Understanding and Improving Disability Identification in Medical Documents

Abstract

Disabilities are a problem that affects a large number of people in the world. Gathering information about them is crucial to improve the daily life of the people who suffer from them but, since disabilities are often strongly associated with different types of diseases, the available data are widely dispersed. In this work we review existing proposal for the problem, making an in-depth analysis, and from it we make a proposal that improves the results of previous systems. The analysis focuses on the results of the participants in DIANN shared task was proposed (IberEval 2018), devoted to the detection of named disabilities in electronic documents. In order to evaluate the proposed systems using a common evaluation framework, a corpus of documents, in both English and Spanish, was gathered and annotated. Several teams participated in the task, either using classic methods or proposing specific approaches to deal effectively with the complexities of the task. Our aim is to provide insight for future advances in the field by analyzing the participating systems and identifying the most effective approaches and elements to tackle the problem. We have validated the lessons learned from this analysis through a new proposal that includes the most promising elements used by the participating teams. The proposed system improves, for both languages, the results obtained during the task.

Advances in Monolingual and Crosslingual Automatic Disability Annotation in Spanish

“…Regarding other languages, several annotated corpora have been used, as the IxaMed-GS corpus [21], conformed by Electronic Health Records (EHR) written in Spanish annotated with drugs and diseases. In addition to all these corpora and tasks, initiatives focused on specific and less studied types of medical entities, such as the corpus used in the DIANN shared task [14] have also emerged.…”

“…The DIANN shared task [14] was dedicated to the detection of disability mentions in biomedical research texts in English and Spanish, with the objective of evaluating the performance of various named entity recognition systems in two different languages. In the first position, [40] presented a neural network-based architecture system consisting of a Bidirectional Long Short Term Memory network (BiLSTM) and a Conditional Random Field (CRF), using static word embeddings for both languages combined with a rule-based module for the detection of disability-related abbreviations, obtaining an F-measure of 0.82 and 0.78 for English and Spanish, respectively.…”

“…In the first position, [40] presented a neural network-based architecture system consisting of a Bidirectional Long Short Term Memory network (BiLSTM) and a Conditional Random Field (CRF), using static word embeddings for both languages combined with a rule-based module for the detection of disability-related abbreviations, obtaining an F-measure of 0.82 and 0.78 for English and Spanish, respectively. More recently, [14] uses a long short-term memory architecture for disabilities, improving the state of the art, with an F-measure of 0.83 and 0.81 for English and Spanish.…”

“…Table 2 presents the different types of systems that will be compared in this paper. On the monolingual part (upper side of [14] + character features (general texts) F LAIRF T [41] FASTTEXT contextual static word & subword (general texts) Spa. F LAIR W iki2V [41] Wikipedia2Vec static word F LAIR SkipN G EHR SkipNG character [45] static word (EHR) F LAIRLM EHR FLAIR contextual [41] character (EHR) Crosslingual approaches…”

LearningToAdapt with word embeddings: Domain adaptation of Named Entity Recognition systems

“…The possibility to easily access and exploit these sophisticated NER systems through APIs or pre-trained models became fundamental for addressing tasks such as Data Integration [34,58], Question Answering [55,5,73], Privacy Protection [20,46] , and Knowledge Base Construction [56].…”