Collaborative annotation and mapping tool for clinical concepts

• Autor(a) principal: Neves, André Sousa
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/33928
• topic: Biomedical text mining; Named entity recognition; Natural language processing; Information retrieval; Information extraction; Mapping; Standard vocabulary concepts

Every day new biomedical information is published in the form of research articles, books and reports, but given its unstructured form it is not useful for knowledge acquisition apart from keyword search. Over the years significant interest has been generated towards text mining and the production of structured data using information retrieval and information extraction techniques, namely named entity recognition. Several natural language processing tools were developed with the main purpose of aiding the manual labor-intensive task conducted by expert curators by implementing automatic pre-processing pipelines that annotate biomedical entities and their relationships in literature, along with interactive interfaces to review and validate them. Moreover, it is essential that the data is harmonized into a common standard that everyone can understand no matter what language, format or encoding it was originally recorded in, in order to provide a collaborative effort among researchers. Some tools provide efficient indexing and searching capabilities to map concepts from various domains into standard vocabulary concepts, or in other words are capable of standardize data into a common format which in turn allow collaborative studies to be conducted. Nevertheless, there is a lack of tools that allow to perform both annotation and mapping. This dissertation presents a web-based tool with the intent to fill this gap by allowing experts to still perform each task individually, but also to form a pipeline and use the output annotations as input for the mapping process. As a result, the tool provides an interactive interface that allows the users to upload text documents and annotate biomedical entities present in them, either manually by selecting portions of text or double clicking words, or automatically with Neji’s web services and manage those generated annotations. For mapping, the users can upload CSV documents containing terms to be mapped to standard vocabulary concepts, using Usagi’s open-source code. Moreover, the users can review and validate suggested mappings based on match score.