The BAALL Ontology – Configuration of Service Robots, Food, and Diet Bernd Krieg-Brückner1,2 , Serge Autexier1 and Mihai Pomarlan2 1 German Research Center for Artificial Intelligence, DFKI, BAALL, Bremen, Germany 2 Collaborative Research Center EASE, Universität Bremen, Germany Abstract The BAALL Ontology, originally motivated by Ambient Assisted Living, now comprises more than 40k OWL axioms to integrate diverse applications, covering a foundational, a variety of general, and several application domain ontologies for configuration of service robots, diets, structured food products and dishes, and cooking assistance. To maintain structural consistency, safe ontology extension is supported by Generic Ontology Design Patterns. Keywords service robots, food, diet, generic ontology design patterns, qualitatively graded relations 1. Introduction The BAALL Ontology has been developed at DFKI’s Bremen Ambient Assisted Living Lab, BAALL, the original work on configuration of mobility assistants (cf. Sect. 3) being motivated by AAL. Its domain has since been extended to cover configuration of diets (cf. Sect. 4) and robotics, food products and dishes (cf. Sect. 5), cooking assistance (cf. Sect. 6), and support for service robots (cf. Sect. 7)1 . It now comprises more than 40k OWL axioms; thus structuring (cf. Sect. 2) and safe extension (cf. Sect. 8) are essential. 2. Structure, Foundation DUL and DULsineA. Consider Fig. 1: the BAALL “hyper-ontology” is modularised into separate ontologies with import relations. Via the module Foundation, compatibility with DUL is maintained. DUL, DOLCE Ultra Light (see www.ontologydesignpatterns.org/ont/dul/) is based on the “Upper Ontology” DOLCE, providing universal concepts as a common ground. DUL is not used directly, but abstracted to DULsineA (“without axioms”), a version of DUL stripped of most relations and axioms, but keeping the essential structuring categories. New FOIS 2021 Ontology Showcase, held at FOIS 2021 - 12th International Conference on Formal Ontology in Information Systems, September 13-17, 2021, Bolzano, Italy " Bernd.Krieg-Brueckner@dfki.de, bkb@uni-bremen.de (B. Krieg-Brückner); Serge.Autexier@dfki.de (S. Autexier); Mihai.Pomarlan@uni-bremen.de (M. Pomarlan) © 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR CEUR Workshop Proceedings (CEUR-WS.org) Workshop Proceedings http://ceur-ws.org ISSN 1613-0073 1 Partial support has come from the German Research Foundation (DFG), as part of the Collaborative Research Center 1320 “EASE - Everyday Activity Science and Engineering” (http://www.ease-crc.org/) DULSINEA FOUNDATION DUL • Universal Concepts (DOLCE Ultra Light) • “Upper Ontology” without Axioms MEDICINE BIOLOGY CHEMISTRY PHYSICS OBJECT • Disease • Eukaryota • ChemicalCompound • PhysicalQuantity • PhysicalAspect • RestrictedDiet • hasIngredient • comprises... FLAVOUR ECONOMY GEOGRAPHY ARTIFACT • FlavourCompound • Price • GeoRegion • PhysicalContainer • CookWare PERSON • hasDisease • isAllowedDish FOD_SOURCE • fromEukaryota FOD_INGREDIENT • hasFlavour- • hasSource FOD_PRODUCT Compound • hasFODIngredient DISH • hasLowValue • hasHighValue DIET • DietRestriction • allowsFOD • allowsDish SHOPPING COOKING • ShoppingList • RecipeTask • processFOD Figure 1: Structure of the BAALL Ontology, with import relations categories, characteristic relations, and axioms are only introduced by Generic Ontology Design Patterns when required, cf. Sect. 8. 