Learn more about the different concepts of general generative AI and knowledgebase topics.
The use of generative AI for answering the incoming user requests is part of moinAI. But what is of importance when using ist? What settings can be made? Are individuel intents still needed or is one big knowledge base sufficient?
This article provides answers to the questions stated above and explains, how generative AI is implemented by moinAI on different levels.
1. General generative AI for answering user requests
The general generative AI is provided from the start. It uses the resources of the knowledge base to generate a correct answer for the individual user requests. More resources, such as URLs, PDFs, Documents, CSV-Charts or Question-Answer-Pairs can be added and enlarge the knowledgebase. This enables the artificial intelligence to answer different enquiries correctly.
Basic influence on the general generative AI is given. A greater amount of resources affects the capability to generate correct responses. Especially question-answer-pairs have an impact on priorly defined questions. If such a question or a question with similar content is recognized, an answer is played that is very similar to the predefined answer. AI communication guidelines make it possible to influence the tone of voice of the texts, the use of emojis, etc..
Going live with the standard settings is possible. It does inherit some risks though: The potential of errors is greater, the customization is limited and an detailed analysis of the performance is not possible.
2. More control options for answer content through specifications
To guarantee the optimal functioning of the Chatbot, the correct allocation of requests to the correct intents plays a key part. The differentiation of intents by specific intentions enables the optimal recognition of user requests.
We recommend the use of multiple knowledge base intents, that are differentiated by their individual intentions. This comes with the following benefits:
- Lesser room for error: By differentiating the intents and their specific resources, the probability of generating false responses gets minimized.
- Better analysis: Differentiated intents enable a specific analysis, providing valuable insights regarding the usage and performance of the intent.
- More customization: Every intent can be customized and optimized individually, which enhances the general performance of the system.
This benefits are connected to the AI-based recognition. First, the AI-Model allocates user requests to a fitting intent. This allocation is based on the specific intention of the intent. The greater the amount of intents, the more precisely is the given answer, since:
specific answers can only be provided, when the intents can be differentiated decidedly.
Knowledge base intents are the same as the general generative AI, at least from a technological standpoint. However, they are more suitable of giving precise responses, since they are specifically differentiated by defined intentions.
2.2 Redactional intents
Redactional intents are intents that provide answers without using generative AI. The answers given are static and priorly defined. These intents are still available.
At the end of this article it is explained, when the use of a redactional intent is recommended an when it makes more sense to use a knowlede base intent.