Fifteenth International Conference on Technology, Knowledge & Society

Abstract

Jargon, assumptions, web pages that look like data dumps, link overload. These are all barriers to connecting users to the information they seek from sources like libraries, archives, and open access repositories. Search term analytics and UI studies help us to better understand what users are looking for and what they click on. But what if we could collect, analyze, and respond to their use of natural language during information seeking and deep research dives? By using chatbots rather than search boxes and then closely analyzing the interactions between searcher and bot we can better understand how users think about and seek out information in systems designed to serve knowledge. Analysis of chatbot transcripts related to information seeking on library websites may provide practical insights into how information can be better organized and labeled for discovery in libraries, archives and digital repositories. Chatbot transcripts from information serving web portals, particularly library websites, are collected, mined and textually analyzed for recurrent terminology, natural language patterns and connective ideas/concepts. Preliminary results show there are patterns in natural language and connective concepts during information seeking, as predicted. Conclusions: This research is ongoing as we seek to enlarge datasets through collection of additional transcripts. Preliminary analysis indicates that examination of natural language expressions and inquiry paths made through chatbot conversations will assist in constructing more intuitive interfaces and taxonomies for information systems, leading to more efficient and effective methods of information sharing, knowledge building and research dissemination.