Wrong top intents in CLU Response

Hello Manoj Kumar,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you are working with a language understanding system and are encountering issues with intent classification for terms related to your “Project” intent.

Among many areas to solve this issue, I will suggest you should Increase Training Data and Evaluate Intent Hierarchy. https://towardsdatascience.com/data-augmentation-for-nlp-a-comprehensive-guide-7508e7f285d5 and https://medium.com/swlh/understanding-and-designing-intent-hierarchies-in-nlp-systems-f6d9a5a8f31e

Also, to review more ways to solve the above, the links below might point you to the right solution:

• https://www.datarobot.com/blog/best-practices-for-labeling-data
• https://medium.com/swlh/how-to-handle-synonyms-and-variations-in-nlp-7e650ad54f1e
• https://towardsdatascience.com/named-entity-recognition-what-it-is-and-how-it-works-4b59662212ec
• https://www.analyticsvidhya.com/blog/2021/07/hyperparameter-tuning-for-nlp-models
• https://www.analyticsvidhya.com/blog/2020/01/how-to-test-and-validate-nlp-models
• https://towardsdatascience.com/transfer-learning-in-nlp-a-comprehensive-guide-5c8f9082d8a0
• https://video2.skills-academy.com/en-us/legal/cognitive-services/clu/clu-characteristics-and-limitations?source=recommendations
• https://video2.skills-academy.com/en-us/azure/ai-services/language-service/conversational-language-understanding/concepts/best-practices?source=recommendations

I hope this is helpful! Do not hesitate to let me know if you have any other questions.

** Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful ** so that others in the community facing similar issues can easily find the solution.

Best Regards,

Sina Salam

@Manoj Kumar I think in this case "PQ Project" or "PQR Project" is an entity rather than an intent. CLU model is best trained to infer an intent from a sentence or an utterance rather than a term or just a key word search. I can recommend to try using longer sentences as utterances to determine the actual intent, for example, in this case you can call intent as getProjectName and add utterances like "What is timeline of XYZ Project" and label XYZ Project as entity and train the model. This will train the model to get the right intent and entity and you will have better results from the model and the service. I hope this helps!!

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.