Papers by Vasu Jindal
Bridging the Embodiment Gap in Agricultural Knowledge Representation for Language Models (2025.acl-srw)
Copied to clipboard
| Challenge: | a paper quantifies the “embodiment gap” between disembodied language models and embodied agricultural knowledge communication . agronomists and researchers examined the embodiment gap in 78 farmers . |
| Approach: | They propose a framework that integrates linguistic patterns from five domains of agricultural expertise and a new metric for evaluating embodied knowledge representation in language models. |
| Outcome: | The proposed frameworks reduce the embodiment gap by 47.3% across agricultural domains . the proposed framework improves tool usage discourse and soil assessment terminology . |
Generating Image Captions in Arabic using Root-Word Based Recurrent Neural Networks and Deep Neural Networks (N18-4)
Copied to clipboard
| Challenge: | Existing studies on image caption generation in English focus on Western languages, ignoring Semitic and Middle-Eastern languages like Arabic, Hebrew, Urdu and Persian. |
| Approach: | They propose to leverage the critical dependency of Arabic to generate Arabic captions using root-word based Recurrent Neural Network and Deep Neural networks. |
| Outcome: | The proposed model outperforms English-Arabic translated captions on a dataset from newspapers in the Middle East. |