Tensor flow accuracy


I have a strange problem with Tensor flow embedding. I have about 200 intents and apporx 30,000 examples. While training, I am getting all entities exactly getting tagged. But, the intents are picked up where the entities do not figure at all. Any pointers, pls.

entity extraction and intent classification aren’t related to each other. 200 intents might cause some problems in classification accuracy. Try running the evaluate script on your training data to see what the major problems are