ExSim at eHealth-KD Challenge 2020 Combining NLP and Word Embeddings for Entity Recognition

This paper describes the system submitted to the eHealth-KD Challenge 2020-Task A: entity recognition. The system utilizes a supervised learning methodology to recognize entities within Spanish texts; namely, it applies NLP and word2vec techniques to create a unique labeled dictionary of entities in the training set. These labels are propagated into new entities that are found in the testing set via semantic similarity measurement. The simplicity of our system shows low performance with F1=0.32, precision=

0.29, and recall=0.34. Finally, the system is discussed from different aspects: challenges, earlier attempts, current system’s characteristics, and possible future work.