Overview

Model

• Supports customizing the models used by the intelligent analysis assistant (NL2SQL) and intelligent insights, allowing you to switch models as needed to choose the one that best fits your business scenarios.
• Continually optimizing and adjusting the used models can enhance the assistant's Q&A capabilities, making it better at understanding and responding to complex queries.

Prompt

• Prompts are user inputs to the large model, which generates corresponding responses or outputs based on these prompts.
• Configuring prompts can guide the model to understand users' specific needs and generate more accurate, relevant, and high-quality outputs, ensuring the model's responses meet the application requirements in different scenarios.

Scene

• A scene represents a Q&A topic and typically comprises multiple datasets. User questions within the same scene usually pertain to the same topic.
• By configuring descriptions and some example questions for each scene, you can help business personnel quickly understand how to find the target data in the scene.

Keyword

• Configure the actual search content for certain words and sentences in Q&A questions. For instance, set the keyword "From January to March" for "the production of apples from January to March" as "greater than or equal to January and less than or equal to March". The assistant can recognize and understand different expressions in user questions.
• Configuring keywords helps the assistant understand diverse expressions in user questions, thus improving its comprehension.

Named Entity

• Configure binding relationships between fields through named entities. For example, associate entity A with fields user_name and product_color. When querying about the user_name field on the Q&A page, it will also display the associated field product_color. For details, see Example Scenario.
• Configuring named entities ensures that specific data entities are recognized and associated during answering, providing more accurate answers.

Question Templates

• When using the intelligent analysis assistant for Q&A, the system automatically assesses the similarity between the input question and the question template. If the similarity meets the threshold, the intelligent analysis assistant returns the question template.
• By configuring question templates, you can ensure that the responses from the intelligent analysis assistant are more targeted and accurate.

Templated Insights

• Configure templated insights for key questions. During Q&A in the assistant interface, you get structured and high-quality intelligent insights as a result.
• Templated insights ensure standard and consistent answers, helping achieve structured and high-quality responses.

Synonym

• Configure synonyms for dataset fields. For example, set product_price as the synonym for product price. When querying product price, the assistant can accurately map it to the product_price field.
• This improves the readability and accuracy of data analysis by allowing the assistant to parse and understand information more precisely.

Time Limit

• When the assistant returns results, it not only answers the current question but also shows related questions and their answers.
• Configuring related questions aims to optimize the Q&A process by predicting potential follow-up questions and providing relevant information in advance.