lshhhhh/textual-emotion-recognition

[CSED499I] Emotion Recognition from Korean Text - 2017 Spring

EmotionRecognition

Emotion recognition from text

Introduction

• Topic: To recognize emotion from text
• Tasks
  • Pre-processing
  • Feature extraction
  • Classification

Implement various kinds of classifiers with various methods and evaluate.

• Consider pre-processing with Morphological analysis, emoticon, punctuation mark, etc.
• Consider feature extractor with Word count, TFIDF, phrase(with polarity), etc.
• Implement various classifiers. - Naïve Bayes, MaxEnt, SVM, deep learning(RNN, CNN)

Requirements

• This program must be able to detect emotion from text.
• The program will include 7 kinds of emotion - Love, joy, surprise, anger, sadness, fear and neutral.
• A text must be classified by one kind of emotion(most likely emotion).

Project Schedule

• Information Gathering & Understanding the Topic (3/6 ~ )
• Identify research topics - the definition of emotions (academic research), etc.
  1. Study Machine Learning and Natural Language Processing
  2. Study Python
  3. Read related articles
• Requirements Analysis (3/14 ~ 3/15)
• Set Direction of Research and Design (3/15 ~ 4/1)
• Simple Implementation (4/1 ~ 4/6)
  • Build emotion data set ver.1
  • Implement simple Naïve Bayes classifier
  • Evaluate performance with 5-fold cross validation
  • Study how to crawl tweets
• Implementation
  • Classifiers
    1. Naïve Bayes
    2. MaxEnt (4/13 ~ 4/19)
    3. SVM (4/20 ~ 4/26)
    4. Deep learning(RNN, CNN) (4/27 ~ 5/15)
  • Consider pre-processing with
    • Morphological analysis, emoticon, punctuation mark, etc. (4/13 ~ 4/26)
  • Consider feature extractor with
    • Word count, TFIDF, phrase(with polarity), etc. (4/27 ~ 5/15)
  • Testing and debugging (5/15 ~ 5/26)
    1. Evaluate performance
    2. Additional data collection and improvement
    3. Final performance evaluation
  • Poster presentation to the CSE professors and students (6/2)
  • Demo and submitting final report (6/2)