consuming operation. In summary, the above issues can be attributed to the task and model mismatch SQL is a general-purpose query language, which makes it a less intuitive and a less useful tool for users in the vertical" application domain of recommender systems, where SQL may not have some specialized capabilities important for recommender systems. Also, SQL is based on the relational data model, and multidimensional recommendations on the multidimensional model 4] would need to be mapped into the relational model to support SQL queries, which leads to various translation problems. To avoid these issues, it is advantageous to develop a specialized recommendation language based on the characteristics of the application domain that also supports the multidimensional recommendation model, as we do in the To provide flexible and user-driven recommendations and to address the previously specified issues with using SQL as a recommendation language, we designed a new recommendation query language REQUEST, which allows its users to express in a flexible manner a broad range of recommendations that are tailored to their own individual needs and, therefore, more accurately reflect their interests. For example, the earlier recommendation for Tom can be expressed in REQUEST as RECOMMEND Movie, Time TO User, Companion USING MoVie Recommendet RESTRICT USer Name=“Tom” AND Time. Timeofweek= weekend” AND Companion.Type=ˇ Girlfriend ASed oN Personal Rating SHOW TOP 3 where Movie Recommender is a 5-dimensional cube of ratings having dimensions User Movie. Time Companion, and Theater; also, Personal Rating represents the ratings measure for the cube The above REqueSt query is based on the olaP paradigm [9], which is a natural choice for querying multidimensional recommender systems, since the data model of REQUEST matches the multidimensional data model of the ratings cube. Besides reQuest, we also present a multidimensional REQUEST is an acronym for REcommendation QUEry STatements. The initial query language, called RQL, was introduced in an earlier workshop paper [5], where only the preliminary ideas of how to define the query language were presented. In this paper, we systematically redesigned the language by formally introducing its syntax, semantics, and the corresponding recommendation algebra. This allowed us to significantly extend capabilities of the language over its preliminary version [5]. To reflect these major changes, we renamed the language from rQl to REQUEST5 consuming operation. In summary, the above issues can be attributed to the task and model mismatch. SQL is a general-purpose query language, which makes it a less intuitive and a less useful tool for users in the “vertical” application domain of recommender systems, where SQL may not have some specialized capabilities important for recommender systems. Also, SQL is based on the relational data model, and multidimensional recommendations on the multidimensional model [4] would need to be mapped into the relational model to support SQL queries, which leads to various translation problems. To avoid these issues, it is advantageous to develop a specialized recommendation language based on the characteristics of the application domain that also supports the multidimensional recommendation model, as we do in the paper. To provide flexible and user-driven recommendations and to address the previously specified issues with using SQL as a recommendation language, we designed a new recommendation query language REQUEST, 1 which allows its users to express in a flexible manner a broad range of recommendations that are tailored to their own individual needs and, therefore, more accurately reflect their interests. For example, the earlier recommendation for Tom can be expressed in REQUEST as RECOMMEND Movie, Time TO User, Companion USING MovieRecommender RESTRICT User.Name = “Tom” AND Time.TimeOfWeek=“weekend” AND Companion.Type = “Girlfriend” BASED ON PersonalRating SHOW TOP 3 where MovieRecommender is a 5-dimensional cube of ratings having dimensions User, Movie, Time, Companion, and Theater; also, PersonalRating represents the ratings measure for the cube. The above REQUEST query is based on the OLAP paradigm [9], which is a natural choice for querying multidimensional recommender systems, since the data model of REQUEST matches the multidimensional data model of the ratings cube. Besides REQUEST, we also present a multidimensional 1 REQUEST is an acronym for REcommendation QUEry STatements. The initial version of our recommendation query language, called RQL, was introduced in an earlier workshop paper [5], where only the preliminary ideas of how to define the query language were presented. In this paper, we systematically redesigned the language by formally introducing its syntax, semantics, and the corresponding recommendation algebra. This allowed us to significantly extend capabilities of the language over its preliminary version [5]. To reflect these major changes, we renamed the language from RQL to REQUEST