3. Configuration of Mobility Assistants The BAALL Ontology was motivated by the project Assistants for Safe Mobility (ASSAM, www. assam-project.eu) [1], where multifaceted variants of smart mobility assistants have been developed on the basis of wheelchairs and walkers. These are intended to compensate for individual age-related impairments: end-users may be afflicted by diseases leading to motoric impairments (e.g. loss of endurance, visibility) or cognitive impairments (e.g. disorientation). With appropriate hardware/software components, the mobility devices not only compensate for motoric impairments, but also provide orientation and navigation assistance. This variety is a considerable challenge for assessment and configuration. About 25 separate interlinked general domain ontologies have been developed (the layer below Foundation, coloured red in Fig. 1) such as Medicine defining Disease, or Artifact defining MobilityDevice as a subclass of Device. Assistance: Intermediate Abstraction Person hasDisease Disease affects hasImpairment Impairment hasAbility requiresAssistance impliesAbility needsMobilityDevice compensatedBy providedBy Ability Assistance Device Figure 2: Composite• configuration for Person needsMobilityDevice Ability compensatedBy Assistance MobilityDevice • Assistance providedBy Device Ontology Development Methodology – KSEM 2016 16 hasDisease_1Slight Person hasDisease_2Moderate hasDisease_3Severe isAllowedDish_recommended Disease isAllowedDish_1discouraged isAllowedDish_2denied hasDietRestriction_1Minor isAllowedDish_3forbidden hasDietRestriction_2Strict requiresDiet isAllowedFOD_recommended isAllowedFOD_1discouraged… RestrictedDiet FOD Dish allowsFOD_recommended comprisesConstituent _0VeryLow_or_greater_Intensity allowsFOD_1discouraged comprisesConstituent _1Low_or_greater_Intensity allowsFOD_2denied comprisesConstituent_2MediumHigh_or_greater_Intensity allowsFOD_3forbidden comprisesConstituent_3High_or_greater_Intensity comprisesConstituent_4VeryHigh_or_greater_Intensity Figure 3: Configuration of Diet comprisesConstituent_5ExtremelyHigh_or_greater_Intensity comprisesConstituent _6ExcessivelyHigh_or_greater_Intensity Composite Configuration. To manage complexity, the relation needsMobilityDevice (Fig. 2) is a stepwise composition of individual relations from Person to separately defined ontologies. Individual Diseases of a particular Person are assessed, captured by the relation hasDisease. A Disease may result in some Impairment modelled in the relation affects. In turn, an Impairment impliesAbility an Ability; an Ability may be compensatedBy some Assistance providedBy some MobilityDevice. For Disease, Impairment, etc., see [2]. Thus the relation needsMobilityDevice is sequentially composed, where each relation is a “triangular” composition, e.g. requiresAssistance ∘ providedBy → needsMobilityDevice. It is sufficient to model the individual relations affects, impliesAbility, etc., separately, and once and for all; the rest is deduced by OWL-DL reasoners. 4. Configuration of Diet: Qualitatively Graded Relations Similarly, the left-hand side of the diagram in Fig. 2 has been re-used and refined for the configuration of diets, see Fig. 3. To achieve a graded valuation according to some qualitative abstraction of a semantic concept, we introduce extra valuation domains with values such as 1Slight, 2Moderate, 3Severe, or some other (arbitrarily fine) qualitative metrics; the number of levels depends on the application. Qualitatively graded relations (cf. [2, 3, 4, 5]) encode such valuations in the names of (a sheaf of) relations hasDisease_1Slight, hasDisease_2Moderate, etc. 5. Structuring Food Products and Dishes In the case of diets, the object to be configured is Food or Drink (abbreviated as FOD in Fig. 3); a Person is recommended, ..., or forbidden certain FOD, e.g. food products containing PorkMeat, or pungent dishes for Gastritis or certain allergies. To facilitate such a modelling, the ontologies containing FOD and Dish are structured in an onion-like fashion (Fig. 1): FOD_Source is the core, where e.g. PorkMeat is characterised as isCutOf some Pig and fromAnimal SusScrofaDomesticus in the Biology ontology. Similarly, FOD_Ingredient relates to FOD_Source only by the relation hasSource, and FOD_Product to FOD_Ingredient by hasFODIngredient. A particular PungentDish can now be defined by comprisesConstituent_2MediumHigh_or_greater_Intensity some PungentFlavourCompound, etc. (Fig. 3). The BAALL Ontology is categorized on the basis of Eurocode-2 (EuroFIR), with constituent relations to Aromas, the usual NutritionCompounds, and FoodAdditives (according to the International Numbering System, INS). 