Skip to content

LXZE/DL_SA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DL_SA (Deep learning on Sentiment Analysis)

The repository of thesis "Thai Comment Sentiment Analysis on Social Networks with Deep Learning Approach"

Supervised by: Asst. Prof. Jumpol Polvichai, Ph.D. and Asst. Prof. Nuttanart Facundes, Ph.D.

This repository mainly contains code for performing sentiment prediction and notebook files for recording the training process and some visualization.

Also, it includes tweet crawler and text proprocessing code which work with Twitter data and Thai sentence specifically.

Thesis abstract

In recent years, many people have posted their comments publicly on social media or websites. As a result, there is a large amount of text data available that could be analysed to gain some insights from users, which can be done by Sentiment Analysis. However, analysing human language data regarding its semantics is difficult because the machine does not have prior knowledge in a language. Therefore, Deep Learning techniques are introduced, as it shows the effectiveness in analysing a massive amount of data. Several applied Deep Learning works have experimented on English and show a satisfying result. Thus, it is an opportunity to explore and perform Sentiment Analysis on Thai online documents using Natural Language Processing techniques and Deep Learning algorithms. There are two main problems to be solved in this work. The first problem is the word ambiguity manipulation, and the other task is an automatic sentiment classification. This work shows our process of Thai online document cleansing to handle errors in raw Thai texts. Also, this work describes the entire process, from data collection, experimentation, evaluation, to the findings of the most suitable Deep Learning algorithm to classify the sentiment polarity from a given document. The results show that every Deep Learning model yields high accuracy and has relatively similar performance of sentiment classification, and a model using the One-dimensional Convolutional Neural Network requires the least time to train. The results can be used for future development in Thai Natural Language Processing and Sentiment Analysis.

Keywords: Deep Learning/ Natural Language Processing/ Sentiment Analysis/ Thai Language

Slide


Requirements

  • python 3.6 ++
  • pandas
  • numpy
  • gensim
  • pythainlp
  • tensorflow
  • keras
  • sklearn
  • tweepy

Involved repository & project

About

Sentiment Analysis with deep learning model

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published