Given thousands or product reviews in local language, from different marketplaces, I need to automatically structure all feedback using a Natural Language Processing algorithm and advanced machine learning techniques. The ideal tool for this purpose is AWS Comprehend, from Amazon Web Services.
The process must be as follows:
1. Identify topics that consumers write on reviews related to a specific category.
2. Perform category sentiment analysis of opinions. Sentiment scores 1 if possitive, 0 if neutral, -1 if negative.
Each review may have many topics and we need to score a sentiment for each topic in each review.
Automatic monthly update of the analysis.
The attached file shows the structure of the data, input and output.