Providing a cooking recommendation for a dish

Abstract

A method (150) for providing a cooking recommendation (110) for a dish to be cooked comprises the steps of acquiring a description (105) of the dish that is intelligible to humans; determining a first classification of the description (105) in relation to a first predetermined taxonomy (140); deriving a second classification of the dish from the first classification and in relation to a second predetermined taxonomy (145); the second taxonomy (145) describing cooking-related properties of the dish; and providing a cooking recommendation (110) for the dish.

Description

[0001] Providing a cooking recommendation for a dish

[0002] The invention relates to the control of a cooking appliance. In particular, the invention relates to the provision of a cooking recommendation for a dish.

[0003] A dish can be prepared according to a recipe, with one of the processing steps involving cooking. Cooking can take place, for example, in an oven, which requires setting the oven to predetermined parameters. These parameters can include a temperature or a heat source. Available heat sources in the oven can include, for example, top heat, bottom heat, convection, or grill. A modern oven can also provide additional heat sources, such as a microwave or a steam generator. Different heat sources can be combined, and a heat source can be activated only during part of a cooking process.For example, a roast can first be treated with a high temperature on all sides to close the pores in the meat, then with a lower temperature from below, possibly with the addition of steam to prevent the roast from drying out, and shortly before the end of cooking with the help of a grill to allow a crust to form.

[0004] A cooking recipe can be assigned parameters for the correct setting of an oven. This usually assumes an oven with only a few heat sources so that the recipe can be used universally. A better-equipped oven cannot be fully utilized in this case, and the cooking result may be good but not optimal. In another case, a familiar cooking recipe may be modified so that it is not clear how a beneficial cooking process could proceed. Empirically determining the best cooking method can be very time-consuming if all possible cooking methods are tried out. Especially if the oven has many different or combinable functions, the effort required can be disproportionately high.

[0005] EP 3909479 A1 proposes, based on a role that is essential for a dish to be prepared, determining ingredients and processing steps for processing the ingredients in order to fulfill at least one role. CN 112369122 A relates to a technique for determining a cooking parameter of a dish based on processing information from a recipe.

[0006] Existing techniques require either large amounts of predefined information or complex processing of information associated with ingredients or processing steps in a recipe. As a result, most known approaches are only partially suitable for use in a household.

[0007] One object underlying the present invention is to provide an improved technique for providing a cooking recommendation for a dish. The invention achieves this object by means of the subject matter of the independent claims. Subclaims specify preferred embodiments.

[0008] According to a first aspect of the present invention, a method for providing a cooking recommendation for a dish to be cooked comprises the steps of capturing a human-understandable description of the dish; determining a first classification of the description with respect to a first predetermined taxonomy; deriving a second classification of the dish from the first classification and with respect to a second predetermined taxonomy; wherein the second taxonomy describes cooking properties of the dish; and providing a cooking recommendation for the dish. Based on the cooking recommendation, a household appliance, such as a stove or an oven, can then be controlled. Accordingly, according to one aspect, the present invention also comprises a method for controlling a cooking appliance, which comprises the method for providing a cooking recommendation.

[0009] It has been recognized that a human-understandable description of a dish can vary greatly, whereas the cooking parameters associated with the dish can change only relatively slightly. For example, different sponge cakes can be prepared with different flavoring ingredients, whereas the type, duration, and temperature of the cooking process can remain virtually unchanged. While a human may perceive a significant difference between a cherry cake and a strawberry cake, this variation may have little relevance in terms of cooking technology. Using the first taxonomy, a dish can be classified in such a way that it is sufficiently precise for subsequent purposes. The second taxonomy can break down processes, ingredients, or characteristics of the dish, allowing a broader determination of a cooking recommendation.The first classification can be converted into the second so that relevant parameters can be extracted for determining the cooking recommendation.

