Projects Data Tools

GraphInsight

Multimedia summarisation using graph ranking over sentences instead of a generative model.

The problem

Generative summarisation invents. For source material where fidelity matters more than fluency, that is the wrong trade.

What it does

Sentences become nodes in a similarity graph, ranked with TextRank, and the highest-ranked are extracted. Every sentence in the summary appears verbatim in the source.

The part that mattered

Extraction over generation. It cannot hallucinate, because it never writes anything new.

Other Projects

Built With Care And A Lot Of Code

WhatsApp