Gaurav Arora

Gaurav Arora, Srujana Merugu, Shreya Jain et al.

Reasoning and linguistic skills form the cornerstone of human intelligence, facilitating problem-solving and decision-making. Recent advances in Large Language Models (LLMs) have led to impressive linguistic capabilities and emergent reasoning behaviors, fueling widespread adoption across application domains. However, LLMs still struggle with complex reasoning tasks, highlighting their systemic limitations. In this work, we focus on evaluating whether LLMs have the requisite representations to reason using two foundational relationships: "equivalence" and "inheritance". We introduce novel tasks and benchmarks spanning six languages and observe that current SOTA LLMs often produce conflicting answers to the same questions across languages in 17.3-57.5% of cases and violate inheritance constraints in up to 37.2% cases. To enhance consistency across languages, we propose novel "Compositional Representations" where tokens are represented as composition of equivalent tokens across languages, with resulting conflict reduction (up to -4.7%) indicating benefits of shared LLM representations.

Mayank Kothyari, Sunita Sarawagi, Soumen Chakrabarti et al.

LLMs are increasingly used as seq2seq translators from natural language utterances to structured programs, a process called semantic interpretation. Unlike atomic labels or token sequences, programs are naturally represented as abstract syntax trees (ASTs). Such structured representation raises novel issues related to the design and selection of in-context examples (ICEs) presented to the LLM. We focus on decomposing the pool of available ICE trees into fragments, some of which may be better suited to solving the test instance. Next, we propose how to use (additional invocations of) an LLM with prompted syntax constraints to automatically map the fragments to corresponding utterances. Finally, we adapt and extend a recent method for diverse ICE selection to work with whole and fragmented ICE instances. We evaluate our system, SCUD4ICL, on popular diverse semantic parsing benchmarks, showing visible accuracy gains from our proposed decomposed diverse demonstration method. Benefits are particularly notable for smaller LLMs, ICE pools having larger labeled trees, and programs in lower resource languages.

Gaurav Arora

This paper describes the system submitted to Dravidian-Codemix-HASOC2020: Hate Speech and Offensive Content Identification in Dravidian languages (Tamil-English and Malayalam-English). The task aims to identify offensive language in code-mixed dataset of comments/posts in Dravidian languages collected from social media. We participated in both Sub-task A, which aims to identify offensive content in mixed-script (mixture of Native and Roman script) and Sub-task B, which aims to identify offensive content in Roman script, for Dravidian languages. In order to address these tasks, we proposed pre-training ULMFiT on synthetically generated code-mixed data, generated by modelling code-mixed data generation as a Markov process using Markov chains. Our model achieved 0.88 weighted F1-score for code-mixed Tamil-English language in Sub-task B and got 2nd rank on the leader-board. Additionally, our model achieved 0.91 weighted F1-score (4th Rank) for mixed-script Malayalam-English in Sub-task A and 0.74 weighted F1-score (5th Rank) for code-mixed Malayalam-English language in Sub-task B.

Gaurav Arora, Chirag Jain, Manas Chaturvedi et al.

Intent Detection systems in the real world are exposed to complexities of imbalanced datasets containing varying perception of intent, unintended correlations and domain-specific aberrations. To facilitate benchmarking which can reflect near real-world scenarios, we introduce 3 new datasets created from live chatbots in diverse domains. Unlike most existing datasets that are crowdsourced, our datasets contain real user queries received by the chatbots and facilitates penalising unwanted correlations grasped during the training process. We evaluate 4 NLU platforms and a BERT based classifier and find that performance saturates at inadequate levels on test sets because all systems latch on to unintended patterns in training data.

Parth Mehta, Gaurav Arora, Prasenjit Majumder

In this work, we present a weakly supervised sentence extraction technique for identifying important sentences in scientific papers that are worthy of inclusion in the abstract. We propose a new attention based deep learning architecture that jointly learns to identify important content, as well as the cue phrases that are indicative of summary worthy sentences. We propose a new context embedding technique for determining the focus of a given paper using topic models and use it jointly with an LSTM based sequence encoder to learn attention weights across the sentence words. We use a collection of articles publicly available through ACL anthology for our experiments. Our system achieves a performance that is better, in terms of several ROUGE metrics, as compared to several state of art extractive techniques. It also generates more coherent summaries and preserves the overall structure of the document.