Regular customer service system
Our common customer service systems basically preset questions and answers, and then use matching rules to locate customer inputto the preset answersOutput.
There are generally two types of rule matching processes:
- Guided style:Fix the question options and let the user choose the fixed questions to get the answers they need. This scenario is generally used in help centers, or scenarios where the knowledge base is very serious and the answers are clear (Insurance recommendations, loan assistance, medical guidance, etc.)
- Pseudo-intelligent mode: Get answers by asking questions in natural language. In fact, it is essentially speech-to-text conversion, and then locates the preset answers through the rule engine. It is just replaced by the rule engine. User's manual selection problem.
The business process is as follows:
The green marked part in the above process is the part that can be optimized.
Intelligent customer service system based on ChatGpt
A few different key points:
-
User questions are optimized through the ChatGpt-Gpt model and searched through ChatGpt-Embeddings, which is more accurate;
-
The establishment of the knowledge base is to convert the conventional knowledge base into vector data and store it in the vector database through ChatGpt-Embeddings, making the answer positioning more intelligent and the establishment of the knowledge base more convenient. For example, it is no longer necessary to use a standard format to create a knowledge base. The input can be text in any format. Greatly improves the establishment of knowledge base;
-
The output is optimized through ChatGpt-Gpt4, which is more intelligent (more personal) and the experience is improved to a higher level.
The green part above needs to call the chatgpt service.