To this end, a method was developed, and a specific application was designed, aimed at processing and enriching the Insights with information about the keywords and themes identified in the different text fields (Input, Implications and Actions) in order to classify them and define the Sentiment Score.
The project involves:
- Extraction and normalization of textual content
- Implementation of a Data Warehouse for the historicization of the processing performed
- Implementation of configuration tables for the management of macro categorizations
- Implementations of PowerBi-based dashboards for analyzing processed information
- Integration of the PowerBi dashboard with the historicized database
Using AI, NLP and Computer Vision services from Google Cloud Platform and Azure, the solution can identify specific metadata for each medical domain, even with limited amounts of data, and define Topics and Sentiment Scores of each input.
The categorization leverages NLP text analytics, combining Artificial Intelligence practices and human decomposition methods: in fact, this approach combines Machine Learning (starting from a generalist Google model) and “human driven” analysis (based on specific contextual expertise) that allows to define and implement a specific ontology (Dictionary Builder).
Once the algorithmic model is properly configured and trained, all input is processed automatically, categorizing it by Topics and attributing its Sentiment Score. The result is then displayed on the Business Intelligence and Data Visualization tool (Control Room).