Sentence Generation using LSTM Based Deep Learning

Abstract

Sentence generation serves the process of predicting relevant words in a specific sequence. The purpose of this research is to come up with a method for generating sentences while maintaining proper grammatical structure. Here, we have implemented a sentence generation system based on Long Short-Term Memory (LSTM) architecture. Our system generally follows the basics of word embedding where words from the dataset get tokenized and turned into vector forms. These vectors are then processed and passed through a Long Short-Term Memory layer. Successive words get generated from the system after each iteration. This process winds up generating relevant words to form a sentence or a passage. The results of the system are pretty convincing compared to different existing methods.