Linguistics-driven intent extraction and semantic variant creation automate the process. This leads to a large dataset of intents and variants and makes the process repeatable for different domains and languages in an automated form. These large training datasets or prebuilt bot templates allow end users to booststrap the process of setting a new bot. Bitext linguistics abstraction results in compact models, deployable at the edge or in a consumer device, for different domains and languages.
0 Comments