CorpusMate 2.0: Integrating AI-powered language analysis into a data-driven learning corpus tool

Abstract

This article presents CorpusMate 2.0 (https://corpusmate.app/), a substantially rebuilt version of the CorpusMate corpus tool for data-driven learning (DDL) first described in Crosthwaite and Baisa (2024). The new version retains the core concordancing, n-gram and disciplinary variation functions of its predecessor while introducing a range of significant enhancements including a rebuilt Python/FastAPI backend, five collocation association measures (including MI2, MI3, and log-likelihood), and - most substantially - a fully integrated large language model (LLM) assistant that reads users' actual search results to provide evidence-grounded linguistic commentary. As a novel feature specific for DDL, the AI assistant also operates across four user roles (Researcher, Teacher, Student, Language learner) each with a distinct pedagogical persona, while the platform also supports responses in eleven languages. This article describes the motivation for these changes, the new platform architecture, and enhanced functionality, while inviting continued use of the tool for DDL research and pedagogy.