From 4-5 December, the CCC team will be in Basel, Switzerland to attend BioData World Congress, the world’s leading event for big data and AI in life sciences and healthcare.
CCC is proud to be featured among the 250+ speakers at this year’s event.
“Biomedical semantic indexing with MeSH”
In our session, CCC’s Principal Solutions Architect Robin Bramley will present a novel neural network architecture that consists of a network-of-networks endowed with an attention mechanism for the purpose of biomedical semantic indexing based on MeSH. This network learns both the appropriate labels for each document and the relationship between the labels for a given text. We’ll share results up to depth three of the MeSH hierarchy. Tagging at the third level of the hierarchy allows for more specific metadata, yet due to the hierarchical nature these tags can still be included in broader category searches.
The session will also cover text pre-processing, hierarchy representation, feature descriptors, and model evaluation. For the latter, in addition to micro- and macro F1, we consider metrics that distinguish between accuracy and precision; a distinction that is of practical importance when evaluating results for hierarchical classification. State-of-the-art classification performance gives greater confidence in automatic tagging.
Robin has over 20 years’ experience in technology consulting across a number of industry sectors, primarily publishing and financial services, but more recently life sciences. Robin is a proponent of open standards, open data and Open Source; his work on text mining and knowledge representation with the labs team makes extensive use of big data processing frameworks, machine learning and graph technologies.
Be sure to stop by Stand #11 and say hello.
We’re here to talk to you about your data and information integration challenges and how solutions from CCC can help.
Follow all of the action with @copyrightclear and the hashtag #BioDataCongress.
Related Resources: