:
After going through various blogs and research papers, I performed sentiment analysis on tweets. As of now, only positive and negative classfication was done.
In today’s world, text is considered to be the most common means of information exchange. But day by day, the heap of text is increasing on internet making it difficult for us to churn out valuable information of it. This is where text mining plays its role by enabling better understanding of information out of these set of texts so that we can come up with smart decisions. Various machine learning techniques can be used for the text classification to get valuable information from the text data. To get better results in terms of enhanced accuracy, data cleaning and feature selection methods can be applied before any text mining technique, which in our case is sentiment analysis. Data cleaning and feature selection can improve the learning process with less computational complexity and storage requirements. In this research we have used a number of data cleaning and feature selection techniques along with the selected classification algorithms like NB, MNB, BNB, LR and SVM. Experimental results show that these techniques are efficient and enhance the classification performance. Here we have used NLTK corpus for movie reviews and an online dataset of tweets for training and testing purpose.
Key Words: Text Mining, Sentiment Analysis, Data Cleaning, Corpus, NLTK, Machine Learning, Feature selection.
NLTK dataset: C:\Windows.old\Users\your_user_name\AppData\Roaming\nltk_data\corpora\movie_reviews\
-
twitter sentiment analysis has been done focussing one feature selection and data cleaning
-
Mainly focussed on naive bayes algorithm