Bangla has a rich written literature, automatically making it replete with examples of creative usage of language. There have been limited efforts to computationally analyze creative text in the Bangla language due to a lack of resources. We present Alankaar, a dataset of 2500 manually annotated examples of text fragments in Bangla containing metaphors. We also provide automatic and manual English translations of these examples. Additionally, we provide 2500 examples of non-metaphorical text in Bangla. We use this dataset to build a metaphor identification system in Bangla. We also use it as a test bed for cross-lingual metaphor translation, finding that not all metaphors translate literally across languages and there are several cultural factors at play in the translation of metaphors. We hope this will advance the field in metaphor translation research and in grounding cultural nuances at work in the process of machine translation.

We introduce ASQ, a tool to automatically mine questions and answers from a sentence using the Abstract Meaning Representation (AMR). Previous work has used question-answer pairs to specify the predicate-argument structure of a sentence using natural language, which does not require linguistic expertise or training, and created datasets such as QA-SRL and QAMR, for which the question-answer pair annotations were crowdsourced. Our goal is to build a tool (ASQ) that maps from the traditional meaning representation AMR to a question-answer meaning representation (QMR). This enables construction of QMR datasets automatically in various domains using existing high-quality AMR parsers, and provides an automatic mapping AMR to QMR for ease of understanding by non-experts. A qualitative evaluation of the output generated by ASQ from the AMR 2.0 data shows that the question-answer pairs are natural and valid, and demonstrate good coverage of the content. We run ASQ on the sentences from the QAMR dataset, to observe that the semantic roles in QAMR are also captured by ASQ. We intend to make this tool and the results publicly available for others to use and build upon.

Open-domain question answering (ODQA) often requires models to resolve conflicting evidence retrieved from diverse sources—a task that remains challenging even for state-of-the-art large language models (LLMs). While single-agent techniques such as self-verification and self-consistency have shown promise across natural language understanding and generation tasks, and multi-agent approaches involving collaborative or competitive strategies have recently emerged, their effectiveness for ODQA in the presence of conflicting contexts remains underexplored. In this work, we investigate these techniques using the QACC dataset as a case study. We find that incorporating a multi-agent verification step—where the best answer is selected from among outputs generated by different LLMs—leads to improved performance. Interestingly, we also observe that requiring explanations during the verification step does not always improve answer quality. Our experiments evaluate three strong LLMs (GPT-4o, Claude 4, and DeepSeek-R1) across a range of prompting and verification baselines.

Figurative language is commonplace in natural language, and while making communication memorable and creative, can be difficult to understand. In this work, we investigate the robustness of Question Answering (QA) models on figurative text. Yes/no questions, in particular, are a useful probe of figurative language understanding capabilities of large language models. We propose FigurativeQA, a set of 1000 yes/no questions with figurative and non-figurative contexts, extracted from the domains of restaurant and product reviews. We show that state-of-the-art BERT-based QA models exhibit an average performance drop of up to 15% points when answering questions from figurative contexts, as compared to non-figurative ones. While models like GPT-3 and ChatGPT are better at handling figurative texts, we show that further performance gains can be achieved by automatically simplifying the figurative contexts into their non-figurative (literal) counterparts. We find that the best overall model is ChatGPT with chain-of-thought prompting to generate non-figurative contexts. Our work provides a promising direction for building more robust QA models with figurative language understanding capabilities.

Figurative language is widespread in human language (Lakoff and Johnson, 2008) posing potential challenges in NLP applications. In this paper, we investigate the effect of figurative language on the task of question answering (QA). We construct FigQA, a test set of 400 yes-no questions with figurative and non-figurative contexts, extracted from product reviews and restaurant reviews. We demonstrate that a state-of-the-art RoBERTa QA model has considerably lower performance in question answering when the contexts are figurative rather than literal, indicating a gap in current models. We propose a general method for improving the performance of QA models by converting the figurative contexts into non-figurative by prompting GPT-3, and demonstrate its effectiveness. Our results indicate a need for building QA models infused with figurative language understanding capabilities.

Prepositional Phrase (PP) attachment is a classical problem in NLP for languages like English, which suffer from structural ambiguity. In this work, we solve this problem with the help of another language free from such ambiguities, using the parse tree of the parallel sentence in the other language, and word alignments. We formulate an optimization framework that encourages agreement between the parse trees for two languages, and solve it using a novel Dual Decomposition (DD) based algorithm. Experiments on the English-Hindi language pair show promising improvements over the baseline.