[0010] The human-readable description can be given in different forms. In particular, the human-readable description can include an image, a label, and / or a textual description. The label or description can alternatively be given in writing or verbally. For example, the food can be referred to as pizza, saffron cake, or lasagne. One or more images can be taken, for example, from a cooking recipe or a serving suggestion. A user can also provide a description, image, or label of a previously prepared food. The human-readable description can essentially comprise information that might be found on a restaurant menu. In another embodiment, a flavor of the food can be linguistically outlined.This could include, for example, a taste, an aroma or a figurative comparison.

[0011] The cooking recommendation preferably comprises a selection of a cooking appliance and / or a parameterization of a cooking appliance. For example, the cooking recommendation can contain an indication as to which of several possible cooking appliances is suitable for preparing the dish. A user can specify which cooking appliances they have at their disposal and the cooking recommendation can select the most suitable appliance for the dish. The parameterization can comprise a mechanical configuration, for example whether a tray, a drip tray or a rack should be used in an oven or at what height the rack or tray should be inserted into the oven. Furthermore, the parameter can comprise a cooking method or a cooking parameter, in particular an operating mode, an operating program or a temperature or a temperature profile.For example, the parameters may include heating an oven to 220°C using top and bottom heat without fan, and placing the food on a middle rack. Furthermore, it may be specified that the oven should be heated to at least approximately 170°C before placing the food inside. The cooking recommendation may also include the use of a cooking program, such as temporarily assisting the cooking process with a built-in steamer, for example, to make a cake particularly moist.

[0012] The invention described herein may also make it possible to propose an innovative, complex or non-intuitive parameterization of a cooking appliance suitable for a dish, so that the dish can be successfully prepared in the sense of its description.

[0013] It is preferred that the second taxonomy be formed on the basis of an ontology that models general cooking-chemical relationships. The ontology can include formal definitions that preferably encompass a variety of dishes. In particular, processing steps that can be referred to as cooking can be modeled. It is generally preferred that the ontology models the largest possible variety of dishes.

[0014] There is preferably a predetermined assignment between the first and second taxonomies. The assignment can be implemented, for example, using a machine learning technique. For this purpose, an artificial neural network (ANN) trained on the basis of a large number of food descriptions can be used. The ANN is preferably trained using training data for which the assignment is already known (supervised learning). Optionally, training data can be related to a predetermined cooking appliance or to predetermined appliance parameters on the cooking appliance.

[0015] The association between the first and second taxonomies can be formed by the ontology. In one embodiment, the ontology allows the determination of cooking-chemical relationships during cooking of the food based on the human-understandable description, and the cooking recommendation can be provided based on these relationships. Providing the cooking recommendation can include steps of determining a food modeled in the ontology that is similar to the food to be cooked; sorting the food into the ontology; determining a process parameter for the food to be cooked; and providing the cooking recommendation based on the process parameter.

[0016] This allows both a cooking recommendation to be determined and the ontological model to be expanded. Optionally, the expansion can be improved based on external feedback. For example, a connection or information rated as good can be confirmed. Information or connections rated as poor can be modified or deleted. The quality of the ontology can be continuously improved by implementing the proposed procedure.

[0017] In one embodiment, the ontological model can be created using any generic data. To achieve good results, the model can be further improved using data on predetermined foods or dishes. Furthermore, the model can be trained or tuned with respect to an available cooking appliance, so that a cooking suggestion for that cooking appliance can lead to particularly accurate cooking recommendations. For example, a manufacturer can provide suitable data for the cooking appliances it produces. This data can be created empirically based on systematic cooking trials. Such data can also be generated or collected from user data. Furthermore, data can be provided for a generic cooking appliance that assumes certain minimum capabilities. For example, a generic oven can generate temperatures of up to approximately 210°C using top and / or bottom heat and has three racks of different heights in the cooking chamber.A specific appliance can also support, for example, convection, grilling, steaming or microwave cooking.

