Reconhecimento de entidades nomeadas em textos de boletins de ocorrências

Abstract

Classifying entities in a given text is a challenging problem in Natural Language Processing (NLP). A common technique that deals with this problem is Named Entity Recognition (NER). Deep Learning techniques have been widely applied to NER tasks because they require little resource engineering and are free of language-specific resources, learning important resources from combinations of words or characters trained in large amounts of data. However, these techniques require a huge amount of data for training. However, these techniques require a huge amount of data for training. This work proposes Human NERD (stands for Human Named Entity Recognition with Deep learning) which addresses this problem by including humans in the loop. Human NERD is an interactive framework to assist the user in NER classification tasks from creating a massive dataset to building/maintaining a deep learning NER model. Human NERD framework allows the rapid verification of automatic named entity recognition and the correction of errors. It takes into account user corrections, and the deep learning model learns and builds upon these actions. The interface allows for rapid correction using drag and drop user actions. The present work also proposes two NER models, one using the Spacy framework and the other the Keras library, both complement each other.