Machine Learning Based Real-Time Sentiment Analysis | TensorFlow

Overview

Inside every tweet resides an explicit and implicit emotion. These tweets can be vastly classified into either Positive or Negative or Neutral based on the core sentiment. But to further understand the Personality of a person deeply, the classification must be based on real-time and more defined sentiments like Happiness, Sadness, Anger, Hate, Confused, etc.

Solution

null

The contents of the tweet are pre-processed, and irrelevant data are dropped.

null

The tweets along with their respective sentiment labels are split into Train and Test set are then passed into 3 models, namely Universal Sentence Encoder (USE), LSTM and doc2vec model.

null

The performance analysis is done on the models, and it is concluded that the USE model works the best among the available options.

Challenges

null

We must be careful that the Sentiment labels considered are not redundant.

null

The programmer must be well versed and have a thorough understanding of the Sentiment Labels himself.

null

LSTM and doc2vec model requires large amount of training data and the performance entirely depends on the scale of the training dataset.

Technology

Related Success Stories

Related Insights

Free Consulting