Debanjan Mahata

Jishnu Ray Chowdhury, Debanjan Mahata, Cornelia Caragea

We study the task of predicting a set of salient questions from a given paragraph without any prior knowledge of the precise answer. We make two main contributions. First, we propose a new method to evaluate a set of predicted questions against the set of references by using the Hungarian algorithm to assign predicted questions to references before scoring the assigned pairs. We show that our proposed evaluation strategy has better theoretical and practical properties compared to prior methods because it can properly account for the coverage of references. Second, we compare different strategies to utilize a pre-trained seq2seq model to generate and select a set of questions related to a given paragraph. The code is available.

Debanjan Mahata, Navneet Agarwal, Dibya Gautam et al.

Identifying keyphrases (KPs) from text documents is a fundamental task in natural language processing and information retrieval. Vast majority of the benchmark datasets for this task are from the scientific domain containing only the document title and abstract information. This limits keyphrase extraction (KPE) and keyphrase generation (KPG) algorithms to identify keyphrases from human-written summaries that are often very short (approx 8 sentences). This presents three challenges for real-world applications: human-written summaries are unavailable for most documents, the documents are almost always long, and a high percentage of KPs are directly found beyond the limited context of title and abstract. Therefore, we release two extensive corpora mapping KPs of ~1.3M and ~100K scientific articles with their fully extracted text and additional metadata including publication venue, year, author, field of study, and citations for facilitating research on this real-world problem.

Sarthak Anand, Debanjan Mahata, Kartik Aggarwal et al.

In this paper we present our approach and the system description for Sub Task A of SemEval 2019 Task 9: Suggestion Mining from Online Reviews and Forums. Given a sentence, the task asks to predict whether the sentence consists of a suggestion or not. Our model is based on Universal Language Model Fine-tuning for Text Classification. We apply various pre-processing techniques before training the language and the classification model. We further provide detailed analysis of the results obtained using the trained model. Our team ranked 10th out of 34 participants, achieving an F1 score of 0.7011. We publicly share our implementation at https://github.com/isarth/SemEval9_MIDAS

Jaemin Cho, Debanjan Mahata, Ozan Irsoy et al. · allen-ai

Document visual question answering (DocVQA) pipelines that answer questions from documents have broad applications. Existing methods focus on handling single-page documents with multi-modal language models (MLMs), or rely on text-based retrieval-augmented generation (RAG) that uses text extraction tools such as optical character recognition (OCR). However, there are difficulties in applying these methods in real-world scenarios: (a) questions often require information across different pages or documents, where MLMs cannot handle many long documents; (b) documents often have important information in visual elements such as figures, but text extraction tools ignore them. We introduce M3DocRAG, a novel multi-modal RAG framework that flexibly accommodates various document contexts (closed-domain and open-domain), question hops (single-hop and multi-hop), and evidence modalities (text, chart, figure, etc.). M3DocRAG finds relevant documents and answers questions using a multi-modal retriever and an MLM, so that it can efficiently handle single or many documents while preserving visual information. Since previous DocVQA datasets ask questions in the context of a specific document, we also present M3DocVQA, a new benchmark for evaluating open-domain DocVQA over 3,000+ PDF documents with 40,000+ pages. In three benchmarks (M3DocVQA/MMLongBench-Doc/MP-DocVQA), empirical results show that M3DocRAG with ColPali and Qwen2-VL 7B achieves superior performance than many strong baselines, including state-of-the-art performance in MP-DocVQA. We provide comprehensive analyses of different indexing, MLMs, and retrieval models. Lastly, we qualitatively show that M3DocRAG can successfully handle various scenarios, such as when relevant information exists across multiple pages and when answer evidence only exists in images.

Roberto Martínez-Cruz, Debanjan Mahata, Alvaro J. López-López et al.

