Computing Latent Taxonomies from Patients' Spontaneous Self-Disclosure to Form Compatible Support Groups

A growing number of Internet sites provide patients with an opportunity to share experiences about their illness. Unfortunately, it depends on the luck of the draw whether patients find compatible others for mutual support. To improve this situation, this paper demonstrates how tools from Information Retrieval can be deployed to discover patients with compatible stories more systematically. We show how finding compatible support is like finding relevant documents on the Internet, and also in what way it is different. We designed an algorithm to produce a 'latent' taxonomy over patients' self-disclosures to aid them in forming support groups. At the end we briefly consider the result of this taxonomy based on only the patients' stories, with one based on explicit questionnaires.