By combining content-matching and collaborative filtering, LIP serves recommendations of the highest quality. It can both recommend content that is thematically relevant to a specific article (based on the text) and also from different interest areas (based on users with shared interests). In addition, it does not suffer from the cold start problem, i.e. the inability to make recommendations when there is a lack of user data, especially for new articles and new users.
A key benefit of recommendations, over traditional search is that users see content they are interested in or need and often also content they did not even know existed, without having to actively try to find it.
The core of LIP’s technology is a unique way of measuring the thematic similarity of documents. Using machine learning, our algorithm is able to automatically process millions of documents and gain an understanding of how words relate to each other. It can then apply this to document databases it has not seen before and make intelligent recommendations of similar content.
This technique looks at a user’s behaviour, e.g. what they viewed, saved and purchased, in order to determine their taste. This is then compared to the behaviour of other users, in order to make relevant, personalised content recommendations. Hence, in a very basic scenario, if user A likes 3 items and user B likes 2 of those items, the likelihood of user B liking the third item user A likes is high.