In this study, we investigate using graph neural network (GNN) representations to enhance contextualized representations of pre-trained language models (PLMs) for keyphrase extraction from lengthy documents. We show that augmenting a PLM with graph embeddings provides a more comprehensive semantic understanding of words in a document, particularly for long documents. We construct a co-occurrence graph of the text and embed it using a graph convolutional network (GCN) trained on the task of edge prediction. We propose a graph-enhanced sequence tagging architecture that augments contextualized PLM embeddings with graph representations. Evaluating on benchmark datasets, we demonstrate that enhancing PLMs with graph embeddings outperforms state-of-the-art models on long documents, showing significant improvements in F1 scores across all the datasets. Our study highlights the potential of GNN representations as a complementary approach to improve PLM performance for keyphrase extraction from long documents.

Mayank Kulkarni, Debanjan Mahata, Ravneet Arora et al.

In this work, we explore how to train task-specific language models aimed towards learning rich representation of keyphrases from text documents. We experiment with different masking strategies for pre-training transformer language models (LMs) in discriminative as well as generative settings. In the discriminative setting, we introduce a new pre-training objective - Keyphrase Boundary Infilling with Replacement (KBIR), showing large gains in performance (upto 8.16 points in F1) over SOTA, when the LM pre-trained using KBIR is fine-tuned for the task of keyphrase extraction. In the generative setting, we introduce a new pre-training setup for BART - KeyBART, that reproduces the keyphrases related to the input text in the CatSeq format, instead of the denoised original input. This also led to gains in performance (upto 4.33 points in F1@M) over SOTA for keyphrase generation. Additionally, we also fine-tune the pre-trained language models on named entity recognition (NER), question answering (QA), relation extraction (RE), abstractive summarization and achieve comparable performance with that of the SOTA, showing that learning rich representation of keyphrases is indeed beneficial for many other fundamental NLP tasks.

Rakesh Gosangi, Ravneet Arora, Mohsen Gheisarieha et al.

In this paper, we study the importance of context in predicting the citation worthiness of sentences in scholarly articles. We formulate this problem as a sequence labeling task solved using a hierarchical BiLSTM model. We contribute a new benchmark dataset containing over two million sentences and their corresponding labels. We preserve the sentence order in this dataset and perform document-level train/test splits, which importantly allows incorporating contextual information in the modeling process. We evaluate the proposed approach on three benchmark datasets. Our results quantify the benefits of using context and contextual embeddings for citation worthiness. Lastly, through error analysis, we provide insights into cases where context plays an essential role in predicting citation worthiness.

Laiba Mehnaz, Debanjan Mahata, Rakesh Gosangi et al.

Code-switching is the communication phenomenon where speakers switch between different languages during a conversation. With the widespread adoption of conversational agents and chat platforms, code-switching has become an integral part of written conversations in many multi-lingual communities worldwide. This makes it essential to develop techniques for summarizing and understanding these conversations. Towards this objective, we introduce abstractive summarization of Hindi-English code-switched conversations and develop the first code-switched conversation summarization dataset - GupShup, which contains over 6,831 conversations in Hindi-English and their corresponding human-annotated summaries in English and Hindi-English. We present a detailed account of the entire data collection and annotation processes. We analyze the dataset using various code-switching statistics. We train state-of-the-art abstractive summarization models and report their performances using both automated metrics and human evaluation. Our results show that multi-lingual mBART and multi-view seq2seq models obtain the best performances on the new dataset

Yaman Kumar, Swati Aggarwal, Debanjan Mahata et al.

In the era of MOOCs, online exams are taken by millions of candidates, where scoring short answers is an integral part. It becomes intractable to evaluate them by human graders. Thus, a generic automated system capable of grading these responses should be designed and deployed. In this paper, we present a fast, scalable, and accurate approach towards automated Short Answer Scoring (SAS). We propose and explain the design and development of a system for SAS, namely AutoSAS. Given a question along with its graded samples, AutoSAS can learn to grade that prompt successfully. This paper further lays down the features such as lexical diversity, Word2Vec, prompt, and content overlap that plays a pivotal role in building our proposed model. We also present a methodology for indicating the factors responsible for scoring an answer. The trained model is evaluated on an extensively used public dataset, namely Automated Student Assessment Prize Short Answer Scoring (ASAP-SAS). AutoSAS shows state-of-the-art performance and achieves better results by over 8% in some of the question prompts as measured by Quadratic Weighted Kappa (QWK), showing performance comparable to humans.

