Machine learning with word embeddings applied to biomedical concept disambiguation

Abstract

Artificial Intelligence (AI) has grown in the last years and it has many applications. Natural Language Processing is one of the AI tasks, which has the objective to endow the machines the capability of understanding human language. This is an important process due to the amount of information stored in textual form. There is a growing need for automatic extraction of knowledge, and NLP comes in this direction helping in tasks such as information extraction and information retrieval. Word sense disambiguation is an important NLP subtask, which is responsible for assigning the proper concept to an ambiguous word or term. In this paper, we present results obtained from applying supervised machine learning algorithms with local features, and word embeddings as global features extracted from Wikipedia and PubMed knowledge sources. These results indicate that word embeddings features are informative and may improve the biomedical word disambiguation accuracy.