Key concepts

Updated at:

This topic defines the key concepts of Intelligent Conversation Analysis.

General concepts

Concept

Description

Rule / general quality inspection rule

A rule is an expression composed of logical operators and conditions. A single rule can contain a logical expression with multiple conditions.

Condition

A condition consists of a scope and an operator. For example, in the condition "the agent's first sentence contains 'Hello'", the scope is "the agent's first sentence", and the operator is "contains the keyword 'Hello'". A condition can contain only one operator.

Operator

An operator evaluates each sentence within a defined scope. Examples include keyword matching, regular expressions, semantic matching, and sentiment detection.

Quality inspection

A quality inspection is the process of analyzing audio or text files based on specified rules.

Review

A review is the manual verification of quality inspection results.

Hit

A hit occurs when parts of a conversation match the conditions defined in a rule.

Appeal

If an agent believes the score or rule hits for a recorded call are incorrect, they can appeal the quality inspection result.

Dataset

A dataset is a collection of one or more recording files.

Call center

The call center is Cloud Connect, which you can use to import audio streams or text for quality inspection.

Large model quality inspection rule

This rule type is based on a deeply optimized large language model for conversations. It provides the quality inspection capabilities needed for intelligent upgrades of marketing and service products. To create this type of rule, simply provide a quality inspection scenario and the corresponding detection dimensions and descriptions.

Session group quality inspection rule

This rule type performs a quality inspection on multiple conversation sessions between a single agent and a customer. It is suitable for scenarios requiring context from multiple conversations for an accurate judgment.

Node

A node represents a step in a process. A process can contain multiple user-defined nodes, each of which can be configured with recognition and detection attributes.

Quality inspection task

A quality inspection task is an activity that evaluates a product or service to ensure it meets specified requirements. Quality inspection tasks are categorized as offline text quality inspection, offline voice quality inspection, real-time voice quality inspection, real-time text quality inspection, and secondary quality inspection of detected tasks.

Quality inspection dimension

In basic quality inspection analysis, data is statistically analyzed based on the following dimensions:

Total number of sessions: The total number of sessions inspected.

Average score: The total score of the selected agents or skill groups for a specific quality inspection result type, divided by the total number of sessions.

Number of completed quality inspections: The total number of successfully inspected sessions.

Quality inspection completion rate: The number of completed quality inspections divided by the total number of sessions.

Number of violations: The total number of sessions that hit point-deduction rules during quality inspection.

Violation rate: The number of violations divided by the total number of sessions.

Number of serious violations: The total number of sessions that contain serious violations.

Billing

Concept

Description

Subscription

With this billing method, you pay for resources in advance for a fixed duration.

Pay-As-You-Go

With this billing method, you pay for resources as you use them, with no upfront costs.

Resource package

When you use offline or real-time voice quality inspection, usage is first deducted from your resource package. Any usage exceeding the package quota is billed on a Pay-As-You-Go basis.

Trial edition of Intelligent Conversation Analysis

For a nominal fee, you get a 30-day trial period. During the trial, you can access all features of Intelligent Conversation Analysis and inspect up to 1,000 sessions.