FAQ

Updated at:

Glossary

1 What are Q&A interfaces and management interfaces?

  • During the public preview, the Q&A interface refers to the following:

    • Real-time Q&A (RecognizeMessage)

    • Test window (RecognizeMessageForTest)

  • During the public preview, management interfaces include the following:

    • Bot management: Create, update, and delete bots.

    • Industry package management: Query package, version, and scenario lists. Subscribe to, unsubscribe from, or upgrade packages and versions. Query subscription relationships.

    • Knowledge category tree management: Query category trees.

    • Knowledge management: Query scenario lists and details. Enable or disable scenarios. Create, retrieve, update, and delete answers.

    • Dialogue data management: Human-to-human and human-machine dialogue data backflow. Labeled data backflow.

    • Other APIs.

2 What are industry packages, scenarios, and industry package versions?

  • Industry package: retailbot uses Alibaba's E-commerce big data to develop hundreds of general scenarios and thousands of industry-specific scenarios for nearly 60 sub-industries. These scenarios are categorized into one general package and multiple industry packages. In addition to high-frequency scenarios, a knowledge package also contains Q&A models and dialogue policies that are designed and optimized for the characteristics of a specific industry. This improves dialogue performance for that industry.

  • Scenario: A collection of similar questions.

  • Industry package version: The built-in industry packages are continuously iterated and upgraded. Upgrades may include adjustments to scenarios and scenario recognition performance. A new version number is released with each upgrade. Subscribe to the latest version to obtain the best performance.

3 What are knowledge, answers, answer constraints, and knowledge categories?

  • Knowledge: In retailbot, "knowledge" is defined as any data, other than rule-based text, that a bot might use to respond to a consumer.

  • Answer: A response behavior to a consumer. The Minimum Viable Product (MVP) version supports knowledge base answers, transfer-to-agent answers, and rule-based text answers.

  • Answer constraint: A combination of trigger conditions for an answer. These can include factors such as the product the consumer is asking about, order status, time, and keywords.

  • Knowledge category: The scenario classification that each scenario belongs to. Developers use it for knowledge management for retail businesses on the Software as a Service (SaaS) side.

FAQ

1 What channels does retailbot support?

  • The public preview MVP version only supports the Taobao and Tmall channels. Before integration, developers must obtain bot qualifications that are officially authorized by Qianniu.

  • Support for multi-channel calls will be added in iterations during the public preview. This will allow developers to integrate with retail channels such as WeChat, official websites, and apps.

2 Is the product free during the public preview? Is a Service-Level Agreement (SLA) provided?

  • During the retailbot public preview, a Free Edition is available for feature evaluation and integration testing. The Free Edition has resource limits, such as Queries Per Second (QPS) and the number of bot instances. To access the full service, you must pay the corresponding fees. For more information, see the billing description.

  • A Service-Level Agreement (SLA) is provided during the public preview.

3. What are the core differences between retailbot and other PaaS-based Q&A chatbots?

  • Capability differences: retailbot provides features, algorithms, scenario-based dialogue capabilities, and data models that are designed specifically for the retail business model. This provides stronger problem-solving capabilities for the retail industry.

  • Out-of-the-box: retailbot provides developers with a semi-finished product that encapsulates multiple scenario-processing capabilities. Developers do not need a deep understanding of the business or to repeatedly build scenarios and dialogue capabilities. It works out-of-the-box and ensures good performance.

4 How do I use the built-in industry packages?

  • After you create a bot instance, you can call the relevant industry package management APIs to subscribe the bot to a specific industry package.

  • After a successful subscription:

    • When this bot instance calls knowledge-related APIs, it automatically retrieves the scenarios included in the corresponding industry package.

    • When this bot instance calls the real-time Q&A API or the test window API, the system automatically applies the Q&A models and policies included in the industry package.

  • Recommendations:

    • All bots are subscribed to the general package by default, and this subscription cannot be canceled. This increases the bot's scenario coverage.

    • Selectively subscribe to relevant industry packages based on the customer's actual business operations.

  • Note:

    • Subscribing to more industry packages is not always better. If a customer does not conduct business in a specific industry, subscribing to unnecessary industry packages may affect detection accuracy.

    • The MVP does not support editing or deleting built-in industry packages. It only supports adjusting the industry package subscription relationship for each bot.

5 What are the limits of the knowledge base?

  • During the public preview, each scenario is limited to 100 answers.

6 What is the effect of enabling or disabling a scenario?

  • Enabling or disabling a scenario only affects a specific bot instance. Other bots that have subscribed to the same industry package and have the same scenario are not affected.

  • The real-time Q&A API does not detect disabled scenarios. However, you can use input parameters to direct the test window API to detect a disabled scenario.

7 What is the difference between the real-time Q&A and test window interfaces?

  • They support different QPS limits and have different Operations and Maintenance (O&M) support policies.

  • In terms of business logic, the test window API can hit a disabled scenario, but it does not support A/B testing.