6. Cooking Assistance Early work on cooking assistance [3] has led to comprehensive modelling of Ingredients (cf. Sect. 5). A recipe is a workflow of RecipeTasks such as preparation, baking, etc. Cooking is modelled based on Ingredients with e.g. flavour properties, while RecipeTasks modify their presence, e.g. when an aroma evaporates at a particular temperature. We hope to be able to eventually predict a flavour composition in the final dish by “virtual cooking”. 7. Organizing Knowledge for Robotic Activities The device configuration from Sect. 3 has later been refined by qualitatively graded configuration (Sect. 4) and become the basis for configuration in robotics, sharing user and device profile modelling, cf. [4]. The structure of an episodic memory has been formalised for domestic service robots per- forming everyday activities (see [5]) by an ontology engineering effort undertaken by the EASE1 project, resulting in a set of ontologies referred to as SOMA [6]. The focus of SOMA is autonomous and cognitive robotics, a still emerging field, which differentiates it from the more industrially focused effort at standardizing robotic knowledge representation undertaken by the ORA group [7]. Researchers involved with the development of SOMA have built upon knowledge and patterns formalized in the BAALL Ontology. An important use of episodic memories is to provide training data for a robot to improve its performance—either by learning from its own activities, or from observing (records of) the activities of other agents, e.g. humans demonstrating how a task should be performed. An episodic memory must be organized so that it will contain at least some useful knowledge for an observing agent, even if the recording agent is different in body and capabilities. Therefore it contains views on an activity from different levels of abstraction, from recordings of sensor inputs and control signals, close to the hardware, to more cognitively motivated sequences of tasks in which entities in the world play roles while obeying restrictions pertaining to the task, or placed on the entities by their own dispositions and/or capabilities. At this cognitive level the rich ontology of RecipeTasks and Cookware from the BAALL Ontology is put to use in EASE’s episodic memories. Gathering a rich variety of episodic memories requires an interdisciplinary collaboration among roboticists, cognitive scientists, and neurologists in EASE1 . This places further critical requirements on the knowledge infrastructure that processes episodic memories. In particular, there must be ways to verify that episodic memories entered into storage obey logical constraints on what is a well-formed episode recording; also important is to check whether an episode recording represents a “good run” of an activity – episodes of failure are an interesting source of learning, as long as failure is recognized as such [5]. 8. Safe Extension Using Generic Ontology Design Patterns Episodes and the other patterns above have been defined by Generic Ontology Design Patterns, GODPs [2, 4, 5, 8], expressed in the language Generic DOL [9], an extension of the Distributed Ontology, Model and Specification Language, an OMG standard (see omg.org/spec/DOL, dol-omg. org), supported by the Heterogeneous Tool Set, Hets [10]. GODPs structure complex ontologies, with local confinement of design choices, ensuring safety for modelling and data with sanity checks on their input and generative internal consis- tency for their interrelation. Thus they are very useful for maintaining structural consistency in extensions of large ontologies, even by non-experts. We may e.g. define a pattern to extend FOD_Source with a new fish while relating it to its counterpart in the Biology ontology; this anchor is important for correct translations to common names in other natural languages, a likely source of error. 