Performing (surprisingly-easy!) Sentiment Analysis on Google Cloud
How to train and deploy a serverless Sentiment Analysis model and API to Google Cloud
Sentiment Analysis takes the text and identifies the emotional meanings behind it. This is useful in situations such as identifying negative support requests, analyzing how people feel about your brand or product, and better understanding customer needs in general.
BERT (Bidirectional Encoder Representations for Transformers) is a wonderful machine learning technique for natural language processing. Due to its architecture, it frequently outperforms other models, making it our “weapon of choice” for performing sentiment analysis.
However, its training process can be costly. Luckily, there are actions you can take to minimize that.
After reading this article, you’ll know how to implement BERT, reduce BERT’s training costs through fine-tuning techniques, and deploy the model to Google Cloud Run to serve it as a production-ready API.
Before you continue — First check whether there is an API that would do a good enough job for your use case. Try existing APIs like Google Cloud Natural Language Sentiment Analysis before going in too deep with a custom ML solution.