Convolution Neural Networks Machine Learning model to classify the polarity of tweet using Gensim Word2Vec

Twitter is a platform used by people to express their opinions and display sentiments on different occasions. Twitteris one of the non-traditional data sources with unlimited potential. It contains a large reserve of data sets which are easier to access and collect when compared to others. One of the available tools is twitter API which lets us collect data and various other information about the tweets. The tweets are collected through Tweepy and the tweets are being labeled through Text-Blob. The process of analyzing the tweets and the texts and emojis in them to classify the opinions of the people and their reaction towards the tweet, along with the viewpoint the owner of the tweet into positive, negative or neutral. To increase the processing of the tweets are preprocessed during collection of data itself. For accuracy we will create a vocab using the W2R (word to vector) model containing all the related words. The words are vectorized with the relationship among them. This makes the prediction easier. Then through implementation of CNN (conventional neural network) we insert the embedding layer which is formed with vectorized words and a LSTM (long-short-term-memory) layer. The accuracy value of the model is created and we can analyze the tweets of the particular domain. TheDataset for each text will be chosen separately by using Entity Identification. The major challenges such as Domain Dependency, World Knowledge and Pragmatics can be handled through selection of major entities and obtaining data related to it. This process can give us the current condition or status of the topic involved so that it will be easier to find out the anomalies. Another major challenge Thwarted Expectations can be handled using emoji and hashtags present in the obtained text. The more detailed preprocessed data helps in better end results and the model is a hybrid approach which will outperform the baseline.