People do indeed love to share information on social media, including health data. Yes, it's true! This is the starting point of the research paper entitled "Exploring Spanish health social media for detecting the effects of drugs on health", which aims to monitor social media conversations to find out how users talk about their relationship with drug consumption.

This allows us to identify possibly adverse effects of these drugs that were unknown till now. Although there is an official protocol to report some undocumented adverse effects, only 5-20% of them are actually reported.

In addition, you can analyze conversations about drugs, symptoms, conditions and diseases to acquire more information. For example, you can find out how users look for certain drugs on social media or how others sell them, often illegally. Still others talk about mixing alcohol with drugs or other illegal substances.

Of course, not everything that appears on the Internet is reliable (that's another issue), but it can certainly motivate users to form new hypotheses.

The research team Group Advanced Databases of the Carlos III University of Madrid has managed to develop hybrid models to obtain the knowledge necessary to identify these adverse effects.

Meaningcloud is the natural language processing platform that allows analysis based on these models. The customization options offered by this platform have been critical for integrating the specific vocabulary found in medicine and other domains.

As we know, names of drugs and symptoms can be very complicated and are often misspelled. The results of the algorithm are promising, because its recall is 10% higher than other known algorithms. You can find more details in this article published in the magazine BMC Medical Informatics and Decision Making Journal.

These developments are part of the TrendMiner project and can be found on the website: TrendMiner Health Analytics Dashboard, which gathers feedback from social media users on antidepressant drugs.

The console displays antidepressants along with related symptoms; by clicking on each of them you can see how the feedback changes over time. The original text appears at the bottom of the page along with the identified names of drugs, symptoms, diseases and any relationship between them.

These relationships can reveal whether a drug is appropriate for a particular symptom or if the drug actually has adverse effects. The prototype also allows searches using the ATC code (Anatomical Therapeutic Chemical Classification System) and the corresponding level according to this classification system. If the "By Active Substance" ['active principle'] option is anabled, any drug containing the active ingredient will be included in the search.

Moreover, the predictive search functionality lets you easily find the right expression for a drug or disease. Please check out the prototype and tell us what you think. And if you find one of the graphics, you can tweet it from anywhere! Any comments are more than welcome!