The use of Textblob for multiple domain tweets for sentiment

Question

I'm building an application that analyze sentiment for news-related tweets in different domains, such as sports, disaster and technology, I'm using Textblob with the default mode (PatternAnalyzer). Does that provide a good sentiment even though domains are different? And how can I evaluate its performance? Or is it better to provide my own training data for each domain and train a classifier?

Answer 1

Textblob is a basic sentiment predictor and it wont be accurate. Its basic training model is trained on movie-review dataset which wont work efficiently for you. I would suggest you to create different datasets for each, if possible.