Closed robknapen closed 3 weeks ago
It's sort of last-minute change, but I can manage to do that in the architecture diagram. Can you take care of renaming in the Technical Documentation + providing the description of functionalities, that you mentioned?
Sure, if the change makes sense and is acceptable I can update it in the Technical Documentation.
If as you say, it is based on the development team discussions it makes sense to me :) But please make sure to finish the changes tomorrow at 12:00.
I don't think I can make that deadline, probably best to do it in a next iteration then.
Ok :) I am leaving it for the next iteration. We can also open this issue on Thursday during Sprint Refinement meeting.
I'll assign it to @roblokers so he can bring it up for the backlog refinement.
Please, now it is time to work on this (deadline mid-september). We will deliver updated version of Technical documentation together with the first prototype.
Currently, we have Interlinker and Metadata augmentation tech component with the following expected functionality. (Note that some of the functionality is expected in next iterations). I think we could keep them separated...
Metadata augmentation
Interlinker
I agree with renaming LLM to Natural language Querying (NLQ).
For interlinker and augmentation, I think I would keep them both and maybe reshuffle functions. For me metadata cleaning would logically fall under augmentation, as this doesn't create links between identical, similar or related items.
Also I agree with Rob that NLP/LLM support (the AI/ML part of metadata augmentation) will be part of it. Although fornow I would not add dedicated functions for it. I would rather think that AI/ML would complement, enforce or replace specific functions along the development path.
For now / after discussion in the group we end up with the following (partly renamed) components and functions:
Knowledge graph
Natural Language Querying
Metadata Augmentation
Needs to be aligned with updated tech doc and model diagram by respective authors (@roblokers @pvgenuchten @DajanaSnopkova )
Based on progression of ideas and development I would like to propose the renaming of two components:
The Metadata Augmentation component can then also include NLP/LLM based functionality for improving metadata, e.g. extracting additional keywords, ontology terms, summarising text, etc. While the NLQ component can be focused on interfacing between the ChatBot UI and the knowledge graph.
Implementation-wise the components might share vector stores with embeddings created from relevant documents or KGs, and access to an instance of a large language model. But they might be different depending on purpose and requirements.