This document describes the basic workflow for performing quality inspection on a dataset in Conversation Intelligence: configuring quality inspection rules, creating and publishing a quality inspection scheme, and creating a quality inspection task.
Step 1: Configure quality inspection rules
Create a rule: On the quality inspection rule configuration page, click Create Rule to add a rule.
Standard quality inspection rules: The three main types are text check, speech check, and model check.
Text check: Common options include keyword, text similarity, and regular expression.
Speech check: Common options include speech rate check and energy detection.
Model check: Easily configure and deploy an agent model or customer model with a single click.
Large model quality inspection rules: Powered by the highly optimized Lingque conversational large model, these rules provide advanced quality inspection capabilities for upgrading marketing and service products.
For more information, see Create a quality inspection rule.
Step 2: Quality inspection scheme
In the left-side navigation pane, select Rules and Schemes to go to the Scheme Management page, where you can create a quality inspection scheme.
Create a quality inspection scheme: At the top of the quality inspection scheme list, click Create Scheme and select either Create custom scheme or Select preset scheme.
Custom scheme: Allows you to add your own quality inspection rules.
Preset scheme: Provides templates for six industries, with general-purpose templates available for other scenarios.
Associate quality inspection rules: On the right, click Associate Rules to associate the quality inspection rules that you previously created.
Publish the quality inspection scheme: At the bottom of the page, click Publish. The scheme goes live and can be used in quality inspection tasks.
For more information, see Create a quality inspection scheme.
Step 3: Quality inspection task
This step shows how to create a quality inspection task from a dataset. You must first create a dataset and then create the task.
1. Create a dataset
In the left-side navigation pane, choose Quality inspection task > Dataset task > Dataset management.
Create a dataset: In the upper-left corner of the dataset list, click Create dataset to open a pop-up window where you can upload your dataset.
Dataset Name: Enter a custom name.
Click Start audio transcription. Verify that the roles are assigned correctly and then click Verify. If the assignments are correct, click Save at the bottom to create the dataset.
2. Create a quality inspection task
In the left-side navigation pane, choose Quality inspection task > Dataset task > Task management.
Create a task: In the upper-right corner, click Create task.
This is a three-step process. In the first two steps, select the dataset and quality inspection scheme you created.
Step 3: For the detection configuration, select the general-purpose model for Mandarin and click Start Inspection.
For more information, see Task management.