Text mining for health knowledge discovery from social media

Suzan Verberne

The Leiden Inst of Advanced Computer Science, Leiden University


Patient forums are forums centered around patient communities. Previous qualitative work has shown that patients gather on patient forums to exchange information and experiences, and support each other emotionally. Patient forums can also be a source for medical hypotheses, e.g. on the effectiveness of medication and side effects. This specifically benefits patients with a rare disease for which clinical trials are often too costly. In the Ph.D. project of Anne Dirkson, we have developed text mining techniques to process and extract information from the large volume of messages on a patient forum. Specifically, we have focussed on extracting the side effects of medications, and the coping strategies of patients who suffer from these side effects. The extraction and aggregation of this information are more challenging than extracting regular named entities (like names and locations) because side effects and coping strategies are not proper nouns; they can be described descriptively with a large variation. For example, one could describe their headache with 'my head is bursting', 'throbbing pain in my head', or 'pounding headache' to name a few. In my presentation, I will explain the challenges of knowledge discovery from patient forum data, the methods that we developed, and the results that we obtained. I will also show how the extracted information relates to results from questionnaire data among patients.

Short bio:

Suzan Verberne is an associate professor at the Leiden Institute of Advanced Computer Science at Leiden University. She is the group leader of Text Mining and Retrieval. She obtained her Ph.D. in 2010 on the topic of Question Answering and has since then been working on the edge between Natural Language Processing (NLP) and Information Retrieval (IR). She has supervised projects involving a large number of application domains: from social media to law and from archaeology to health. Her research focus is to advance NLP "beyond the benchmark", addressing challenging problems in specific domains. She is highly active in the NLP and IR communities, holding chairing positions in large worldwide conferences. See link to bio profile

Week 20 2022/2023

Thursday 23rd March 2023

Microsoft Teams - request a link via email

For queries, contact Chloe: c.humphreys@lancaster.ac.uk or Ignatius: i.ezeani@lancaster.ac.uk