Training the Assistant to recognize additional data
Customizing Assistant with Domain-Specific Data
While Slang CONVA strives to provide the highest accuracy and coverage over recognition of your user’s utterances, there may be certain words or phrases that CONVA doesn’t recognize, because they are specific to a customer’s particular app or brand. In such cases, customers are given the opportunity to augment the in-built data available, with their own data.
Adding custom data
In the CONVA platform, larger domains, such as travel and retail, are subdivided into smaller ones, such as Bus, Flight, or Train (for travel) or Pharmacy or Grocery (for retail).
Each subdomain has specific in-built data, allowing users to fine-tune the recognition quality by selecting only the subdomains that matter to them.
For each of these subdomains, additional data can be uploaded as many comma-separated files, each of which having a specific format and functionality described below. The user has the ability to augment existing data or entirely replace CONVA’s internal data entirely with their own. If the developer chooses to augment the in-built data recognition will be based on both in-built data and user data.
Uploading subdomain data:
The upload option for each subdomain data file
To upload subdomain-specific data click on the up arrow button for each subdomain data type, and upload the corresponding CSV file.
Editing Subdomain data:
Click this button to view the subdomain data
The data viewing modal and the button to click to download the data
To edit an already uploaded csv file, download the file first using the download button, edit the file and then upload it again.
Replace/Augment data option:
The checkbox to tick to toggle between replace or augmentation options for the data
As mentioned above the data uploaded can be used to augment or replace existing CONVA domain data. To replace the data check the replace checkbox for each data file.