2020 IASA - FIAT/IFTA Joint Conference

Since several years RAI is digitalizing the Italian video heritage library. Rai started video broadcasting on  3rd of January 1954 and nowadays a big digital library is available as “Le teche RAI”.
Such content is available on the RAI OTT service (available on any digital device) called RaiPlay, where, for example, contents from the 60s and 70s and later are easily available.


ATOS has integrated for RaiPlay a Content Recommendation platform which examines the available content through semantic analysis algorithms and then catalogs them based on their characteristics. Each registered user is profiled in his actions such as playing videos, adding to his list, sharing on social networks, etc., and all these actions build a profile that is useful for identifying and predicting what content to propose.
The RaiPlay applications mix in the same page new contents, popular contents, archive contents and contents from different categories in different stripes, each strip is composed by editorial suggestions and profile based suggestions.

The solution implements an analytics engine that is capable to analyze different users’ attitudes toward different kind of content including heritage material from “Le Teche RAI”.
Furthermore, ATOS realized a prediction engine based on classification methods that use logistic regression algorithms implemented in Python: these prediction algorithms are used to assess in advance the future popularity of a content and understand the effectiveness of the metadata used to classify the content. Metadata accuracy is even more important when ingesting content produced several years ago.
In addition the solution implements traditional recommendation algorithms based on content affinity and is able to guide the user in finding even more heritage content, this helps RAI to give more value to historical contents and to effectively increase their cultural and economic importance.