Description:
Fast and scalable analysis of reviews of a company's mobile application, automation of marking these reviews into categories and harmonization of the taxonomy of these categories involves creating a solution in the format of an exploration tool, implying automatic categorization and subsequent visualization of the analysis results. The solution allows you to evaluate review sentiment and establish data trends to help improve a company's mobile app.
Context:
The company has a mobile application in the App Store. In the reviews of this application, users write their feedback, which is categorized manually. Based on this, analysts create tasks for the development department.
Solution:
a model was created for clustering and searching for new topics in the review stream to automate the work of analysts.
Results:
Integration into BI of the customer service department Automation of review categorization:
Taxonomy Harmonization:
Sentiment assessment and trend analysis:
To build the model we used:
Simulation results:
Customer: Restaurant, Retail
Technology stack: BigARTM, Python, Flask, PyTorch, nltk, gensim.
Fast and scalable analysis of reviews of a company's mobile application, automation of marking these reviews into categories and harmonization of the taxonomy of these categories involves creating a solution in the format of an exploration tool, implying automatic categorization and subsequent visualization of the analysis results. The solution allows you to evaluate review sentiment and establish data trends to help improve a company's mobile app.
Context:
The company has a mobile application in the App Store. In the reviews of this application, users write their feedback, which is categorized manually. Based on this, analysts create tasks for the development department.
Solution:
a model was created for clustering and searching for new topics in the review stream to automate the work of analysts.
Results:
Integration into BI of the customer service department Automation of review categorization:
- Increased accuracy by 15% (reduced human error);
- Reduce analysis time by 85%.
Taxonomy Harmonization:
- Reducing the time to identify a new category by 60%;
- Adding and merging 15% of categories.
Sentiment assessment and trend analysis:
- Identification of growth points when analyzing reviews is 40% more;
- Reducing the time to identify critical and reputation-impacting reviews by 70%.
To build the model we used:
- Reviews of the mobile application from the AppStore, Google Play, and other sources;
- Review ratings provided by the user;
- Initial taxonomy of review categories.
Simulation results:
- A model for categorizing reviews by taxonomy and predicting sentiment;
- Recommendations for taxonomy harmonization;
- Web application with aggregation of analytics based on reviews.
Customer: Restaurant, Retail
Technology stack: BigARTM, Python, Flask, PyTorch, nltk, gensim.
