Using the chroma library and the bge-large-zh-v1.5 model, when recalling certain words, completely irrelevant slices are recalled.

[tanghaichen]

Search before asking

• [X] I had searched in the issues and found no similar issues.

Operating system information

Linux

Python version information

3.10

DB-GPT version

main

Related scenes

• [ ] Chat Data
• [ ] Chat Excel
• [ ] Chat DB
• [X] Chat Knowledge
• [ ] Model Management
• [ ] Dashboard
• [ ] Plugins

Installation Information

• [ ] Installation From Source

• [ ] Docker Installation

• [ ] Docker Compose Installation

• [ ] Cluster Installation

• [ ] AutoDL Image

• [ ] Other

Device information

GPU 96G

Models information

bge-large-zh-v1.5

What happened

使用的是bge-large-zh-v1.5模型和chroma向量库，在检索某些词语的时候，召回的切片分数很高但是是和词语完全无关的。但只有某个词语是这样的，其他的绝大部分词语的召回还是比较准的。 目前文档存在pdf、csv和word，切片数量大概6000个左右。 示例： 词语：“水资源” 存在20个文档，900个切片，直接出现了水资源词语。其他文档均未出现这三个字。 但在询问水资源时，召回的切片中出现的均是与其无关的切片 目前未发现其他词语出现这个问题。

What you expected to happen

正常应该是从完全出现这个词语的切片中进行召回才是合理的。

How to reproduce

未知复现方法

Additional context

No response

Are you willing to submit PR?

• [ ] Yes I am willing to submit a PR!

[Aries-ckt]

what kind of your document type and could you show some bad cases for us?