CodiEsp: AICOS team stands out in natural language processing competition



A Fraunhofer AICOS team – The Mental Strokers –, led by researcher João Costa, was runner up in an international competition on automatic clinical coding (CodiEsp - CLEF eHealth 2020). From a total of 168 submitted runs by 22 teams from 11 different countries, AICOS’ Mental Strokers stood out.

The yearly CLEF eHealth - Multilingual Information Extraction task represents one of the first efforts to promote the development and evaluation of automatic clinical coding systems for medical documents. This year’s competition was divided in three sub-tracks: Diagnosis Coding, Explainable AI, and Procedure Coding. In the later sub-track, AICOS’ Mental Stroker team classified second.

Information extraction from EHRs written by healthcare professionals is a rather complex task. The rapid data accumulation, leading to large amounts text charact high diversity of types, structures, formats, together with the rich medical vocabulary make this process particularly challenging. This complexity highlights the need to translate the information locked in free text into structured clinical information. Additionally, transforming this information into coded clinical documents, is a task currently performed by experts only but increasingly relevant. The task was performed in Spanish text highlighting the importance of clinical coding in languages other than English to promote automatic clinical coding systems.


You can read the full paper here.


*Photo: Wordcloud used by FhP-AICOS's team for the presentation which shows the most frequent terms of the analyzed reports.