AskSport: Web Application for Sports Question-Answering

TL;DR

AskSport addresses the need for domain-specific sports QA by combining a BM25-based retriever with a RoBERTa reader over the QASports dataset within a BigQA-inspired architecture. The system retrieves up to ten documents and returns the top three answers with confidence and provenance, presented through a Streamlit interface and hosted on HuggingFace. The paper demonstrates qualitative scenarios (e.g., Rookie of the Year, team title counts, historical best players) and discusses accessibility and potential future extensions to other sports and different QA components. The contribution lies in a modular, accessible sports QA pipeline that can be extended with alternative retrievers/readers. The practical impact is providing fans, coaches, and media professionals with quick, structured answers tied to source documents.

Abstract

This paper introduces AskSport, a question-answering web application about sports. It allows users to ask questions using natural language and retrieve the three most relevant answers, including related information and documents. The paper describes the characteristics and functionalities of the application, including use cases demonstrating its ability to return names and numerical values. AskSport and its implementation are available for public access on HuggingFace.

Figures (3)

• Figure 1: Implementation Pipeline of AskSport.
• Figure 2: AskSport Interface. Demonstrating key capabilities.
• Figure 3: Details of Tabs 2 and 4 of AskSport.