Azure AI Content Safety for Red flag comment analysis
Use Cases and Deployment Scope
We use Azure AI Content Safety in our AI Agent model for red flag comment analysis process. Mainly the comments are from surveys called open ended survey where the users of our client products will be taken survey and comments will be given by the users in text in that the words which are not good or unparlimentary words used means Azure AI Content Safety model mark them as Red flag comments and store in db that the comment is vulgar.
Pros
- Azure AI Content Safety correctly finds the unparlimentary words maximum.
- Azure AI Content Safety endpoint and api is easy to create from Azure portal and integrate with our code
- Azure AI Content Safety can do things which Azure AI Foundry model can't do
Cons
- Azure AI Content Safety model is very costly that should be reduced
- Azure AI Content Safety sometimes, very few time couldn't handle most vulger contents that should be trained to model
- Azure AI Content Safety metrics and scalability is sometime not politely visible, which should be improved
Return on Investment
- One time it cost 7000 US Dollars when we ran to test, which is very costly, then we connected with the Azure team and sorted it out and got back 4000 US Dollars. This should not happen
- Azure AI Content Safety helped to find red flagged comments for lakhs of comments for our analysis
- Azure AI Content Safety prevents those comments in our product frontend and store that as report helps us a lot.
Usability
Alternatives Considered
Azure OpenAI Service and ChatGPT
Other Software Used
Azure Blob Storage, Azure AI Document Intelligence, Azure AI Search