Creating a knowledge graph with your LLM
The construction of a KG is generally a multistep process consisting of the following:
- Knowledge creation: The first step, in which we define the purpose of this KG, is to gather the sources from which to extract knowledge. In this step, we have to decide how we build our KG but also where we maintain it. Once built, the KG has to be stored, and we have to have an efficient structure to query it.
- Knowledge assessment: In this step, we assess the quality of the KG obtained.
- Knowledge cleaning: There are several steps and procedures to make sure there are no errors and then correct them. This step can be conducted at the same time as knowledge assessment, and some pipelines conduct them together.
- Knowledge enrichment: This involves a series of steps to identify whether there are gaps in knowledge. We can also integrate additional sources (extract information from other datasets, integrate databases, or merge multiple KGs). ...