Unanswered

Reading one word as a synonym as another word?

Like (2) Share

Report

Posted on 11 Oct 2024 19:06:55 by TA-11101903-0

I was wondering if there was anyway for Copilot studio to read a word that has a similar meaning like reading "car" as "truck" or vice versa, so when if someone puts in a query to the agent and refers to a car as a truck, it reads it as the same word and can search through the knowledge base with an unbiased viewpoint and finds relevant documents. Just curious.

Categories:

General

All responses (2)

Answers (0)

Suggested answer

SaiRT14 1,983 Super User 2025 Season 1 on 20 Oct 2024 at 13:19:21

Like (0)

Report
Reading one word as a synonym as another word?

You can set up a custom glossary or synonym dictionary where you define that "car" and "truck" should be treated as interchangeable terms. This would allow Copilot to understand queries with either term as the same concept.

In Copilot Studio, you can configure a synonym list where certain words are treated as equivalent.

For example, you could map "car," "truck," "automobile," and "vehicle" together so the AI interprets them as the same entity.
Suggested answer

Vinoth Selvam 1,527 Super User 2025 Season 1 on 13 Oct 2024 at 11:17:49

Like (1)

Report

Reading one word as a synonym as another word?

Hi TA-11101903-0,

I think you can try some ways to make it work.

You can try use synonym expansion in the query or prompt before it reaches the knowledge base. When a user inputs a word like "truck," the system can expand the query to include synonyms like "car" or "vehicle.

"When a user mentions a 'truck,' 'car,' or 'vehicle,' treat them as the same concept."

You can also use Named Entity Recognition and help the system recognize that both "car" and "truck" belong to the entity category of "vehicles." This way, it would treat both terms under the same category during search or response formulation.

One more complex method ( I have not tried ):

Microsoft Azure Cognitive Search has a built-in semantic search capability that ranks results based on the meaning of words rather than exact matches. This ensures that when someone refers to "truck," documents with "car" could also be returned if they are relevant based on the context.

Thanks.