LLM in Public Service Applications: CPGRAMS

Abstract

CPGRAM Portal is a government platform introduced for the citizens to file their grievances related to specific ministries. With the increasing population, it’s difficult to handle and manage a large number of complaints. Also manual addressing to basic user queries is also time taking and requires a lot of labour. This project proposes an AI/ML driven chatbot designed to handle users will ease the grievances filing process. This chatbot is powered by Large Language Models(LLMs) which can address user queries efficiently. This chatbot will be made for specific ministries, guiding users step by step in filing grievances and also address frequently asked questions. Use of machine learning enables the chatbot to continuously learn and adapt to user queries which will help in increasing the response time and accuracy. So this will help to reduce the manual support, enhance user satisfaction and optimize the overall experience of the user. This solution focuses on addressing common queries faced by user, timely and efficient support for grievance resolution in ministry-specific context.

Introduction

Purpose:
CPGRAMS allows citizens to file grievances related to government ministries, but rising complaint volumes have made timely resolution difficult. Integrating AI/ML, especially large language models (LLMs), can streamline grievance handling, improve response times, and enhance citizen satisfaction by providing real-time, ministry-specific support via a chatbot.

Key Points:

• AI/ML Benefits:

  • Continuously learns and improves accuracy

  • Identifies query patterns to optimize efficiency

  • Reduces manual workload, speeds up responses

  • Cloud-hosted for scalability and reliability

• Chatbot Features:

  • Real-time assistance for grievance submission

  • Answers frequently asked questions

  • Tracks grievance status

  • Improves accessibility and responsiveness in citizen-government interaction

Literature Survey Highlights:

• AI can transform public services by enhancing decision-making, data analysis, and reducing bureaucracy but raises ethical, privacy, and trust issues.

• LLMs show promise in government and legal domains but need safeguards against misinformation and bias.

• Cognitive computing and personalized assistants improve complaint management and user interaction.

• Retrieval-Augmented Generation (RAG) enhances accuracy by integrating external knowledge with LLMs.

Technical Approach:

• Algorithms Used:

  • SentenceTransformer (all-MiniLM-L6-v2): Converts text into semantic embeddings for meaningful search

  • FAISS (Facebook AI Similarity Search): Fast nearest-neighbor retrieval of relevant information

  • Mistral-7B Transformer: Generates coherent, context-aware natural language responses

  • MongoDB: Manages user data, credentials, and grievance records

• Methodology:

  • Collect and preprocess grievance-related documents (FAQs, guides) into searchable embeddings

  • Build a FAISS index for fast retrieval of relevant info

  • Use RAG framework to combine retrieved data with user queries for accurate answers

  • Deploy chatbot on cloud infrastructure (Vertex AI, Compute Engine) for scalability

  • Provide RESTful API endpoints for signup, login, query, and status checking

  • Operational testing confirms effective retrieval, response generation, and database integration

Technology Stack:

• Python (3.11), Flask for backend API development

• PyPDF2 for PDF text extraction

• Hugging Face’s Mistral-7B for language modeling

• Torch for machine learning inference

• Cloud platforms for deployment and scalability

Impact:

The AI-powered CPGRAMS chatbot offers an efficient, scalable, and user-friendly system to handle public grievances. By automating query handling and improving accuracy through continuous learning, it aims to transform citizen-government engagement into a more effective, transparent, and responsive process.

Conclusion

This project showcases the creation and implementation of an AI enabled chatbot that improves the functionality and user experience of CPGRAMS grievance filing process. Using recent advancements, such as Large Language Models (LLMs), FAISS-based vector retrieval, and cloud computing, the chatbot effectively helped users navigate the grievance process and answer frequently asked questions. Through the use of natural language processing for training and semantics, combined with a cloud-based, scalable backend platform for generating responses and other functionality, the project demonstrated the chatbot\'s ability to adapt and respond quickly in a public service situation. By providing a remote and scalable intelligent user interface to users while minimizing the need to interact with manual support, the proposed solution has dramatically improved users\' access to government services. Moreover, the chatbot has the capability of continuing to improve over time as users interact it and provide feedback, ensuring its relevance in a dynamic and complex operational environment. Overall, the project demonstrates how the use of AI and ML technologies can be used to change the engagement of citizens and modernize public sector engagement to make it more efficient and agile, and is an important example for future projects leveraging AI-based governance solutions.

References

[1] The Role of Artificial Intelligence in Government Services: A SystematicLiterature ReviewHamirul1*, Darmawanto1, Nova Elsyra1, Syahwami1 1Setih Setio Institute of Administration and Health, Muarabungo,Indonesia [2] On Large Language Models in National Security Applications William N. Caballeroa, Phillip R. Jenkinsa aDepartment of Operational Sciences, Air Force Institute of Technology,WPAFB, OH 4543 [3] Paul Henman (2020): Improving public services using artificial intelligence: possibilities, pitfalls, governance, Asia Pacific Journal of Public Administration, DOI: 10.1080/23276665.2020.1816188 To linkto thisarticle:https://doi.org/10.1080/23276665.2020.1816188 [4] Data Management and AI in E-government? Tibor Vámos and István Soós Computer and Automation Research Institute, Hungarian Academy of Sciences H-1111 Budapest, Lágymányosi u. 11., Hungary vamos@sztaki.hu [5] APPLYING COGNITIVE COMPUTING TOLEGAL SERVICES Deborah Whittle Faculty of Technology University of Sunderland, Sunderland. bg19tz@student.sunderland.ac.uk Lynne Hall Faculty of Technology University of Sunderland, SunderlandLynne.hall@sunderland.ac.ukhttp://dx.doi.org/10.14236/ewic/HCI2022.35 [6] A Personalized Assistant for Customer Complaints Management Systems Boris GalitskySchool of Computer Science and Information Systems Birkbeck College, University of London Malet Street, London WC1E 7HX, UK galitsky@dcs.bbk.ac.ukhttp://www.dcs.bbk.ac.uk/~galitsky/ComplaintE ngineIntro.html [7] A Large Language Model based legal assistant for governance applicationsMamalis, Marios Evangelos? marios.mamalis@uom.edu.gr Fitsilis, Fotios† fitsilisf@parliament.gr Kalampokis, Evangelos? ekal@uom.edu.gr Theodorakopoulos, Georgios‡ g.theodorakopoulos@nsk.gr Tarabanis,Konstantinos?kat@uom.edu.gr [8] Mistral 7BAlbert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, ThibautLavril, Thomas Wang, Timothée Lacroix,William El Sayed [9] ASurvey on RAGMeetingLLMs:Towards Retrieval-Augmented Large Language Models Wenqi Fan wenqifan03@gmail.com The Hong Kong Polytechnic University, HK SAR Shijie Wang arXiv:2405.06211v3 [cs.CL] 17 Jun 2024 shijie.wang@connect.polyu.hk The Hong Kong Polytechnic University, HK SAR Tat-Seng Chua Yujuan Ding? dingyujuan385@gmail.com The Hong Kong Polytechnic University, HK SAR Hengyun Li neilhengyun.li@polyu.edu.hk The Hong Kong Polytechnic University, HK SAR dcscts@nus.edu.sg National University of Singapore, Singapore Liangbo Ning BigLemon1123@gmail.com The Hong Kong Polytechnic University, HK SAR Dawei Yin yindawei@acm.org Baidu Inc, China Qing Li csqli@comp.polyu.edu.hk The Hong Kong Polytechnic University, HK SAR

Copyright

Copyright © 2025 Sudarshan Goyal, Rimjhim Singh. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.