#Sentiment Analysis on Movie Reviews - Using Support Vector Machines with Linear Kernel This project is part of the course CZ4041 (Machine Learning) at NTU. Kaggle challenge: "Classify the sentiment of sentences from the Rotten Tomatoes dataset"

https://www.kaggle.com/c/sentiment-analysis-on-movie-reviews

##Kaggle Project Description "There's a thin line between likably old-fashioned and fuddy-duddy, and The Count of Monte Cristo ... never quite settles on either side."

The Rotten Tomatoes movie review dataset is a corpus of movie reviews used for sentiment analysis, originally collected by Pang and Lee [1]. In their work on sentiment treebanks, Socher et al. [2] used Amazon's Mechanical Turk to create fine-grained labels for all parsed phrases in the corpus. This competition presents a chance to benchmark your sentiment-analysis ideas on the Rotten Tomatoes dataset. You are asked to label phrases on a scale of five values: negative, somewhat negative, neutral, somewhat positive, positive. Obstacles like sentence negation, sarcasm, terseness, language ambiguity, and many others make this task very challenging.

##Abstract From Project Report This paper covers an implementation of a sentiment analysis on movie reviews. The implementation involved Support Vector Machine (SVM) with linear kernel, evaluating two different strategies - One-VS-One and One-VS-All, with different pre-processing methods.

The goal of this project was to implement an accurate prediction in terms of the domain for the Kaggle competition, "Sentiment Analysis on Movie Reviews". The best result generated using the following pre-processing methods: Only consider letters and lowercase symbols, stemming, normalized TF-IDF and removing 0.00004 % of the least frequent words. One-VS-One generated a higher score than One-VS-All, with C set to 0.32. The final result scored the 30:th place on the leader board resulting in the top 3.48 % of all the submissions. (For more information, read the Project Report)