Can I train my own aspect based sentiment analysis (opinion mining) model?
Hi, I'm considering using the sentiment analysis and opinion mining services. However in case I need to correct and fix misclassifications - I need to be able to train also my own model for aspect based sentiment analysis. I can see in the documentation that I can train sentiment analysis on sentence level, but there is no option to label aspects/targets for the opinion mining part: https://learn.microsoft.com/en-us/azure/ai-services/language-service/sentiment-opinion-mining/custom/concepts/data-formats
I need to be able to label the concept/target that the model needs to extract with its sentiment (positive/negative/neural). For example:
{ "sentence": { "id": "2846", "text": "Not only was the food outstanding, but the little 'perks' were great.", "aspectTerms": { "aspectTerm": [ { "term": "food", "polarity": "positive", "from": "17", "to": "21" }, { "term": "perks", "polarity": "positive", "from": "51", "to": "56" } ] }, "sentiment": "positive" } }
Is there a way to train a custom model like this?
Thanks!