Defining pedagogical indexation of texts for language learning as an indexation allowing users to query for texts in order to use them in language teaching requires to take into account the influence of the properties of the teaching situation we define as " pedagogical context ". We propose to justify the notions of prisms and facets on which our model rely through the description of material selection in the task of planing a language class as an adaptation of Yinger's model of planing. This interpretation of Yinger's model is closely intertwined with the elaboration of the notion of pedagogical context. The latter provides sounder bases on which to build our model on. This resulted in improvements in the potentialities of the model compared to its first published version. The MIRTO project, started in 2001, stemmed from the observation of various recurrent issues in Computer Assisted Language Learning (CALL) systems: rigidity, inability to adapt the learning sequences to learners and unavailability of means to manipulate concepts pertaining to the teachers' field of expertise (language didactics) [1]. The aim of MIRTO was to promote the use of Natural Language Processing (NLP) to address those problems by adding an abstraction layer between the user and the material. Antoniadis et al. consider that the formulation of problems in didactics relevant terms depends on handling language not as character sequences but as a system of forms and concepts [1]. MIRTO thus proposes to separate treatments (e.g. gap-filling exercise generation script) and the data on which they are to be applied (a text in this case).