[Question]: knowledge graph，failure to correctly identify graph relationships

[huashaoxu]

Describe your problem

原始文档，"李某首先想到了其表哥郑某（某食品有限公司法人代表），多次打电话向郑某借钱，但郑某未同意。" 解析使用的实体，是person，enterprise 图谱未识别出李某和郑某的关系，而且多了一个名为”LI MOU“的人出来

chunk结果 { "directed": false, "multigraph": false, "graph": {}, "nodes": [ { "entity_type": "人", "description": "李某是故事中的主要人物，他想到向他的表哥借钱。", "source_id": "0", "weight": 1, "rank": 0, "id": "李某" }, { "entity_type": "人", "description": "郑某是李某的表哥，他是某食品有限公司的法人代表，并且拒绝了李某的借款请求。", "source_id": "0", "weight": 1, "rank": 0, "id": "郑某" }, { "entity_type": "企业", "description": "某食品有限公司是一家企业，由郑某担任法人代表。", "source_id": "0", "weight": 1, "rank": 0, "id": "某食品有限公司" } ], "links": [] }

{ "id": "李某借款事件分析", "children": [ { "id": "1. 人物关系", "children": [ { "id": "1.1 主要人物", "children": [ { "id": "1.1.1 李某", "children": [] }, { "id": "1.1.2 郑某", "children": [] } ] } ] }, { "id": "2. 行为动作", "children": [ { "id": "2.1 借款行为", "children": [ { "id": "2.1.1 联系方式", "children": [] }, { "id": "2.1.2 借款请求", "children": [] }, { "id": "2.1.3 反应结果", "children": [] } ] } ] }, { "id": "3. 结果与影响", "children": [ { "id": "3.1 郑某的反应", "children": [ { "id": "3.1.1 拒绝原因", "children": [] } ] } ] } ] }

只有”实体“没有”关系“

[KevinHuSh]

That depends on the NLP capability of LLM.

[fengyunweia]

That depends on the NLP capability of LLM.

graph chat模型使用 qwen2-vl-instruct 回复的结构不包括relationship_strength ，没有动提示词，是为什么，graph默认的chat模型是什么啊

[fengyunweia]

That depends on the NLP capability of LLM.

graph chat模型使用 qwen2-vl-instruct 回复的结构不包括relationship_strength ，没有动提示词，是为什么，graph默认的chat模型是什么啊

看了源码发现是根据租户来的，但是我使用的这个千问的为什么回复结构不包括relationship_strength 呢