Conversational QA Agents with Session Management

In the age of advanced artificial intelligence, this thesis attempts to extend conversational question-answering (QA) agents' power by integrating them with session management and document retrieval functionality. The first goal is to give conversational agents the ability to get and extract information from a company's document repository which enables it with greater power to provide answers based on the organizational information. The research starts by doing a complete investigation of current conversational QA agents' situation and outlines their weaknesses with their ability to understand the context and information retrieval and exploiting and better utilization of the organization’s internal knowledge repository and the need for more efficient session management. This paper proposes a novel conversational question-answering (QA) agent architecture with advanced session management capabilities. The agent architecture provides access to internal documents that can be searched using Elasticsearch to find relevant information to respond to user inquiries. An experimental evaluation was conducted to assess the system's performance and usefulness. The proposed conversational QA agent and its session management features were tested by analyzing how well the system could respond to users' questions by searching enterprise documents and maintaining contextual dialog responses. Findings show a considerable improvement in the conversational agent’s responses to user requests when using a multi-layer architecture with prompt tuning. In addition, session management integration ensures the proper flow of a dynamic conversation. The applied impact of this research can be determined in an array of fields where gaining reliable access to internal knowledge bases is critical for decision-making and problem-solving. The conclusion emphasizes what this thesis brought to the field of Conversational QA Agents, and underlines how it could transform organizational practices in managing such information. However, as conversational agents gain their importance as tools for knowledge experts, using novel approaches to tackle current problems would be a significant step towards context-aware and intelligent interaction between users and the agent.