Please use this identifier to cite or link to this item:
|Title:||REAL-TIME TWEET SENTIMENT ANALYSIS BASED ON DEEP LEARNING||Authors:||FENG PIAOPIAO||Keywords:||Real-time system, Sentiment analysis, Deep learning, Optimization||Issue Date:||20-Aug-2018||Citation:||FENG PIAOPIAO (2018-08-20). REAL-TIME TWEET SENTIMENT ANALYSIS BASED ON DEEP LEARNING. ScholarBank@NUS Repository.||Abstract:||Sentiment analysis finds the sentiment polarity (e.g., positive or negative) of text documents, which is widely used in commercial practice for marketing, product feedback, etc, aiming to improve the products or services. Deep learning models have shown state-of-the-art accuracy for sentiment. However, they incur high computational cost, which is a big challenge for real-time deployment over large-scale data streams. Towards this challenge, in this thesis, I propose an efficient sentiment analytical pipeline, which consists of three phases. The first phase samples tweets of popular entities to reduce the workload of latter phases; the second phase uses cheap shallow models to prune “easy” tweets; the third phase applies the state-of-the-art deep learning models to process “hard” tweets. A distributed system, namely ESAP, is developed with my proposed optimization pipeline. The experimental results over the public Twitter dataset confirm that my system can trade off between high efficiency and high accuracy.||URI:||http://scholarbank.nus.edu.sg/handle/10635/151097|
|Appears in Collections:||Master's Theses (Open)|
Show full item record
Files in This Item:
|FengPP.pdf||1.42 MB||Adobe PDF|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.