[0018] A cooking suggestion can be provided for a cooking appliance specified by a user. If no specific data is available for this cooking appliance, the cooking recommendation can be based on a generic appliance. The cooking recommendation may then not utilize special capabilities of the available appliance, but may still represent an acceptable recommendation for classic use of the appliance. Data or connections of the representation of the dish to be cooked in the ontology can be enriched using inference. This allows information or connections of the dish to be cooked to be deduced from already modeled information or connections. Techniques of subsumption, induction, or abduction can be used for inference.

[0019] In another embodiment, missing data about the food to be cooked is generated in the ontology and linked to the representation of the food to be cooked. The data can be determined, for example, by interpolation.

[0020] The ontology can model various cooking-chemistry relationships. In one embodiment, the ontology comprises reaction kinetics of food. More preferably, reaction kinetics of different orders are modeled. Reaction kinetics can be approximated, for example, using an Arrhenius equation. Other expressions are also possible. An Arrhenius equation approximately describes a quantitative temperature dependence in physical and, above all, chemical processes in which an activation energy must be overcome at the molecular level. It describes a phenomenological relationship and applies to many chemical reactions, especially in culinary or cooking chemistry.

[0021] The ontology can include threshold values ​​at which a cooking-chemical effect occurs. For example, a time-dependent browning effect of the Maillard reaction can be stored starting at a threshold temperature of ~140°C. This allows a predetermined cooking result of the food to be cooked to be specifically controlled by a parameter of the cooking appliance. For example, browning, thorough cooking, braising, and sous-vide cooking can be specifically controlled. It is further preferred that a threshold value is specified approximately and / or vaguely. Precise control of appliance parameters of the cooking appliance can allow the intensity of a cooking-chemical effect to be controlled. In the above example, for example, the intensity of the Maillard reaction can be varied by slightly deviating from the specified temperature.

[0022] The ontology can further include indicators for the achievement of predetermined cooking-chemical effects. For example, it can be modeled that a meat's core temperature is related to the denaturation of certain proteins and this is related to the achieved degree of doneness. Such a relationship can be quantified, so that a cooking-chemical effect can also be achieved or exploited gradually.

[0023] It is generally preferred that a chemical relationship be modeled as an approximation in the ontology. It is common for a specific chemical reaction to not proceed exactly as theoretically predicted. By using an approximation, such individual deviations can be accounted for.

[0024] According to a further aspect of the present invention, a device for providing a cooking recommendation for a food to be cooked comprises the following elements: an input device for capturing a human-understandable description of the food; an output device for providing a cooking recommendation; a memory with a first and a second taxonomy; wherein the first taxonomy describes a food in a human-understandable manner; and the second taxonomy describes cooking properties of the food; a processing device which is configured to determine a first classification of the description with respect to the first taxonomy; to derive a second classification therefrom with respect to the second taxonomy; and to provide a cooking recommendation for the food on the basis of the classification.

[0025] The processing device can be configured to fully or partially execute a method described herein. For this purpose, the processing device can be implemented electronically and, for example, comprise a programmable microcomputer or microcontroller, and the method can be in the form of a computer program product with program code means. The computer program product can also be stored on a computer-readable data carrier. Features or advantages of the method can be transferred to the device, or vice versa.

[0026] The processing device is further preferably configured to determine a second taxonomy with respect to an ontology; the ontology models general cooking-chemical relationships.

[0027] In one embodiment, the human-understandable description includes a desired cooking result. For example, a degree of cooking can be specified from warm to rare, rosy to well-done. A consistency can be specified, for example, as crispy, soft, juicy, or dry. A cooking depth can be indicated by terms such as grilled, gratinated, or roasted. This can improve the description of the expected cooking result. The cooking recommendation can be determined based on such keywords.

[0028] A cooking property of the second taxonomy may include one of a physiochemical property, a physical state of the food before cooking, a dimension of the food or an ingredient, a number of pieces or portions to be prepared at the same time, a cooking vessel to be used, a common ingredient, a common cooking method and a specific culinary concept for the food.

