On 22nd of March, Meltwater’s Stockholm NLP Meetup hosted a session about the technological development in a startup company called Kickborn. In this meetup Harinder Bedi, the founder of Kickborn, described a cognitive information retrieval system that understands the intent of questions and provides answers to them.
The algorithm of the Kickborm project uses NLP technologies like tokenizer, sentence splitter, part-of-speech tagger to compute basic feature set for intent detection from user defined text corpus. These features are then stored in a graph-like database, developed in-house, to obtain precise answers for a given question. Kickborn also has the capability to identify named entities and coreference which helps it to deliver a robust inference. The whole system is still in its earlier phase of development and does not provide extensive benchmarking. Currently it only supports English language. It also has other side products like document summarisation and topic based text understanding, etc.
The slide deck of this talk is published here.
We have lots of interesting talks upcoming in our NLP meetup event. Stay tuned :)