Data Mining Virtual Reference Chats

Abstract

From surveying the literature, most chat transcript analysis for academic libraries has been based on sampling. This project departs from that approach. Three years of raw data (15,441 chat transcripts) from LibraryH3lp was examined to better understand user experience. Open and proprietary software was used to text mine, process, and analyze the transcripts to identify and visualize patterns. Using this “big data” provides a window into unique community needs and has measurable applications to web technology, reference, and assessment.