Research Companion App

Onyebuchukwu, Oke (2024). Research Companion App. Knowledge Media Institute, The Open University, Milton Keynes, UK.

URL: https://kmi.open.ac.uk/scholarship/

Abstract

This paper introduces the Research Companion App, designed to assist students in comprehending research papers through interactive features. Utilising Retrieval Augmented Generation (RAG) and prompt engineering, the app enhances response accuracy, scalability, and flexibility. The app, built with Streamlit and Hugging Face, and leveraging Microsoft Azure OpenAI models, includes functionalities such as summarisation, visual aids generation, quizzes, and a chatbot. Evaluation of the chatbot's performance across different temperature settings reveals that higher temperatures generally yield more consistent and reliable outputs. However, the quality of responses is contingent on the retrieval process. Future improvements aim to refine summary generation and explore better QA evaluators.

Plain Language Summary

The Research Companion App helps students understand research papers with interactive tools. It uses advanced technology to improve accuracy and flexibility. Built with Streamlit and Hugging Face, and using Microsoft Azure OpenAI models, the app can summarize papers, create visual aids, generate quizzes, and includes a chatbot. Tests show that the chatbot works better at higher settings, but its quality depends on how well information is retrieved. Future updates will focus on better summaries and question-answer evaluations.

Viewing alternatives

Item Actions

Export

You can export this page using these formats