Section: New Results

Around the Taaable research project

Participants : Julien Cojan, Valmi Dufour-Lussier, Inaki Fernandez, Emmanuelle Gaillard, Laura Infante-Blanco, Florence Le Ber, Jean Lieber, Amedeo Napoli, Emmanuel Nauer, Yannick Toussaint.

The Taaable project (http://taaable.fr ) has been originally created as a challenger of the Computer Cooking Contest (CCC, organized during the ICCBR Conference). A candidate to this contest is a system whose goal is to solve cooking problems on the basis of a recipe book (common to all candidates), where each recipe is a shallow XML document with an important plain text part. The size of the recipe book (about 1500 recipes) prevents from a manual indexing of recipes: this indexing is performed using semi-automatic techniques.

After being ranked twice second, in the 2008 and 2009 CCCs organized during the ICCBR conference, Taaable won the first price and the adaptation challenge, in 2010 (note that no contest was organized in 2011). Beyond its participation to the CCCs, the Taaable project aims at federating various research themes: case-based reasoning, information retrieval, knowledge acquisition and extraction, knowledge representation, minimal change theory, ontology engineering, semantic wikis, text-mining, etc.

The most important original features of this version are:

A module for refining the domain ontology for improving the case retrieval.

In Taaable, the retrieval of similar cases is based on a query generalization using an ontology of the cooking domain. In order to make the case retrieval more progressive and more precise, a enrichment of the domain ontology, and especially the ingredient hierarchy, has been studied and implemented [42] . The refinement process consists in inserting intermediate classes into the initial hierarchy of the system for better distinguishing classes that were initially not distinguishable. In order to introduce new classes into the initial hierarchy, the initial classes of ingredients have been characterized with additional properties. These additional properties are cooking actions that can be applied to ingredients and that have been extracted from the texts of recipes. FCA has been used on these new properties for restructuring the initial hierarchy.

A module for computing adaptation knowledge.

Adaptation knowledge discovery has been performed for better adapting cooking recipes to user constraints. This paper extends the approach proposed in 2009 [80] for extracting this kind of adaptation knowledge. The adaptation knowledge comes from the interpretation of closed itemsets whose items correspond to the ingredients that have to be removed, kept, or added. An original approach focusing on a restrictive binary context building and on a specific ranking based on the form of the closed itemsets has been proposed [47] .

Several theoretical studies have been carried out that should be applied to some future versions of Taaable:

• The representation of preparations in temporal representation formalisms [63] .

• An algorithm for adapting cases defined in the expressive description logic $\mathrm{𝒜ℒ𝒞}$ [43] , [11] .

• The study of the relations between adaptation based on belief revision and other approaches to adaptation [61] , [11] .

• The study of the extension of the domain ontology to make the retrieval step of a case-based reasoning system more accurate [42] .

• The study of adaptation knowledge discovery based on variation of ingredients between pairs of recipes [42] .