Using recurrent neural networks for part-of-speech tagging and subject and predicate classification in a sentence

Abstract

In natural language processing the use of deep learning techniques is very common. In this paper, a technique to identify the subject and predicate in a sentence is introduced. To achieve this, the proposed technique completes POS tagging identifying in a later stage the subject and the predicate in a sentence. Two different deep neural networks are used to complete this process. A first one to establish a correspondence between individual words and part-of-speech (POS) tags and a second one that, taking as input these tags, identifies relevant elements of the sentence such like the subject and the predicate. To validate the architecture of our proposal a set of tests over public datasets have been designed. In these experiments, this model achieves high rates of accuracy in POS tagging and in subject and predicate classification. Finally, a comparison of the results obtained for each individual network with similar tools such as NLTK, pyStatParser and spaCy is made.