The Human Language and Accessibility Technologies Group (HULAT) is part of Computer Science and Engineering Department of Universidad Carlos III de Madrid. We participate in R & D projects related to natural language processing (NLP) in biomedical domain (biomedical literature, clinical narrative and social media), usability and accessibility of user interfaces. We work in close collaboration with different companies.
Research is focused in two main lines:
- Natural Language Processing (NLP): information extraction, named entities recognition, text simplification, semantic knowledge representation (ontologies and corpora) with special focus in biomedical domain.
- Accessibilty: methodological frameworks for accessible applications as well as accessitiliby to products and services, integration of accessibility requirements in the software development process and in user interfaces for people with disabilities.
In the Big Data era there is an increasing need to fully exploit and analyse the huge quantity of information available about health. The adoption of the Electronic Medical Record (EMR) was a relevant milestone to keep trace of patients, but not all its information is available yet to improve protocols and decision making. Moreover, other information sources, such as biometrical data or publications on recent research on pathologies and experimental treatments, could be integrated in order to get a better picture of the situation and take action in a more informed and scientific way. Also patients do need support to fully understand their pathologies, treatments and medication. New techniques and tools can contribute to improve communication and interaction among health workers, patients and data in order to support better decision making. New Natural Language Processing (NLP) techniques to analyse and process biomedical literature, clinical narrative and health-related documents are required to identify patterns, correlations or semantics (by means of semantic modelling, machine learning and other AI approaches) that might be relevant to understand and structure data and take the right decision
Currently, the group work in DeepEMR project(TIN2017-87548-C2-1-R), devoted to apply deep learning methods to process clinical notes from EMR of a Spanish hospital (HUFA) that collaborates with us. HULAT, has a prolonged experience in competitive research projects related to NLP applied to biomedical domain. Some of these projects are ISSE project (FIT-350300-2007-75), for enhancing semantic interoperability in electronic health systems, MULTIMEDICA project (TIN2010-20644-C03-01), whose aim was to develop information extraction techniques applied to scientific and informative literature about health and TrendMiner european project (FP7-ICT 287863), to research if social networks could be a valuable source of information to detect drug adverse events not detected in previous clinical trials.
In the last years we have developed several resources related to NLP in the scope of IMI such as the Drug-Drug Interaction (DDI) corpus and DINTO ontology to help in name entity recognition an relation extraction tasks from health-related texts (Medline abstracts, Drug Label, clinical notes,…) . Concerning drugs related domain, we have organized and participated in some important shared tasks as DDI extraction 2011, SEMEVAL Extraction of Drug-Drug Interactions from Biomedical Texts 2013, Adverse Drug Reaction Extraction from Drug Labels 2017 and FDA Adverse Drug Even Challenge 2019.
HULAT has also participated in organizing conferences and workshops, such as SEPLN 2008 and BioSEPLN 2010 related to NLP and DSAI 2013, Interacción 2016 and DSAI 2016 as well as all editions of AMADIS related to accessibility.
Regarding Accessibility, the group has a wide experience in multimodal and multiplatform user interfaces design folowing Model-Based User Interface Development (MBUID) frameworks that allow the definition of abstract interfaces in design time and that provide an solution space to face the complexity of application user interfaces taking into account user heterogeneity. This experience has been developped in projects such as eGovernability (TIN2014-52665-C2-2-R) to reseach accessible and usable user interfaces for electronic public administration (eAdministration)
We have experience in technology solutions which allow presenting the information more effectively serving the needs of different groups of users (people with disabilities, elderly people, etc.). In this line, we work in the design of user-friendly and tailored user interfacesBesides, an important research area that combines accessibility to information and NLP is text simplification to help patients to better understand health-related information. In this line we have proposed methods for lexical simplification of adverse effect in Spanish drug package inserts.
Members of the group belong to the Spanish Centre of Subtitling and Audiodescription (CESYA) in collaboration with Real Patronato sobre Discapacidad, focused on promoting and developing tools and services related to the audiovisual accessibility. One of the objectives is providing accessibility to audiovisual media for people with disabilities. In the area of subtitles we have developed several tools for generating and broadcasting subtitles. We also work on systems for synchronizing the subtitles with the audio of an event. Regarding of accessibility services in the TDT, we have developed tools for monitoring the accessibility services provided by the TV broadcasters. In the area of education, we are working on including accessibility services in the classes.
CESyA provides enterprises and public organizations with consulting and training services of audiovisual accessibility. As members of CESyA we collaborate in public initiatives relatives to the accessibility in Spain.