A Survey on Machine Learning Based Keyphrase Generation in Natural Language Processing

Abstract

With the exponential increase of available textual data, transforming the natural language content into potential information becomes crucial to assist the prominent application domains. In Natural Language Processing (NLP) applications, keyphrase generation has recently become an increasingly prominent research topic. Even though numerous advancements are realized in the current keyphrase generation models, the proliferation of neural network models profoundly impacted natural language generation to a new era. Over the past several years, various studies on keyphrase extraction and generation have been developed that deliver significant contributions to the current state of keyphrase generation research. The researchers confront understanding the deep insights for further developments from the conventional keyphrase generation research while adopting the deep neural network models. Hence, this work notably focuses on studying the keyphrase generations with the impact of the exploding deep learning methods. This survey briefly introduces keyphrase generation in NLP and reviews the recent abstractive methods using deep learning for meaningful keyphrase generation that achieves stateof-the-art performance. By determining and discussing the shortcomings in the previous machine learning and deep learning-based keyphrase generation, this work facilitates strong groundwork for understanding the recent developments in keyphrase generation. Also, it discusses the research challenges in keyphrase generation from the perspective of both text mining and deep learning models. Thus, this study lays the foundation for the researchers to develop potential solutions to resolve the research constraints in keyphrase generation.