil*, Yaman Kumar Singla*, Rajiv Ratn Shah, Mika Hama, Roger Zimmermann; (2021). Towards Modelling Coherence in Spoken Discourse [arXiv]

While there has been significant progress towards modelling coherence in written discourse, the work in modelling spoken discourse coherence has been quite limited. Unlike the coherence in text, coherence in spoken discourse is also dependent on the prosodic and acoustic patterns in speech. In this paper, we model coherence in spoken discourse with audio-based coherence models. We perform experiments with four coherence-related tasks with spoken discourses. In our experiments, we evaluate machine-generated speech against the speech delivered by expert human speakers. We also compare the spoken discourses generated by human language learners of varying language proficiency levels. Our results show that incorporating the audio modality along with the text benefits the coherence models in performing downstream coherence related tasks with spoken discourses.