9. Conclusion The BAALL Ontology is primarily used for semantic integration. Depending on the appli- cation, several application domain ontologies provide restricted “views” to the structured hyper-ontology (cf. Fig. 1). As examples consider Person importing Assistance and Artifact for MobilityDevice on the one hand (cf. Sect. 3), or Person importing Medicine and Diet on the other (cf. Sect. 4). Assistance for mobility devices, FOD, Diet, RecipeTask and Cookware provide a fair coverage for extensive academic use; completeness for commercial use is not claimed. The structured BAALL Ontology and some examples of application domain ontologies derived from it can be found at http://ontologies.baall.de, freely available to the academic community, with a dual licence for commercial use. It is presently being re-constructed by applying GODPs in a systematic fashion (cf. Sect. 8). GODPs also support populating the BAALL Ontology by data in a safe way (cf. Sect. 8 and [5]), e.g. for robot or human activity recordings (cf. Sect. 7), or for specific food products and dishes (cf. Sect. 5). Acknowledgments We are grateful for their early contributions to Sidoine Ghomsi (aromas), Martin Rink [3], Philipp Kolloge (diet), Dmytro Kozha (shopping assistance), and Marc Robin Nolte (bakery); to Mihai Codescu and Till Mossakowski (Generic DOL); José deGea Fernandez (configuration in robotics); and Daniel Beßler, Robert Porzel (episodes). References [1] B. Krieg-Brückner, C. Mandel, C. Budelmann, A. Martinez, Indoor and Outdoor Mobility Assistance, in: Ambient Assisted Living - Advanced Technologies and Societal Change, volume 2, Springer Verlag, 2014, pp. 33–52. [2] B. Krieg-Brückner, Generic Ontology Design Patterns: Qualitatively Graded Configuration, in: F. Lehner, N. Fteimi (Eds.), KSEM 2016, The 9th International Conference on Knowl- edge Science, Engineering and Management, volume 9983 of Lecture Notes in Artificial Intelligence, Springer International Publishing, 2016, pp. 580–595. [3] B. Krieg-Brückner, S. Autexier, M. Rink, S. G. Nokam, Formal Modelling for Cooking Assistance, Essays Dedicated to Martin Wirsing, in: R. D. Nicola, R. Hennicker (Eds.), Software, Services and Systems, Springer International, 2015, pp. 355–376. [4] B. Krieg-Brückner, M. Codescu, Deducing Qualitative Capabilities with Generic Ontology Design Patterns, in: M. F. Silva, J. Lima, L. Reis, A. Sanfeliu, D. Tardioli (Eds.), Robot 2019, number 1092 in AISC, Springer Nature, 2020, pp. 391–403. [5] B. Krieg-Brückner, M. Codescu, M. Pomarlan, Modelling Episodes with Generic Ontology Design Patterns, in: K. Hammar, O. Kutz, A. Dimou, T. Hahmann, R. Hoehndorf, C. Masolo, R. Vita (Eds.), JOWO 2020, The Joint Ontology Workshops, volume 2708 of CEUR Workshop Proceedings, 2020. [6] D. Beßler, R. Porzel, M. Pomarlan, A. Vyas, S. Höffner, M. Beetz, R. Malaka, J. Bateman, Foundations of the Socio-physical Model of Activities (SOMA) for Autonomous Robotic Agents, in: B. Brodaric, F. Neuhaus (Eds.), Formal Ontology in Information Systems - Proceedings of FOIS 2021 (Bozen-Bolzano, Italy), Frontiers in Artificial Intelligence and Applications, IOS Press, 2021. [7] E. Prestes, S. Fiorini, J. Carbonera, Core Ontology for Robotics and Automation, in: Standardized Knowledge Representation and Ontologies for Robotics and Automation, 2014. [8] B. Krieg-Brückner, T. Mossakowski, M. Codescu, Generic Ontology Design Patterns: Roles and Change Over Time, in: Advances in Pattern-based Ontology Engineering, volume 51 of Studies on the Semantic Web, IOS Press, Amsterdam, 2021. [9] M. Codescu, B. Krieg-Brückner, T. Mossakowski, Extensions of Generic DOL for Generic Ontology Design Patterns, in: A. Barton, S. Seppälä, D. Porello (Eds.), Proceedings of the Joint Ontology Workshops 2017, CEUR-WS.org, 2019. [10] T. Mossakowski, C. Maeder, K. Lüttich, The Heterogeneous Tool Set, Hets, in: O. Grumberg, M. Huth (Eds.), TACAS 2007, volume 4424 of LNCS, Springer, 2007, pp. 519–522.