[0029] The second taxonomy can be based primarily on food chemistry considerations. The structure of the food can be classified according to the cooking properties of its ingredients. Information unrelated to the cooking process, such as aroma or color, can be neglected.

[0030] It is preferred that multiple cooking properties be determined, with the cooking recommendation being provided based on the multiple determined cooking properties. In one embodiment, multiple cooking properties of a component can be determined, for example, a mass and a liquid-to-dry matter ratio. In another embodiment, properties of different components can be determined, for example, a moisture content of a dough or a moisture content of raisins in the dough.

[0031] Especially when the human-understandable description comprises several parts, information from all parts can be evaluated to create the cooking description. For example, a picture of the finished dish can be used to estimate a quantity, a preferred cooking vessel, or a degree of doneness. A basic classification can be derived from the name of the dish.

[0032] It is particularly preferred that the first and / or second taxonomy be hierarchically organized. Mapping between the first and second taxonomies can occur at different levels of the respective hierarchy. Mapping between elements of different hierarchy levels can also occur. Such a mapping can involve a 1:1, a 1:n, or an n:m relationship.

[0033] A node in the second taxonomy can be assigned information about the preparation of the food. The preparation can, in particular, relate to a processing step prior to cooking. The cooking recommendation can preferably be determined based on the assigned information. For example, the information can relate to a mixing process of ingredients, an intermediate step relevant to culinary chemistry, or the tempering of the uncooked food. The cooking recommendation can take corresponding parameters into account.

[0034] In a further preferred embodiment, the second taxonomy comprises limits for a permissible deviation of a cooking characteristic from a specification. For example, if a cooking recommendation for a predetermined amount of a dish is known, and the human-understandable description indicates a significantly different amount, the prepared cooking recommendation can be discarded. In another example, the dish can be assigned a predetermined ratio of ingredients. If the described dish deviates from this ratio more than permitted, a cooking recommendation assigned to the predetermined ratio can be adjusted or discarded.

[0035] It is particularly preferred that a mapping of the first taxonomy to the second taxonomy be predetermined. Endpoints of the taxonomies can be associated with each other. It is preferred that the mappings target as few nodes or endpoints of the second taxonomy as possible.

[0036] Each endpoint of the second taxonomy can be assigned a cooking recommendation. To determine the cooking recommendation, it may be sufficient to prepare any dish whose description is assigned to this endpoint. In this way, a large number of dishes can be mapped to a relatively small number of cooking recommendations. A cooking recommendation assigned to an endpoint of the second taxonomy can, for example, be determined empirically. To do so, one of the dishes can be prepared and cooked. Optionally, different cooking methods or temperatures can be tried out. In one embodiment, the recommendation indicates one of several predetermined cooking appliances. For example, a first cooking appliance can comprise an oven, a second cooking appliance a steamer, a third cooking appliance a microwave oven, and a fourth cooking appliance a cooktop with temperature control. Multiple cooking recommendations assigned to different cooking appliances can also be provided.This allows a user to decide for themselves which cooking appliance they would like to use to prepare the food.

[0037] In another embodiment, the cooking recommendation can be determined based on a cooking technique supported by a predetermined cooking appliance. For example, if a combination appliance is available that combines the technologies of a conventional oven with those of a microwave or steamer, the preferred cooking technique can be specified.

[0038] In general, the cooking recommendation can include a temperature and a cooking time. The cooking recommendation can also be complex and span several consecutive cooking processes. A single cooking process can specify a cooking technique, a temperature, and / or a cooking time.

