As I consider my time as an ISG GovernX summer 2023 research intern, although it was based in Stamford, Connecticut, USA, I worked remotely from Virginia. I am swept away into a domain where contract management, innovation, and the application of cutting-edge technologies intersect with the application of coding expertise. Engaging in the development of a Clause Recommendation System during my internship not only enhanced my technical skills but also expanded my understanding of reality-based implementations about natural language processing and data analysis.
My primary objectives were: to assist the team with my coding expertise and convert client requirements into coding tasks. By testing various methodologies and applying my solutions to resolve tangible challenges, I developed professionally as a programmer and demonstrated the efficacy of my contributions to the group.
One key project I worked on was the Clause Recommendation System. I played a role in building its essential parts: Contract Indexing, Clause Similarity Analysis, Clause Correction, and Clause Authoring. These elements aimed to simplify contract management and drafting processes using NLP, indexing, and various algorithms.
As a part of these projects, using state-of-the-art indexing methodologies such as Faiss (Facebook AI Similarity Search) which is developed by FAIR (Fundamental AI Research) at Facebook, and Annoy (Approximate Nearest Neighbors) which is also used by Spotify for music recommendations, I led the complete indexing of contract clauses and entire contracts. By making use of these complex indexing methods, I developed a methodical strategy to effectively arrange and access contract clauses. The implementation process encompassed several steps: converting textual data into indexable numerical representations, generating high-dimensional vector representations using the Annoy and Faiss libraries, and optimizing search functionalities to ensure efficient retrieval of relevant clauses. By strategically employing indexing methods, the system's performance was greatly improved. This resulted in the ability to retrieve clauses quickly and accurately, which in turn facilitated users' seamless navigation through contracts for the purposes of analysis and management.
I led the development of the Clause Authoring component within the project scope, which sought to generate a curated list of upcoming clauses based on previous clauses inside unfinished contracts. I designed a unique strategy to forecast and generate contextually appropriate clauses by combining a model based on transformer architecture with other strategic methods. This included training the transformer-based model on large datasets, allowing it to understand the complexity and connections within clauses. I created a system capable of intelligently recommending suitable follow-up clauses by using the power of transformers and complementary techniques, allowing for a more efficient and contextually aligned contract drafting process.
As part of this project, I was involved in the implementation of the Clause Similarity Analysis, a critical component meant for analyzing and calculating the similarity between various provisions inside contracts. I created a systematic framework to identify the similarity levels of clauses using complex algorithms and Natural Language Processing (NLP) approaches. I supported the comparison of textual material by leveraging cosine similarity and other similarity measures, allowing the system to discover and categorize phrases based on their semantic closeness. This method not only sped up contract analysis, but it also improved the Clause Recommendation System's capacity to extract key clauses quickly, resulting in a more efficient and accurate contract management process.
During my time on the project, I made significant improvements to the creation of the Clause Correction function, which is intended to discover and correct problems in contract provisions. I began developing a system capable of identifying inconsistencies, grammatical faults, and potential legal ambiguities inside contract terms using a combination of rule-based methods and preliminary examination of machine learning techniques. I began training models on annotated datasets and began developing rule-based validation procedures to establish the groundwork for automatic error detection. While the function was still in its infancy, their efforts provided the framework for future additions and the ability to recommend adjustments and enhancements to maintain the correctness and integrity of contract terms. The Clause Correction feature might have been able to make contract papers safer by reducing legal risks and making contract management stronger, but it wasn't fully developed while I was an intern.
In addition to technical expertise, I gained the ability to articulate complex ideas with clarity and recognized the value of collaborative efforts. The internship provided an opportunity for personal and professional development.
Throughout the internship, I worked in an iterative development process, refining and improving the Clause Recommendation System. I participated in rigorous testing and fine-tuning of the system's features while working closely with the team. We iteratively increased the accuracy of clause recommendations, fine-tuned indexing approaches, and improved the overall efficiency of the system using continual feedback loops and agile methodologies. This iterative process made sure that our system stayed flexible, responsive, and able to give clients consistent value in the complex field of contract analysis and management.
The ISG GovernX Internship was a life-changing learning experience that not only expanded my technical knowledge but also gave me an improved understanding of technology's transformational possibilities in contract management. Looking ahead, I am excited to continue contributing to the growth of contract management systems, as well as to explore new paths for innovation and to leverage emerging technology to further modernize the area. The internship laid the groundwork for future projects that will continue to push the boundaries of what is feasible in contract management and analysis.
It was a great opportunity to work alongside such bright and dedicated people as Kashyap Puranik, Rekha Acharya, Aravind S, Varun Gopal, Praveen Rajagopal, Sourabh Dhaker, and Anchal Panda. I want to thank Aravind S and Varun Gopal for providing me with this fantastic chance to be a part of the team. Special thanks to Kashyap Puranik and Rekha Acharya for believing in me and providing me with this rewarding internship. In addition, I'd like to thank Stanley Paul and Cady Griggs for their great HR help throughout the internship. Furthermore, I want to express my heartfelt gratitude to Dr. Vikas Ashok, my Ph.D. adviser. His continuous support, advice, and encouragement have been crucial throughout my academic path and have made a significant contribution to my professional development.
– Mohan Krishna Sunkara (@mk344567)