Sarthak Anand, Pradyumna Gupta, Hemant Yadav et al.

This paper presents our submission to the SemEval 2020 - Task 10 on emphasis selection in written text. We approach this emphasis selection problem as a sequence labeling task where we represent the underlying text with various contextual embedding models. We also employ label distribution learning to account for annotator disagreements. We experiment with the choice of model architectures, trainability of layers, and different contextual embeddings. Our best performing architecture is an ensemble of different models, which achieved an overall matching score of 0.783, placing us 15th out of 31 participating teams. Lastly, we analyze the results in terms of parts of speech tags, sentence lengths, and word ordering.

Gyanesh Anand, Akash Gautam, Puneet Mathur et al.

Twitter is a social media platform where users express opinions over a variety of issues. Posts offering grievances or complaints can be utilized by private/ public organizations to improve their service and promptly gauge a low-cost assessment. In this paper, we propose an iterative methodology which aims to identify complaint based posts pertaining to the transport domain. We perform comprehensive evaluations along with releasing a novel dataset for the research purposes.

Akash Gautam, Puneet Mathur, Rakesh Gosangi et al.

In this paper, we present a dataset containing 9,973 tweets related to the MeToo movement that were manually annotated for five different linguistic aspects: relevance, stance, hate speech, sarcasm, and dialogue acts. We present a detailed account of the data collection and annotation processes. The annotations have a very high inter-annotator agreement (0.79 to 0.93 k-alpha) due to the domain expertise of the annotators and clear annotation instructions. We analyze the data in terms of geographical distribution, label correlations, and keywords. Lastly, we present some potential use cases of this dataset. We expect this dataset would be of great interest to psycholinguists, socio-linguists, and computational linguists to study the discursive space of digitally mobilized social movements on sensitive issues like sexual harassment.

Dhruva Sahrawat, Debanjan Mahata, Mayank Kulkarni et al.

In this paper, we formulate keyphrase extraction from scholarly articles as a sequence labeling task solved using a BiLSTM-CRF, where the words in the input text are represented using deep contextualized embeddings. We evaluate the proposed architecture using both contextualized and fixed word embedding models on three different benchmark datasets (Inspec, SemEval 2010, SemEval 2017) and compare with existing popular unsupervised and supervised techniques. Our results quantify the benefits of (a) using contextualized embeddings (e.g. BERT) over fixed word embeddings (e.g. Glove); (b) using a BiLSTM-CRF architecture with contextualized word embeddings over fine-tuning the contextualized word embedding model directly, and (c) using genre-specific contextualized embeddings (SciBERT). Through error analysis, we also provide some insights into why particular models work better than others. Lastly, we present a case study where we analyze different self-attention layers of the two best models (BERT and SciBERT) to better understand the predictions made by each for the task of keyphrase extraction.

Yaman Kumar, Debanjan Mahata, Sagar Aggarwal et al.

In this paper, we introduce the first and largest Hindi text corpus, named BHAAV, which means emotions in Hindi, for analyzing emotions that a writer expresses through his characters in a story, as perceived by a narrator/reader. The corpus consists of 20,304 sentences collected from 230 different short stories spanning across 18 genres such as Inspirational and Mystery. Each sentence has been annotated into one of the five emotion categories - anger, joy, suspense, sad, and neutral, by three native Hindi speakers with at least ten years of formal education in Hindi. We also discuss challenges in the annotation of low resource languages such as Hindi, and discuss the scope of the proposed corpus along with its possible uses. We also provide a detailed analysis of the dataset and train strong baseline classifiers reporting their performances.

Avinash Swaminathan, Raj Kuwar Gupta, Haimin Zhang et al.

In this paper, we present a keyphrase generation approach using conditional Generative Adversarial Networks (GAN). In our GAN model, the generator outputs a sequence of keyphrases based on the title and abstract of a scientific article. The discriminator learns to distinguish between machine-generated and human-curated keyphrases. We evaluate this approach on standard benchmark datasets. Our model achieves state-of-the-art performance in generation of abstractive keyphrases and is also comparable to the best performing extractive techniques. We also demonstrate that our method generates more diverse keyphrases and make our implementation publicly available.

