Improved Deep Learning Model for Aspect Words Extraction

Deep learning has recently been successful in numerous academic areas, including social network filtering, image processing, machine translation, and text mining. There is a lot of user-generated content available on various social media sites. that can be used for precise opinion analysis of the shifting market dynamics. Aspect word extraction (AWE) task detects the targets of user opinions from the text in aspect-level sentiment analysis (ABSA). The majority of aspect retrieval methods depend on a certain domain or context. Additionally, it is difficult to identify the most pertinent aspect phrases that are domain-specific because there is a deficiency of topic and situational knowledge. The current methods fail to extract some useful aspect terms because they are unable to identify noun phrase dependencies over time. In order to enhance the AWE challenge, this research suggests a combined language patterns with deep learning methods. On typical review datasets, results from studies and experiments are examined for many kinds of aspects.

I. Introduction

People now depend on the Internet and social media platforms to get their reviews on a daily basis. The enormous increase in processing power and data accessibility has significantly altered the way and the degree to which organizations base their decisions on technologies. Businesses are increasingly wanting to employ advanced technologies, which is a distinct trend to support daily decision-making and service delivery [1]. Unstructured content from such consumer comments has been recognized as an important form for knowledge to advance several application fields, but processing enormous amounts of user-generated unstructured text is time-consuming and impractical. These reviews are an important resource for manufacturers to raise the caliber of their goods [2]. It's been more typical in the recent past for user-generated web material to focus on topics covered in reviews.