Natural Language Queries to Wikidata: A Naïve Prototype
|SMWCon Fall 2023|
|Natural Language Queries to Wikidata: A Naïve Prototype|
|Description:||In this lightning talk, we'll delve into a prototype I've been tinkering with, which is all about enabling more intuitive interactions with Wikidata through natural language queries.|
|Audience:||Everyone, Developers, Community|
|Event start:||2023/12/13 15:30:00|
|Keywords:||llm, prototype, rag, wikidata|
In this lightning talk, I'll be showcasing a prototype focused on enhancing user interaction with Wikidata through natural language processing. This prototype's aim is simple: to enable users to pose questions in a conversational manner and to receive responses backed by and directly linking to Wikidata statements.
The system uses a retrieval-augmented generation method, where the it initially retrieves data from Wikidata in textual format. This data is then assessed for its relevance to the user's query, and integrated into a large language model prompt. Currently, the project is in its experimental phase.
During the presentation, I'll demonstrate some code snippets, provide a basic walkthrough of the prototype, and discuss both its potential and the limitations I've faced with this approach. I'm eager to share ideas and receive feedback on this initial version of the tool.