docs/how_to/graph_constructing/

[giscus[bot]]

docs/how_to/graph_constructing/

In this guide we'll go over the basic ways of constructing a knowledge graph based on unstructured text. The constructured graph can then be used as knowledge base in a RAG application.

https://python.langchain.com/docs/how_to/graph_constructing/

[chocwaffles]

Is there a drop in replacement of ChatOpenAI with a local LLM model? I have tried chatollama, chatllamacpp, and they don't quite work as well as the soon to be deprecated ollamafunction.

[chocwaffles]

I have issues using chatollama to replace ollamafunctions as below
Swapped

llm = OllamaFunctions(model="mistral:7b-instruct-v0.3-q8_0", temperature=0, format="json")

for

llm = ChatOllama(model="llama3.1:8b-instruct-q5_K_M", temperature=0)

to perform

from langchain_experimental.graph_transformers import LLMGraphTransformer
llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(documents)

What am i missing? thanks.

[Lovi07]

What is the error you are getting?

[alibama]

is this still an issue? i'm interested in testing

[ladislav-urban]

We have tested if LLMGraphTransformer is able to construct a Knowledge Graph of quality comparable to WikiData on 1200 random Wikipedia pages. These are our results:
The node types are of good quality and expected names such as ‘Person’, ‘Organization’, ‘Document’, ‘Event’. The canonicalization process would eliminate types with the same meaning such as ‘Location’ and ‘Place’ or ‘Literary_work’ and ‘Book’.

The querying process needs to be aware of hierarchy of types such as ‘Organization’ -> ‘Educational_institution’. There are also nodes that end up without any type and are marked as ‘Entity’ . The post processing needs to look at the nodes that are of this type and categorize them into suitable types.

The node types do have exponential count distribution with a long tail of node types that do have 1 or few nodes. The node type ‘Document’ includes references to source data with relation type ‘mentions’. The nodes such as person do have attribute ‘id’ with a person's name as well as optional attributes such as ‘role’ .

If you are interested in more details you can see it at: https://www.dynocortex.com/news-and-blog/good-kg-created-by-ai/

[alibama]

great review = any insight on running with local models?