word2vec for text similarities

Presenter(s): Damian Murezzan (SRF)

Being able to finding articles that belong to the same topics is very useful for content-based recommendations, but good metadata is needed to enable this. Determining the similarity between articles is an alternative way to do this without labels or categories. SRF has developed an API service which allows the retrieval of similar articles based on word2vec encodings. This talk explains the motivation to use such a service and how SRF implemented it in their systems.