This repository contains translations of a 150 000 randomly selected entries from Amazon dataset originally created by Julian McAuley and Jianmo Ni, containing over 20GB of data. The goal is to create smaller datasets for sentiment analysis on languages other than English, for which there are many publicly available datasets already. Translation is performed with Microsoft Azure’s Translator cloud service.
Original text values have been sanitized from double quotes characters. English original with 150k entries is in aws_sentiment_input.zip. Column sentiment_score indicates review attitude for particular item on a scale from negative (1) to positive (5). Two columns that can be used for model training are summary and reviewtext. Original column overall is renamed to sentiment_score. In order to join translations with original, you can use original_id column in translation dataset. Data is probably skewed in terms of Amazon’s shopping category, for which there is no indicator, but it is not skewed regarding sentiment score distribution. Each sentiment category on a scale from one to five contains slightly more than 30k entries.
German translation (parquet file)
Spanish translation (parquet file)
Croatian translation (parquet file)
Bengali translation (parquet file)
Dutch translation (parquet file)
Italian translation (parquet file)
Norwegian translation (parquet file)
Arabic translation (parquet file)
French translation (parquet file)
Sentiment score distribution view from PowerBI:
Review date from PowerBI:
