Innoradiant's help for research on Covid-19:

AsI-Health, an open text mining platform over Covid related scientific texts.

What does AsI-Health provide ?

We believe that in this period everybody’s skills and competences should be devoted to give a contribution to the battle that human kind is engaging against Covid-19. On our side the only contribution we can give is in terms of semantic analysis of medical texts. For this reason we worked hard to provide an application (AsI-Health) which could help scientists involved in research on Covid-19 to find a path across the huge amount of scientific literature published so far.

What we did was to configure our technology for working on very specific medical texts. In doing that we integrated our AsI technology with a set of open source software (reported below) in order to provide a maximal precision to our analyses.

The basic idea is that researchers in the field might need something more than just a textual search engine to dig in the amount of available literature and gather paper or claims useful to their research. AsI-Health implements and extends state of art technologies in Open Information Extraction, in order to capture most salient relations expressed in sentences and link the actors who play a role in this relations.

News & Updates

- Integration of Elsevier Dataset (papers since 01/01/2020) 18/05/20
- New version with negated relations 18/03/2020
- The bioRxiv dataset was updated to 25/03/2020
- The bioRxiv dataset was updated to 14/04/2020

Which data did we use?

The dataset is constantly evolving. As of today, we include all preprint related to Covid-19 and SARS-CoV-2 made available via bioRxiv. Updates are performed regularly and systematically reported in the “News and Updates” panel. Needless to say they are preprints, so they did not necessarily undergo peer review: as the bioRxiv site says “They should not be regarded as conclusive, guide clinical practice/health-related behavior, or be reported in news media as established information.”

In the next days the Elsevier dataset will be added to the document base, with scientific paper on Coronavirus related matters since 2010 (the choise of this subset is due to a limitation of computational capabilities and to a priority given to novel information). In the next weeks we hope to include also contributions from other publishers, such as  Springer Nature | Wiley | NEJM | BMJ | American Society for Microbiology | American College of Cardiology| Chongqing VIP Information

Who is the application for?

We want to to offer the state of the art NLP research to research against Covid-19, so the application is mainly targeted towards people involved in studies and experimentation on that. Unfortunately, as a small enterprise, we can't rely on the computational power needed to grant fully unlimited access to the application. As a consequence we ask all people willing to use it to contact us at the address below with their professional email address. Upon reception we will release username and password for free unlimited access.  


How could anyone contribute?

It is not easy for us to contact all the scientists involved in research against Covid-19. So the best contribution you could give to this initiative is just to circulate the link to this page, in the hope that it reaches someone who really need the application to advance in her findings. Another contribution can be provided by companies which decide to sponsor this initiative. In this case we would ask them to help us in building a more responsive infrastructure, both in terms of navigation fluidity (to help the user to find information more rapidly) and in terms of background processing (in order to integrate more sources and perform updates more regularly)

Who helped us?

As we said, we integrate a lot of open source software in AsI-Health. Here is a non-exhaustive list of projects from which we took modules in order to make the processing more precise and “intelligent”.

Stanford CoreNLP is used for dependency parsing.

Apache cTAKES™ is used for extracting medical concepts out of scientific literature

Dexter is used to detect concepts entering into relations and disambiguating them according to Wikipedia concepts

Kibana/Elastic is used to index all the extracted information and explore the information space

To ask technical questions
To request access credentials or propose sponsorship