Sentiment Analysis using Machine Learning

Our project briefly explores achieving the task of sentiment analysis / sentiment classification, utilizing several machine learning techniques.

Github Repository Link: Sentiment Analysis using Machine Learning (Github Repository)

Findings

Both supervised and unserpervised methods can be used to achieve sentiment analysis with moderate success rates.

Notes:

• The number one difficulty in sentiment analysis is discovering and preparing a dataset to train on.
• Supervised sentiment analysis suffers from oversimplification in data sets. We found inconsistencies in text samples containing sarcasm for example.
• Unsupervised sentiment analysis suffers from inability to interpret results. The word clusters seperated nicely, but what were supposed to be positive, negative, and neutral clusters, were more like positive, negative, and Spanish clusters.

Authors

Name	Email
Domin Sweet	domin.sweet@my.utsa.edu
John Le	john.le@my.utsa.edu
Jack Xu	jack.xu@my.utsa.edu

References

• 7. TrackMyHashTag (Twitter Dataset)
• Unsupervised Sentiment Anlysis (Tutorial)
• FastText Vectorization (Gensim Library)
• Supervised Sentiment Analysis (Preprocessing)