Fake feature test: LLM NLP for Querying Neo4j KG using an English->Cypher Translation Layer

[Arnedeklerk]

Background

NeoDash has released an update introducing the use of Latent Linguistic Modelling (LLM) Natural Language Processing (NLP) for querying a Neo4j Knowledge Graph (KG) using an English->Cypher translation layer. This innovative feature aligns with Keywan's previous interests and differs from our GPT summary Fake Feature Test (FFT), as it functions as a translation layer rather than a summariser.

You can read more about this feature:

• Here
• Their implementation method here
• Detailed explanation here

Action Plan

Conduct a Fake Feature Test (FFT), ideally in the KnetMiner 6.0 release. This will involve integrating Google Analytics to assess the success probability. It's important to note that the numbers from this test should serve as a rough guide rather than a definitive result.

• Update Figma designs to show FFT button. Think of realistic relevant logic for where the user ends up when clicking the button.
• Bind GA4
• Monitor outcomes and report back here

After running the FFT, we can observe the outcomes and decide our next steps accordingly. The completion of this step should also provide us with useful insights that can help us advance the implementation of LLM NLP, assuming the FFT was successful.

[Arnedeklerk]

Empat: we can discuss this next week, but for now don't worry about it too much. This will be a part of the KnetMiner update for 6.0 and have little to do with KnetSpace, other than perhaps a form to submit interest which should be captured somewhere (@SvAnastasiia).

First things first, and that's for me to take a look at Figma. Likely it'd be a similar solution to #758