Categorisation is about making it easier to find opportunities.
Getting it right means being better than humans.
For it be really useful, we have to be able to take source documents and, where necessary improve on the categorisation provided by the publisher.
If a publisher gives us a record that is classified as construction, that is useful, but if you’re a roofing contractor, it’s not much use. Yes, it stops you looking at opportunities for taxis, it doesn’t give you *exactly* what you want. Even if you add a filter on construction tenders with ‘roof’ as a keyword, you’ll be better off, but you’ll miss any contracts that fail to mention the word ‘roof’. So a contract to replace guttering might get missed.
We’re striving to take existing broad classifications and give them better, more detailed classifications. So that users don’t just look at construction tenders, but tenders for roofing. That means writing code that does a better job of classifying the records than the original publisher.
That’s a challenge, but one we think we are equal to.
Our new categorisation algorithm is getting some really good results and will soon go into testing with some of our customers.