[0039] Non-limiting embodiments of the invention will now be described in more detail with reference to the accompanying figures, in which:

[0040] Figure 1 shows a device for providing a cooking recommendation for a dish;

[0041] Figure 2 shows an exemplary first taxonomy;

[0042] Figure 3 shows an exemplary second taxonomy;

[0043] Figure 4 shows an exemplary first taxonomy with additional information;

[0044] Figure 5 shows an exemplary second taxonomy with information and recommendations;

[0045] Figure 6 shows an exemplary mapping between a first and a second taxonomy;

[0046] Figure 7 shows a preferred embodiment of a device

[0047] Figure 8 shows a first exemplary ontology; and

[0048] Figure 9 illustrates a second exemplary ontology. Figure 1 shows a device 100 for providing a cooking recommendation for a dish. The device 100 is configured to receive a human-understandable description 105 of a dish and to provide a cooking recommendation 110. The description 105 may include a name, an image, or a description of the dish. The description 105 may be in graphical, textual, or linguistic form. The dish includes food prepared from ingredients for humans. The description 105 may, for example, include a recipe, an excerpt from a menu, or a descriptive statement about a dish.

[0049] The cooking recommendation 110 typically includes a temperature and a duration for which the uncooked food should be exposed to the temperature in a cooking appliance 115. Optionally, the cooking recommendation 110 includes a target value that a sensor value from a sensor of a cooking appliance in use should assume. For example, an oven may have a thermometer for determining a core temperature of a food, and the cooking recommendation may state that the core temperature should reach a value of 55°C. For this purpose, the cooking recommendation 110 may include a temperature acting on the food by means of a heat source, here, for example, 110°C. The cooking recommendation 110 may also specify a heat source, for example, top or bottom heat in an oven, convection heat, or a grill. Furthermore, a cooking technique may be specified, for example, radiant heat, convection heat, steam cooking, or microwave cooking.Different cooking techniques can be assigned to different cooking appliances 115; however, a cooking appliance 115 can also be configured to perform different cooking techniques. The cooking recommendation can also include a sequence of individual cooking recommendations that follow one another.

[0050] The device 100 comprises an input device 120, an output device 125, a processing device 130, and a memory 135. The input device 120 is illustrated by way of example as a microphone for capturing human speech; a different input device 120 can be used to capture the description in a form other than acoustic. For example, an alternative input device 120 can comprise a camera, an interface for receiving images or text, or an input keyboard. The output device 125 is also illustrated by way of example as a loudspeaker for outputting vocal information; however, the cooking recommendation 110 can also be displayed in another way, for example numerically or graphically, for which purpose a textual or graphical output device 125 can be provided. The processing device 130 is preferably embodied as a microcomputer.A first taxonomy 140 and a second taxonomy 145 are stored in memory 135. Note that in another embodiment, the contents of memory 135 may be stored not locally in device 100, but rather at a remote location. The remote location may include another processing device that can perform portions of the operation of processing device 130 of device 100 described herein.

[0051] The first taxonomy 140 describes a food in a human-understandable way or allows the classification of a food specified by the description 105. Thus, a large number of foods can be assigned to one of several predetermined classes of the first taxonomy 140. This process can also be called classification.

[0052] The second taxonomy 145 is designed to describe the cooking properties of foods or to allow a food to be classified into one of several predetermined categories, whereby foods within a category may have similar or identical cooking properties. A cooking recommendation can be assigned to each such class of foods. Alternatively, a mapping or derivation rule can be specified, by means of which a cooking recommendation can be derived from the class. The derivation rule can contain parameters that include, for example, a quantity, a number, or a mass ratio of ingredients.

[0053] Both taxonomies 140 and 145 are preferably organized hierarchically and can encompass different categories or levels. Connections between taxonomies 140 and 145 can occur at any level of the respective hierarchy.