Nilay Shrivastava, Astitwa Saxena, Yaman Kumar et al.

Visual speech recognition (VSR) is the task of recognizing spoken language from video input only, without any audio. VSR has many applications as an assistive technology, especially if it could be deployed in mobile devices and embedded systems. The need of intensive computational resources and large memory footprint are two of the major obstacles in developing neural network models for VSR in a resource constrained environment. We propose a novel end-to-end deep neural network architecture for word level VSR called MobiVSR with a design parameter that aids in balancing the model's accuracy and parameter count. We use depthwise-separable 3D convolution for the first time in the domain of VSR and show how it makes our model efficient. MobiVSR achieves an accuracy of 73\% on a challenging Lip Reading in the Wild dataset with 6 times fewer parameters and 20 times lesser memory footprint than the current state of the art. MobiVSR can also be compressed to 6 MB by applying post training quantization.

Haimin Zhang, Debanjan Mahata, Simra Shahid et al.

In this paper we present our approach and the system description for Sub-task A and Sub Task B of SemEval 2019 Task 6: Identifying and Categorizing Offensive Language in Social Media. Sub-task A involves identifying if a given tweet is offensive or not, and Sub Task B involves detecting if an offensive tweet is targeted towards someone (group or an individual). Our models for Sub-task A is based on an ensemble of Convolutional Neural Network, Bidirectional LSTM with attention, and Bidirectional LSTM + Bidirectional GRU, whereas for Sub-task B, we rely on a set of heuristics derived from the training data and manual observation. We provide detailed analysis of the results obtained using the trained models. Our team ranked 5th out of 103 participants in Sub-task A, achieving a macro F1 score of 0.807, and ranked 8th out of 75 participants in Sub Task B achieving a macro F1 of 0.695.

Yaman Kumar, Dhruva Sahrawat, Shubham Maheshwari et al.

Visual Speech Recognition (VSR) is the process of recognizing or interpreting speech by watching the lip movements of the speaker. Recent machine learning based approaches model VSR as a classification problem; however, the scarcity of training data leads to error-prone systems with very low accuracies in predicting unseen classes. To solve this problem, we present a novel approach to zero-shot learning by generating new classes using Generative Adversarial Networks (GANs), and show how the addition of unseen class samples increases the accuracy of a VSR system by a significant margin of 27% and allows it to handle speaker-independent out-of-vocabulary phrases. We also show that our models are language agnostic and therefore capable of seamlessly generating, using English training data, videos for a new language (Hindi). To the best of our knowledge, this is the first work to show empirical evidence of the use of GANs for generating training samples of unseen classes in the domain of VSR, hence facilitating zero-shot learning. We make the added videos for new classes publicly available along with our code.

Nupur Baghel, Yaman Kumar, Paavini Nanda et al.

There has been upsurge in the number of people participating in challenges made popular through social media channels. One of the examples of such a challenge is the Kiki Challenge, in which people step out of their moving cars and dance to the tunes of the song, 'Kiki, Do you love me?'. Such an action makes the people taking the challenge prone to accidents and can also create nuisance for the others traveling on the road. In this work, we introduce the prevalence of such challenges in social media and show how the machine learning community can aid in preventing dangerous situations triggered by them by developing models that can distinguish between dangerous and non-dangerous challenge videos. Towards this objective, we release a new dataset namely MIDAS-KIKI dataset, consisting of manually annotated dangerous and non-dangerous Kiki challenge videos. Further, we train a deep learning model to identify dangerous and non-dangerous videos, and report our results.

Debanjan Mahata, Jasper Friedrichs, Rajiv Ratn Shah et al.

