Feature extraction and selection (PCA/SVD) + implementation of weak classifiers

[LDiana22]

The main idea is to extract many features from text and perform feature selection to determine what are the most promising ones. Ideas:

• useful feature - most informative unigram and bigram collocations: https://www.researchgate.net/publication/268817500_Sentiment_Analysis_for_Social_Media
• frequency of: nouns, adjectives, punctuation, emoji, elongated words
• frequency of positive/negative words
• frequency of positive/negative emojis
• number of elongated words <- positive/negative words (correct misspelling + identification of the sentiment)

[LDiana22]

Implement classifiers such as Naive Bayes, SVM, Decision tree and report performance

[LDiana22]

Grid search SVM using Weka - https://www.researchgate.net/publication/324835821_SVM_Optimization_for_Sentiment_Analysis

[LDiana22]

https://pdfs.semanticscholar.org/0f4d/f38e308a65ff3fe42878c7fdfa68e17f9212.pdf

[LDiana22]

SVM example implementation https://github.com/Ankushr785/Emotion-recognition-from-tweets/blob/master/emotion_recognizer_svm.py