Introducing the concept of enhancing ASR (Automatic Speech Recognition) system accuracy and efficiency using data type hints.

​

Hints are used to bias the ASR system towards predicting transcriptions within specific contexts.

Adding Data Type Hints for Voice Bots

• When gathering specific details (e.g., mobile numbers), include data type hints for the ASR system.

  • Use a hint such as 'numeric' to prioritize the recognition of numeric values.

  • Example: Improve extraction accuracy by prompting customers for their mobile number with the 'numeric' hint.

    ​

​

Enhancing Keyword Queries with Language and Entity Values

• Deploy language and entity values to augment the voice bot's comprehension of keyword queries.

  • Create specific entity values for various card types (e.g., Debit Cards, Visa Cards, Master Cards).

  • Example: Customize entity values to accurately interpret user requests, such as identifying the intent to block a debit card when referencing a "Visa Card."

​

​

Optimizing ASR System for Targeted Keywords

• Enhance the ASR system's ability to recognize specific keywords or phrases.

  • Provide relevant keywords to the ASR system to enhance accuracy in recognizing them.

  • Example: Use keywords like "previous month" or "current month" to ensure precise extraction of the desired time frame.

Adding Additional Hints

• Tailor hints according to specific use cases within the Voicebot.

• Contact the Voicebot Product Team to discuss requirements, and enable hints by emailing tickets@sprinklr.com.

  • Examples:

    • Numeric_5 (Hint for recognizing numeric values)

    • City (Hint for identifying city names)