Using Co-occurrence Analysis to Expand Consumer Health Vocabularies from Social Media Data

Abstract

As health consumerism in the United States has remarkably risen over the past decade, more and more health consumers are actively seeking health related information on their own. However, health consumers' efforts in seeking healthcare information could be challenging due to the lack of professional knowledge in medicine, and the gap between health professional vocabulary and health consumer vocabulary is one of the biggest issues. It has long been recognized that consumers and health professionals often express health concepts in different ways and consumers usually find it difficult to understand some medical terminologies or express themselves using the professional language for information seeking. To address this problem, many researchers have been working on developing Consumer Health Vocabulary (CHV) to bridge the language gap. One crucial step in developing CHV is the identification of consumer health expressions. Most of the methods used by existing studies in identifying consumer health expressions involve human efforts, which is very time-consuming. In this study, we propose an automatic method to identify consumer health expressions from consumer-contributed content, which represent consumers' language the best. Co-occurrence analysis was used to identify terms that co-occur frequently with a set of seed terms. A corpus containing 120,393 discussion messages was used as a dataset and co-occurrence analysis was used to extract the most related consumer expressions. The experiment results demonstrated that the proposed method achieved a good performance in identifying consumer health expressions.

Symptom clusters in women with breast cancer: an analysis of data from social media and a research study

“…Similarly, variations of the standard symptom such as “changes mood” or “change moods” are counted as well. This reduces the possibility of missing symptoms that are not described using a standard nomenclature [51, 52]. …”

Enriching consumer health vocabulary through mining a social Q&A site: A similarity-based approach

“…To simultaneously extract consumer and professional term pairs, Vydiswaren et al [19] leveraged pre-defined lexical patterns such as “A, also known as B” to automatically extract the synonymous pairs of terms A and B from Wikipedia, and labeled them as either consumer or professional terms. In another study, over 120,000 discussion messages on MedHelp were examined to identify consumer expressions that co-occurred with pre-selected terms associated with adverse drug reactions [16]. These studies often require labor-intensive manual reviews by domain experts or the crowd, and/or rely on ad hoc lexical syntactic patterns, limiting their scalability.…”

“…The development of professional vocabularies (e.g., SNOMED CT) receives suffiicient support from teams of domain experts as well as professional organizations (e.g., SNOMED International and NLM). In contrast, the CHV employs an open-access and collaborative approach [13] to identify or extract lay people’ health-related terms from a variety of consumer-generated text corpora on various platforms such as PatientsLikeMe [14], MedHelp [15, 16], MedLinePlus [17, 18], and Wikipedia [19]. …”

Mining Heterogeneous Healthcare Networks Extracted from Health-Consumer-Contributed Contents for Adverse Drug Reaction Detection

“…Therefore, high quality CHV is able to help with capturing more consumers' expressions and better extracting ADR terms. Some studies are dedicated to expanding CHV by using social media data (Jiang and Yang 2013;Jiang and Yang 2015). For D, we search for diseases that are treated by each of the 20 drugs in SIDER database 5 to construct our disease lexicon.…”