Building an NLP Engine to Identify, Analyze, and Recommend Actions to Psychological Violence

Violetta seeks to positively impact the population that experiences a situation of violence to a lesser degree, which is why it is often not identified as violence. Violetta provides information in an empathetic, friendly, and unbiased manner to promote prevention and early recognition of patterns of violence.

The impact of the platform also transcends the gender issue, since “Violetta” is aimed at women and men between 18 and 45 years of age who use social networks to interact and who seek some kind of support when they detect something unusual or some unpleasant behavior on the part of the people with whom they live, either in their family, partner and work nuclei.

The project outcomes

Help to develop a Multi-Label Text categorization model to identify psychological violence within a text sent by the user through a conversational platform. Next, the models give recommendations based on the specific category of violence. 

Outcomes:

• A model trained with the given data and external violence data with accuracy near to 93% 
• Categorize the text inputs into different categories.
• Give recommendations to the user with the percentage of risk of psychological violence based on text classification and the model. 
• Develop a dashboard to analyze the data. The dashboard measures the sentiment and shows the topics that are being talked about to understand trends in real-time.

Further: The NLP model will be integrated into a chatbot.