Weakly Supervised Deep Learning for Customer Review Sentiment Classification

Abstract


Sentiment analysis is one of the key challenges for mining online user generated content. In this work, we focus on customer reviews which are an important form of opinionated content. The goal is to identify each sentence’s semantic orientation (e.g. positive or negative) of a review. Traditional sentiment classification methods often involve substantial human efforts, e.g. lexicon construction, feature engineering. In recent years, deep learning has emerged as an effective means for solving sentiment classification problems. A neural network intrinsically learns a useful representation automatically without human efforts. However, the success of deep learning highly relies on the availability of large-scale training data. In this paper, we propose a novel deep learning framework for review sentiment classification which employs prevalently available ratings as weak supervision signals. The framework consists of two steps: (1) learn a high level representation (embedding space) which captures the general sentiment distribution of sentences through rating information; (2) add a classification layer on top of the embedding layer and use labeled sentences for supervised fine-tuning. Experiments on review data obtained from Amazon show the efficacy of our method and its superiority over baseline methods.


Subscribe to access this work and thousands more
Overall Rating

0

5 Star
(0)
4 Star
(0)
3 Star
(0)
2 Star
(0)
1 Star
(0)
APA

Chandan Nune, S. (2021). Weakly Supervised Deep Learning for Customer Review Sentiment Classification. Afribary. Retrieved from https://tracking.afribary.com/works/deep-learning-for-product-reviews-sentiment-analysis

MLA 8th

Chandan Nune, Saathwik "Weakly Supervised Deep Learning for Customer Review Sentiment Classification" Afribary. Afribary, 04 Sep. 2021, https://tracking.afribary.com/works/deep-learning-for-product-reviews-sentiment-analysis. Accessed 06 Nov. 2024.

MLA7

Chandan Nune, Saathwik . "Weakly Supervised Deep Learning for Customer Review Sentiment Classification". Afribary, Afribary, 04 Sep. 2021. Web. 06 Nov. 2024. < https://tracking.afribary.com/works/deep-learning-for-product-reviews-sentiment-analysis >.

Chicago

Chandan Nune, Saathwik . "Weakly Supervised Deep Learning for Customer Review Sentiment Classification" Afribary (2021). Accessed November 06, 2024. https://tracking.afribary.com/works/deep-learning-for-product-reviews-sentiment-analysis

Document Details
By: Saathwik Chandan Nune Field: Marketing Type: Paper 7 PAGES (3002 WORDS) (pdf)