Semantika on Bloomberg TV: search that understands meaning

Summary

On Business Start, Ivan Nonchev asks whether a Bulgarian search engine can compete with the world's names in the field, and introduces Rossen Genchev as the founder of Semantika, a search engine based on artificial intelligence. Yes and no, he answers: what sets it apart is that it does not rely on the presence of keywords but on their meaning within a given segment. A standard search engine drops the noise words, cuts every word to its root and looks for the roots in every document; ask a fashion site for Gucci, Prada and shoes and the result is often nothing, because no single product mentions all three.

Semantika adds what the product does not say, which he calls information enrichment. Enter a BMW 120 and it knows the brand is German and premium, and that the car is a 1 Series, compact and made for the city; so it can answer "give me a dynamic car for city driving", a request without a single keyword in it. In fashion it takes the colours from the picture, reads the relations between them and recommends pieces that go together; told the colours of the season, it answers "give me fashionable shoes". The aim, he says, is to make software understand people rather than make people learn software.

The technology applies well beyond shops: real estate; genetics, where Semantika is in talks with a company; and search in medical records, one of its prototypes. The company is five people, its site has been live for a month, and five prospective clients are in negotiation. The team jokes that it has no competitor without $30–40 million of investment; Google too, he says, started as students, with a simple algorithm that ranked sites by their links. Bulgaria has the people for it: an IT business needs internet, not roads, taxes are low, and funds and angel groups are appearing.

Key moments

1. 00:24What sets it apart: the meaning of words within a segment, not their presence.
2. 01:18Gucci, Prada and shoes: why a keyword search so often returns nothing.
3. 02:01Information enrichment: a BMW 120 becomes a compact premium German city car.
4. 03:13Applicable in every domain; the site has been live for a month, with five prospective clients.
5. 05:49Fashion: the colours taken from the picture, and the relations between them.
6. 06:38Make software understand people, not people learn software.
7. 07:55Genetics and medical records: where the next applications are.
8. 10:30Google started small too, with a simple algorithm that ranked sites by their links.
9. 12:08Five people, and no competitor without $30–40 million of investment.
10. 13:23Bulgaria: an IT business needs internet, not roads, and taxes are low.