Mining social media messages such as tweets, articles, and Facebook posts for health and drug related information has received significant interest in pharmacovigilance research. Social media sites (e.g., Twitter), have been used for monitoring drug abuse, adverse reactions of drug usage and analyzing expression of sentiments related to drugs. Most of these studies are based on aggregated results from a large population rather than specific sets of individuals. In order to conduct studies at an individual level or specific cohorts, identifying posts mentioning intake of medicine by the user is necessary. Towards this objective we develop a classifier for identifying mentions of personal intake of medicine in tweets. We train a stacked ensemble of shallow convolutional neural network (CNN) models on an annotated dataset. We use random search for tuning the hyper-parameters of the CNN models and present an ensemble of best models for the prediction task. Our system produces state-of-the-art result, with a micro-averaged F-score of 0.693. We believe that the developed classifier has direct uses in the areas of psychology, health informatics, pharmacovigilance and affective computing for tracking moods, emotions and sentiments of patients expressing intake of medicine in social media.

Debanjan Mahata, John Kuriakose, Rajiv Ratn Shah et al.

Keyword extraction is a fundamental task in natural language processing that facilitates mapping of documents to a concise set of representative single and multi-word phrases. Keywords from text documents are primarily extracted using supervised and unsupervised approaches. In this paper, we present an unsupervised technique that uses a combination of theme-weighted personalized PageRank algorithm and neural phrase embeddings for extracting and ranking keywords. We also introduce an efficient way of processing text documents and training phrase embeddings using existing techniques. We share an evaluation dataset derived from an existing dataset that is used for choosing the underlying embedding model. The evaluations for ranked keyword extraction are performed on two benchmark datasets comprising of short abstracts (Inspec), and long scientific papers (SemEval 2010), and is shown to produce results better than the state-of-the-art systems.

Mayank Meghawat, Satyendra Yadav, Debanjan Mahata et al.

Multiple modalities represent different aspects by which information is conveyed by a data source. Modern day social media platforms are one of the primary sources of multimodal data, where users use different modes of expression by posting textual as well as multimedia content such as images and videos for sharing information. Multimodal information embedded in such posts could be useful in predicting their popularity. To the best of our knowledge, no such multimodal dataset exists for the prediction of social media photos. In this work, we propose a multimodal dataset consisiting of content, context, and social information for popularity prediction. Specifically, we augment the SMPT1 dataset for social media prediction in ACM Multimedia grand challenge 2017 with image content, titles, descriptions, and tags. Next, in this paper, we propose a multimodal approach which exploits visual features (i.e., content information), textual features (i.e., contextual information), and social features (e.g., average views and group counts) to predict popularity of social media photos in terms of view counts. Experimental results confirm that despite our multimodal approach uses the half of the training dataset from SMP-T1, it achieves comparable performance with that of state-of-the-art.

Debanjan Mahata, Jasper Friedrichs, Hitkul et al.

Mining social media messages for health and drug related information has received significant interest in pharmacovigilance research. Social media sites (e.g., Twitter), have been used for monitoring drug abuse, adverse reactions of drug usage and analyzing expression of sentiments related to drugs. Most of these studies are based on aggregated results from a large population rather than specific sets of individuals. In order to conduct studies at an individual level or specific cohorts, identifying posts mentioning intake of medicine by the user is necessary. Towards this objective, we train different deep neural network classification models on a publicly available annotated dataset and study their performances on identifying mentions of personal intake of medicine in tweets. We also design and train a new architecture of a stacked ensemble of shallow convolutional neural network (CNN) ensembles. We use random search for tuning the hyperparameters of the models and share the details of the values taken by the hyperparameters for the best learnt model in different deep neural network architectures. Our system produces state-of-the-art results, with a micro- averaged F-score of 0.693.

Jasper Friedrichs, Debanjan Mahata, Shubham Gupta

This paper describes Infosys's participation in the "2nd Social Media Mining for Health Applications Shared Task at AMIA, 2017, Task 2". Mining social media messages for health and drug related information has received significant interest in pharmacovigilance research. This task targets at developing automated classification models for identifying tweets containing descriptions of personal intake of medicines. Towards this objective we train a stacked ensemble of shallow convolutional neural network (CNN) models on an annotated dataset provided by the organizers. We use random search for tuning the hyper-parameters of the CNN and submit an ensemble of best models for the prediction task. Our system secured first place among 9 teams, with a micro-averaged F-score of 0.693.