You provision a LUIS service in Azure and use Machine Learning features to identify new items for a concept or a distinguishing trait.
For example add new medical terms to a subject domain.
What are the two types of features that are supported as part of schema design?
Click on the arrows to vote for the correct answer
A. B. C. D.Correct Answers: B and C.
Option A is incorrect because patterns are used to help improve prediction accuracy.
You can reduce the number of utterances while improving the overall accuracy.
Example of pattern utterance is “Who does {EmployeeListEntity} report to[?]”
Option B is correct because the phrase list as a feature is supported.
They act as domain specific vocabulary and enhance the quality of intents and entities.
Examples of phrase lists are Airport Codes and Medical terms.
Option C is correct because Model (intent and entity) as a feature is supported.
You can add an entity as a feature if that entity is significant for the intent.
Likewise you can add an entity as a feature to another entity if it enhances prediction of that entity.
Option D is incorrect because Prediction is not a supported feature.
Reference:
To learn more about use of features in a LUIS application, use the link given below:
In the context of LUIS service in Azure, there are two types of features that can be used as part of schema design to identify new items for a concept or distinguishing trait:
Patterns: Patterns are a type of regular expression that can be used to identify and extract entities from user input. Patterns can be created for a specific concept or a set of related concepts. For example, in the medical domain, patterns can be created to identify symptoms, diagnoses, and treatments. Once a pattern is defined, LUIS can use it to recognize entities in user input.
Phrase lists: Phrase lists are a collection of words or phrases that are related to a specific concept or entity. These lists can be used to improve the accuracy of the LUIS model by providing additional context to help identify the correct entity. For example, in the medical domain, a phrase list can be created for common medical terms and medications. When a user enters a medical term that is part of the phrase list, LUIS can use it to identify the correct entity.
In addition to patterns and phrase lists, LUIS also supports machine learning models to identify and classify entities. These models are trained using labeled data to identify patterns and relationships in user input. Once the model is trained, it can be used to recognize entities in new input.
Finally, prediction is not a type of feature that is supported as part of schema design in LUIS. Prediction refers to the process of predicting the intent or outcome of user input based on the trained LUIS model.