Generate Contextual Insight of Product Review Using Deep LSTM and Word Embedding

• Published in: Journal of physics. Conference series Vol. 1577; no. 1; pp. 12006 - 12015
• Main Authors: Hanafi, Suryana, N, Basari, ASH
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.07.2020
• Subjects: Customers; Embedding; Internet; Physics; Product reviews; Recommender systems; Social networks; User behavior
• ISSN: 1742-6588; 1742-6596
• DOI: 10.1088/1742-6596/1577/1/012006

Nowadays, in every day live, majority people face in many internet options. For example, what meal to eat, what news to read, what vehicle to ride, what the best path to travelling, what the best group in social network to joint, what the best video to watch, what the best video in YouTube to watch and etc. The best way to recommend the internet content to customer is by using recommender system. Recommender system calculate product recommendation by detecting user behaviour in the past. The user behaviour in the past was being variable to compute similarity between many customers. One of the majority user behaviour is in the term of document. Most of document interpret model in recommender system use traditional NLP model such as TF-IDF, LDA model. According to NLP point of view, traditional NLP face the weakness in contextual understanding. Aims to handle the problem on above, we proposed novel model to generate contextual understanding by involve two important aspect considered subtle word and word sequence. We implemented word embedding based on GLOVE and detecting word sequential using RNN-LSTM. According to qualitative evaluation report, our model successful to capture contextual insight of the document of movie review by IMDB. This model suitable to integrated with latent factor based on matrix factorization to generate product recommendation in Collaborative filtering model.