Intentions and attention in exploratory health search

Abstract

We study information goals and patterns of attention in exploratory search for health information on the Web, reporting results of a large-scale log-based study. We examine search activity associated with the goal of diagnosing illness from symptoms versus more general information-seeking about health and illness. We decompose exploratory health search into evidence-based and hypothesis-directed information seeking. Evidence-based search centers on the pursuit of details and relevance of signs and symptoms. Hypothesis-directed search includes the pursuit of content on one or more illnesses, including risk factors, treatments, and therapies for illnesses, and on the discrimination among different diseases under the uncertainty that exists in advance of a confirmed diagnosis. These different goals of exploratory health search are not independent, and transitions can occur between them within or across search sessions. We construct a classifier that identifies medically-related search sessions in log data. Given a set of search sessions flagged as health-related, we show how we can identify different intentions persisting as foci of attention within those sessions. Finally, we discuss how insights about foci dynamics can help us better understand exploratory health search behavior and better support health search on the Web.

Characterizing Health-Related Community Question Answering

“…The charts demonstrate an acceptable (given the data volume) association between the extracted topics and real-life events. Question and answer parts of the CQA pages allow us to detect different question types analogous to evidence-directed and hypothesis-directed queries in the Web health search [1]. For example, hypothesis-directed search intent can be associated with a template "disease in question -therapy in answers".…”

“…The site was launched in 2006 and has accumulated almost 80 million questions and more than 400 million answers by August 2012. 1 The most remarkable difference from Yahoo! Answers is the two-level directory used at Otvety@Mail.Ru.…”

Interactions between health searchers and search engines

“…To answer them, we studied the search logs of 190 thousand consenting users of a major commercial Web search engine. Search logs are a valuable resource for studying information seeking in a naturalistic setting, and such data has been used by several studies to explore how searchers obtain medical information [8,32]. The pipeline used to process this data and the features extracted for analysis are further described in Section 3.…”

“…A user experiencing anxiety about a headache might first spend some time searching for information on a serious condition such as "brain tumor" and then switch to a symptom query about headache (a type of transition that prior work shows occurs frequently [8]). The user might choose the query wording "severe headache explanations" because of the subjective concern they experience at query time.…”

Towards Automatic Evaluation of Health-Related CQA Data

“…In contrast to these studies we experiment with an automatic method for quality assessment focused on health-related CQA. The idea of the approach is to perform evaluation on a narrowed subset -questions asking for medication for a specific symptom or disease, for example (all examples are originally in Russian): According to different studies [1,9,7,21] this type of information need is among the most common ones. We hypothesize that this type of questions is exemplary enough to reflect the overall quality of CQA health-related content.…”