MAPP-Reviews (Monitoring App Reviews) is software that:
-
- Extracts requirements with a negative rating from app reviews;
-
- Generates time series based on the frequency of negative reviews; and
-
- Trains predictive models to identify requirements with higher negative rating trends.
In MAPP-Reviews, the primary intention is to detect negative reviews of a software requirement that are starting to happen and make a forecast to observe if they will worsen in subsequent periods, i.e., high frequency of negative reviews.
MAPP-Reviews uses the pre-trained RE-BERT model to extract software requirements from app reviews. RE-BERT uses a cross-domain training strategy, where the model was trained in 7 apps and tested in one unknown app for the test step.
MAPP-Reviews uses the k-means algorithm to obtain a clustering model of semantically similar software requirements.
MAPP-Reviews uses the Prophet Forecasting Model. Prophet is a model from Facebook researchers for forecasting time series data considering non-linear trends at different time intervals, such as yearly, weekly, and daily seasonality.