Pharma’s forays into the brave new world of social media have not always been successful (tmm.txp.to/0214/brave) – but mining social media for information has attracted a lot of attention. But how exactly do you extract meaningful data from online chatter?
“Most systems for extracting adverse drug reactions (ADRs) follow a dictionary-based approach. The main drawback of these systems is that they fail to recognize terms which are not included in the dictionary,” wrote Isabel Segura-Bedmar, Ricardo Revert and Paloma Martinezin in a recent paper (1). “In addition, the dictionary-based approach is not able to handle the large number of spelling and grammar errors in social media texts.”
To help overcome these issues, the Spanish researchers have developed a system that mines social media and specialized blogs to detect potential ADRs. They have developed a prototype system (2), which uses the framework of Project TrendMiner (a project funded by the European Commission to deliver open-source, real-time methods for mining and summarizing online media) and a linguistic processor based on Daedalus’s commercial MeaningCloud technology. Put simply, the system analyzes comments on social media with natural language processing techniques that can “translate” colloquial descriptions into more structured information. As well as identifying drug names, illnesses and effects, the system registers co-occurrences too; for example, when looking at anti-anxiety drugs, the system can take into account references to the active ingredient, generic name, or commercial brand name, and also pick out references to therapeutic effects and adverse effects.
The researchers envision the technology being used by pharma companies to listen in on what people are saying online about their drugs, or to gather information on suspected ADRs that could be used to supplement existing information sources.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
References
- I. Segura-Bedmar, R. Revert and P. Martinez,“Detecting Drugs And Adverse Events From Spanish Health Social Media Streams”, Proceedings of the 5th International Workshop on Health Text Mining and Information Analysis (2014). I. Segura-Bedma et al., “Exploring Spanish Health Social Media for detecting drug effects”, BMC Medical Informatics and Decision Systems (in press - 2015).
