Pre-trained Generative models such as BART, T5, etc. have gained prominence as a preferred method for text generation in various natural language processing tasks, including abstractive long-form question answering (QA) and summarization. However, the potential of generative models in extractive QA tasks, where discriminative models are commonly employed, remains largely unexplored. Discriminative models often encounter challenges associated with label sparsity, particularly when only a small portion of the context contains the answer. The challenge is more pronounced for multi-span answers. In this work, we introduce a novel approach that uses the power of pre-trained generative models to address extractive QA tasks by generating indexes corresponding to context tokens or sentences that form part of the answer. Through comprehensive evaluations on multiple extractive QA datasets, including MultiSpanQA, BioASQ, MASHQA, and WikiQA, we demonstrate the superior performance of our proposed approach compared to existing state-of-the-art models.

Extracting relational triples from text is a crucial task for constructing knowledge bases. Recent advancements in joint entity and relation extraction models have demonstrated remarkable F1 scores (≥ 90%) in accurately extracting relational triples from free text. However, these models have been evaluated under restrictive experimental settings and unrealistic datasets. They overlook sentences with zero triples (zerocardinality), thereby simplifying the task. In this paper, we present a benchmark study of state-of-the-art joint entity and relation extraction models under a more realistic setting. We include sentences that lack any triples in our experiments, providing a comprehensive evaluation. Our findings reveal a significant decline (approximately 10-15% in one dataset and 6-14% in another dataset) in the models’ F1 scores within this realistic experimental setup. Furthermore, we propose a two-step modeling approach that utilizes a simple BERT-based classifier. This approach leads to overall performance improvement in these models within the realistic experimental setting.

When answering natural language questions over knowledge bases, missing facts, incomplete schema and limited scope naturally lead to many questions being unanswerable. While answerability has been explored in other QA settings, it has not been studied for QA over knowledge bases (KBQA). We create GrailQAbility, a new benchmark KBQA dataset with unanswerability, by first identifying various forms of KB incompleteness that make questions unanswerable, and then systematically adapting GrailQA (a popular KBQA dataset with only answerable questions). Experimenting with three state-of-the-art KBQA models, we find that all three models suffer a drop in performance even after suitable adaptation for unanswerable questions. In addition, these often detect unanswerability for wrong reasons and find specific forms of unanswerability particularly difficult to handle. This underscores the need for further research in making KBQA systems robust to unanswerability.

We explore the task of automated generation of technical interview questions from a given textbook. Such questions are different from those for reading comprehension studied in question generation literature. We curate a context based interview questions data set for Machine Learning and Deep Learning from two popular textbooks. We first explore the possibility of using a large generative language model (GPT-3) for this task in a zero shot setting. We then evaluate the performance of smaller generative models such as BART fine-tuned on weakly supervised data obtained using GPT-3 and hand-crafted templates. We deploy an automatic question importance assignment technique to figure out suitability of a question in a technical interview. It improves the evaluation results in many dimensions. We dissect the performance of these models for this task and also scrutinize the suitability of questions generated by them for use in technical interviews.

We explore the task of generating long-form technical questions from textbooks. Semi-structured metadata of a textbook — the table of contents and the index — provide rich cues for technical question generation. Existing literature for long-form question generation focuses mostly on reading comprehension assessment, and does not use semi-structured metadata for question generation. We design unsupervised template based algorithms for generating questions based on structural and contextual patterns in the index and ToC. We evaluate our approach on textbooks on diverse subjects and show that our approach generates high quality questions of diverse types. We show that, in comparison, zero-shot question generation using pre-trained LLMs on the same meta-data has much poorer quality.

Predicting difficulty of questions is crucial for technical interviews. However, such questions are long-form and more open-ended than factoid and multiple choice questions explored so far for question difficulty prediction. Existing models also require large volumes of candidate response data for training. We study weak-supervision and use unsupervised algorithms for both question generation and difficulty prediction. We create a dataset of interview questions with difficulty scores for deep learning and use it to evaluate SOTA models for question difficulty prediction trained using weak supervision. Our analysis brings out the task’s difficulty as well as the promise of weak supervision for it.

Given the diversity of the candidates and complexity of job requirements, and since interviewing is an inherently subjective process, it is an important task to ensure consistent, uniform, efficient and objective interviews that result in high quality recruitment. We propose an interview assistant system to automatically, and in an objective manner, select an optimal set of technical questions (from question banks) personalized for a candidate. This set can help a human interviewer to plan for an upcoming interview of that candidate. We formalize the problem of selecting a set of questions as an integer linear programming problem and use standard solvers to get a solution. We use knowledge graph as background knowledge in this formulation, and derive our objective functions and constraints from it. We use candidate’s resume to personalize the selection of questions. We propose an intrinsic evaluation to compare a set of suggested questions with actually asked questions. We also use expert interviewers to comparatively evaluate our approach with a set of reasonable baselines.

Existing approaches for table annotation with entities and types either capture the structure of table using graphical models, or learn embeddings of table entries without accounting for the complete syntactic structure. We propose TabGCN, that uses Graph Convolutional Networks to capture the complete structure of tables, knowledge graph and the training annotations, and jointly learns embeddings for table elements as well as the entities and types. To account for knowledge incompleteness, TabGCN’s embeddings can be used to discover new entities and types. Using experiments on 5 benchmark datasets, we show that TabGCN significantly outperforms multiple state-of-the-art baselines for table annotation, while showing promising performance on downstream table-related applications.

Question answering (QA) over a knowledge graph (KG) is a task of answering a natural language (NL) query using the information stored in KG. In a real-world industrial setting, this involves addressing multiple challenges including entity linking, multi-hop reasoning over KG, etc. Traditional approaches handle these challenges in a modularized sequential manner where errors in one module lead to the accumulation of errors in downstream modules. Often these challenges are inter-related and the solutions to them can reinforce each other when handled simultaneously in an end-to-end learning setup. To this end, we propose a multi-task BERT based Neural Machine Translation (NMT) model to address these challenges. Through experimental analysis, we demonstrate the efficacy of our proposed approach on one publicly available and one proprietary dataset.

We study the problem of schema discovery for knowledge graphs. We propose a solution where an agent engages in multi-turn dialog with an expert for this purpose. Each mini-dialog focuses on a short natural language statement, and looks to elicit the expert’s desired schema-based interpretation of that statement, taking into account possible augmentations to the schema. The overall schema evolves by performing dialog over a collection of such statements. We take into account the probability that the expert does not respond to a query, and model this probability as a function of the complexity of the query. For such mini-dialogs with response uncertainty, we propose a dialog strategy that looks to elicit the schema over as short a dialog as possible. By combining the notion of uncertainty sampling from active learning with generalized binary search, the strategy asks the query with the highest expected reduction of entropy. We show that this significantly reduces dialog complexity while engaging the expert in meaningful dialog.

Recent work has addressed the problem of detecting relevant claims for a given controversial topic. We introduce the complementary task of Claim Stance Classification, along with the first benchmark dataset for this task. We decompose this problem into: (a) open-domain target identification for topic and claim (b) sentiment classification for each target, and (c) open-domain contrast detection between the topic and the claim targets. Manual annotation of the dataset confirms the applicability and validity of our model. We describe an implementation of our model, focusing on a novel algorithm for contrast detection. Our approach achieves promising results, and is shown to outperform several baselines, which represent the common practice of applying a single, monolithic classifier for stance classification.