ABSTRACT
With immense amounts of text data to be analyzed by professionals, analyzing and mining opinion reviews has become a tedious task for e-commerce networks. It also consists of redundancies as the reviews consist of a limited number of aspects. The same problem is faced during sentiment analysis and text summarization too. This project aims to reduce the computational cost by mining the aspects and analyzing the nodes for a representative node among all the review nodes for a crowdsourcing network model. This is done using the bidirectional long short-term memory (LSTM) neural network model, which mines the aspects from the review text data. Review representation is implemented using a PageRank algorithm, which calculates the relativity among the review nodes and selects the representative node. Aspect term extraction is mainly used for the extraction of aspect terms explicitly mentioned in the documents. This provides a huge potential to the Fourth Industrial Revolution (Industry 4.0) by helping in more efficient business analysis procedures with the selection of representative nodes of information for an aspect resulting in time and workforce wastage reduction. A bidirectional LSTM encoder, along with an attention mechanism, is used for the consideration of the relationship between the words and the long-term dependencies between words as a feature. The accuracy of the aspect mining, along with the mapping of aspects according to their respective review nodes which are present in the e-commerce network, increases the accuracy due to the use of an LSTM network. This novel approach can be applied for ranking products and services-based industries.
