Publications

You can also find my articles on my Google Scholar profile.

Enhancing Systematic Decompositional Natural Language Inference Using Informal Logic

Published in arXiv.org, 2024

Discovered that training an RDTE-oriented entailment classifier via knowledge distillation and employing it in a modern neuro-symbolic reasoning engine significantly improves results over other entailment classifier baselines, illustrating the practical benefit of this advance for textual inference.

Recommended citation: Weir, N., Sanders, K., Weller, O., Sharma, S., Jiang, D., Jiang, Z., Dalvi, B., Tafjord, O., Jansen, P.A., Clark, P., & Durme, B.V. (2024). Enhancing Systematic Decompositional Natural Language Inference Using Informal Logic. ArXiv, abs/2402.14798. https://arxiv.org/abs/2402.14798

AnaloBench: Benchmarking the Identification of Abstract and Long-context Analogies

Published in arXiv.org, 2024

This work proposes ANALOBENCH, a benchmark to determine analogical reasoning ability in LMs, and tests a broad collection of proprietary models and open source models, finding that scale offers minimal gains when, (i) analogies involve lengthy scenarios, or (ii) recalling relevant scenarios from a large pool of information.

Recommended citation: Ye, X., Wang, A., Choi, J., Lu, Y., Sharma, S., Shen, L., Tiyyala, V., Andrews, N., & Khashabi, D. (2024). AnaloBench: Benchmarking the Identification of Abstract and Long-context Analogies. ArXiv, abs/2402.12370. https://arxiv.org/abs/2402.12370

Comparative Analysis of Entity Identification and Classification of Indian Epic

Published in ICMI 22: Proceedings of the 2022 International Conference on Multimodal Interaction, 2022

This study addresses the lack of Natural Language Processing (NLP) research on Indian epics like the Mahabharata. By analyzing state-of-the-art supervised Machine Learning (ML) methods for Named Entity Recognition (NER) on a labeled dataset of Mahabharata tokens, the research aims to revitalize NLP exploration in this domain. Challenges include the presence of English and Sanskrit words with different characterizations of characters throughout the narrative. Findings reveal shortcomings in existing methods like NLTK, spaCy, and Stanford NER due to difficulties in distinguishing entities with the same name. Context-driven methods like BERT show promise but tend to overfit the dataset. Overall, the study underscores the need for further research to develop NER techniques suited to the complexities of Indian epics.

Recommended citation: Shreya Sharma and Mukesh Mohania. 2022. Comparative Analysis of Entity Identification and Classification of Indian Epics. In Proceedings of the 2022 International Conference on Multimodal Interaction (ICMI 22). Association for Computing Machinery, New York, NY, USA, 404–413. https://doi.org/10.1145/3536221.3556573 https://dl.acm.org/doi/abs/10.1145/3536221.3556573

Integrative Analysis and Machine Learning Based Characterization of Single Circulating Tumor Cells

Published in Journal of Clinical Medicine, 2020

This study compiled single-cell expression data of circulating tumor cells (CTCs) from various cancers, revealing a continuum of epithelial to mesenchymal transition. Analysis also identified an inverse gene expression pattern between PD-L1 and MHC, relevant to cancer immunotherapy. A classifier trained on CTC and peripheral blood mononuclear cell (PBMC) transcriptomes accurately recognized diverse CTC phenotypes. Additionally, this classifier validated circulating breast tumor cells captured using a novel label-free microfluidic system for CTC enrichment.

Recommended citation: Iyer A, Gupta K, Sharma S, Hari K, Lee YF, Ramalingam N, Yap YS, West J, Bhagat AA, Subramani BV, et al. Integrative Analysis and Machine Learning Based Characterization of Single Circulating Tumor Cells. Journal of Clinical Medicine. 2020; 9(4):1206. https://doi.org/10.3390/jcm9041206 https://www.mdpi.com/2077-0383/9/4/1206

Predictive Maintenance of Air Conditioning Systems Using Supervised Machine Learning

Published in 2019 20th International Conference on Intelligent System Application to Power Systems (ISAP), 2020

The importance of predictive maintenance for air conditioners due to various potential faults is highlighted in this paper. Gas leakage and capacitor malfunction, two common issues, are detected using the decision tree machine learning algorithm. Data collected from sensors and microcontrollers are analyzed using MATLAB Classification App Learner Toolbox. The decision tree method shows higher prediction accuracy compared to the support vector machine. This research aims to identify faults early and enable proactive maintenance to mitigate efficiency loss, energy consumption increase, and maintenance costs.

Recommended citation: S. Trivedi, S. Bhola, A. Talegaonkar, P. Gaur and S. Sharma, "Predictive Maintenance of Air Conditioning Systems Using Supervised Machine Learning," 2019 20th International Conference on Intelligent System Application to Power Systems (ISAP), New Delhi, India, 2019, pp. 1-6, doi: 10.1109/ISAP48318.2019.9065995. https://ieeexplore.ieee.org/abstract/document/9065995