Abstract
Investigation records are distributed across varied sources, and their facts may be incomplete, duplicated, or inconsistent, making evidence tracking, legal reference checking, and charge-sheet preparation difficult. This paper introduces Byomkesh AI, an investigation intelligence system that organizes data derived from documents into a reviewable, evidence-linked case representation. The system has nine stages: document ingestion, OCR, document parsing, investigation graph construction, canonical projection, evidence validation, case context construction, legal-reference reasoning, and charge-sheet population. During validation, there are three possible dispositions: draft allowed, review required, and final blocked. The mean extraction confidence score for the project is reported to be 0.99 (99%), but no information has been provided regarding the process used for aggregation and calibration of this score. It is not appropriate to interpret this extraction confidence score as the score for extraction accuracy. There is no benchmark for mapping accuracy, evaluation dataset, denominator or ground truth for BNS-mapping. The Section 106 example given in the project presentation is merely a demonstration, and not an evaluation of accuracy.References
- Mathew, Minesh, Dimosthenis Karatzas, and C. V. Jawahar. ”Docvqa: A dataset for vqa on document images.” In Winter Conference on Applications of Computer Vision (WACV),IEEE, 2021,pp. 2199-2208.
- Huang, Yupan, Tengchao Lv, Lei Cui, Yutong Lu, and Furu Wei. ”Layoutlmv3: Pre-training for document ai with unified text and image masking.” In Proceedings of the 30th ACM international conference on multimedia,20222, pp. 4083-4091.
- Appalaraju, Srikar, Bhavan Jasani, Bhargava Urala Kota, Yusheng Xie, and R. Manmatha. ”Docformer: End-to-end transformer for document understanding.” In 2021 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE, 2021, pp. 973-983.
- Neudecker, Clemens, Konstantin Baierer, Mike Gerber, Christian Clausner, Apostolos Antonacopoulos, and Stefan Pletschacher. ”A survey of OCR evaluation tools and metrics.” In Proceedings of the 6th International Workshop on Historical Document Imaging and Processing, 2021, pp. 13-18.
- Chalkidis, Ilias, Abhik Jana, Dirk Hartung, Michael Bommarito, Ion Androutsopoulos, Daniel Katz, and Nikolaos Aletras. ”LexGLUE: A benchmark dataset for legal language understanding in English.” In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2022, pp. 4310-4330.
- Joshi, Abhinav, Shounak Paul, Akshat Sharma, Pawan Goyal, Saptarshi Ghosh, and Ashutosh Modi. ”IL-TUR: Benchmark for Indian legal text understanding and reasoning.” In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, 2024, vol.10, pp.11460-11499.
- Paul, Shounak, Pawan Goyal, and Saptarshi Ghosh. ”Lesicin: A heterogeneous graph-based approach for automatic legal statute identification from indian legal documents.” In Proceedings of the AAAI conference on artificial intelligence, 2024, vol. 36, no. 10, pp. 11139-11146.
- Shukla, Siddharth, Tanuj Tyagi, Abhay Singh Bisht, Ashish Sharma, and Basant Agarwal. ”NyayGraph: A knowledge graph enhanced approach for legal statute identification in Indian law using large language models.” In Proceedings of the Natural Legal Language Processing Workshop, 2025, pp. 147-156.
- Lewis, Patrick, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Kuttler et al. ”Retrieval-augmented generation for knowledge- ¨ intensive nlp tasks.” Advances in neural information processing systems 2024, vol.33 pp. 9459-9474.
- Gao, Tianyu, Howard Yen, Jiatong Yu, and Danqi Chen. ”Enabling large language models to generate text with citations.” In Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2023, pp. 6465-6488.
- Nushi, Besmira, Shamsi T. Iqbal, Paul Bennett, Jaime Teevan, Eric Horvitz, Dan Weld, Ruth Kikin-Gil et al. ”Guidelines for human-AI interaction.” In Proceedings of the CHI Conference on Human Factors in Computing Systems. 2019, no.3 pp. 1-13.
- Klie, Jan-Christoph, Richard Eckart De Castilho, and Iryna Gurevych. ”From zero to hero: Human-in-the-loop entity linking in low resource domains.” In Proceedings of the 58th annual meeting of the association for computational linguistics, 2020, pp. 6982-699.
- Government of India. The Bharatiya Nyaya Sanhita, 2023. Act No. 45 of 2023, enacted December 25, 2023.
- Government of India. The Bharatiya Nagarik Suraksha Sanhita, 2023. Act No. 46 of 2023, enacted December 25, 2023.
- Government of India. The Bharatiya Sakshya Adhiniyam, 2023. Act No. 47 of 2023, enacted December 25, 2023.

Journal of Information Technology and Digital World