[0054] The second taxonomy 145 may, for example, include, in a first exemplary category, physico-chemical properties, for example

[0055] - a weight

[0056] - a heat capacity

[0057] - a density

[0058] - a ratio of solid to liquid ingredients a ratio of water-binding, dry ingredients to liquid ingredients

[0059] In a second exemplary category, a physical state of the dish before preparation can be specified, for example

[0060] - firmly

[0061] - fluid

[0062] - semi-frozen

[0063] - frozen

[0064] In a third exemplary category, a geometry of a cooking vessel can be specified, for example

[0065] - a volume or dimensions

[0066] - a form in which an ingredient is present, for example whole or cut into small pieces

[0067] A fourth example category may specify the number of dishes or portions prepared simultaneously. For example, preparing four pizzas simultaneously may require different cooking parameters than preparing just one pizza simultaneously.

[0068] A fifth exemplary category may specify a cooking utensil or cooking vessel in which the cooking process is to take place.

[0069] In a sixth exemplary category, common ingredients of a dish can be mentioned, for example

[0070] - a range of a quantity ratio (e.g. a weight proportion of flour)

[0071] - a range of proportions according to food groups (e.g. a weight percentage of all types of flour used in the dish)

[0072] - a range of a quantity ratio according to food properties (e.g. a ratio of solid to liquid components)

[0073] A seventh exemplary category may include common preparation methods for a dish, such as stirring, frying, or braising. An eighth exemplary category may include specific culinary concepts for a dish, such as the type of dough, such as shortcrust pastry, sourdough, yeast dough, quark-oil dough, puff pastry, or batter.

[0074] Also shown in Figure 1 are method steps of a method 150 for providing a cooking recommendation 110 for a dish. In a step 155, a human-understandable description 105 of the dish can be recorded. In a step 160, a first classification of the dish with respect to the first taxonomy 140 can be performed. In this case, a class of the first taxonomy 140 into which the dish falls can be determined. Multiple classes into which partial aspects of the dish fall can also be determined.

[0075] In a step 165, a second classification can be derived from the determined first classification, which is related to the second predetermined taxonomy. The second classification can be assigned to the first classification. The second classification can include one or more cooking properties of the food or one of its ingredients.

[0076] In a step 170, a cooking recommendation 110 for the food can be determined based on the second classification. In a step 175, the provided cooking recommendation can be output. In one embodiment, the cooking recommendation is directed to a user and can be presented textually, numerically, or symbolically. In another embodiment, a cooking appliance 115 can be controlled directly based on a parameter of the cooking recommendation. For example, an oven can be controlled to reach a predetermined temperature specified in the cooking recommendation 110.

[0077] Figure 2 shows an exemplary first taxonomy 140. The illustrated first taxonomy 140 relates to specific cakes and does not claim to be exhaustive. The illustrated first taxonomy 140 may be part of a larger first taxonomy 140.

[0078] The illustrated first taxonomy 140 includes, in an exemplary first category, cakes 205 from a user's perspective. Pound cakes 210 are specified in an exemplary second category. A third exemplary category can distinguish between chocolate cakes 215, lemon cakes 220, and marble cakes 225 without icing. A fourth category can be classified according to the type of cake icing. Examples available here include no icing 230, chocolate icing after baking 235, chocolate pieces in batter 240, or lemon icing after baking 245.

[0079] In a fifth category, which is also exemplary, a distinction can be made between molds in which the cakes 205 can be cooked or baked. For example, a loaf pan 250 and a Bundt pan 255 can be selected.

[0080] From a human perspective, pound cakes are a common type of cake with many variations. Each of the cakes mentioned in Figure 2 (chocolate, lemon, or marble cake) can be considered a different cake because they differ in appearance and flavor. Even if a similar batter is used for a chocolate cake, adding chocolate chips to the batter or adding chocolate coating after baking will, from a human perspective, produce a different result. Pound cakes are usually baked in a loaf or bundt pan, which affects their final shape. In this example, twelve different cakes are grouped under the pound cake category.

[0081] Figure 3 shows an exemplary second taxonomy 145. In a first exemplary category, cakes 305 are collected from a preparation or cooking perspective. In a second exemplary category, pound cake dough types 310 are defined. A composition of different ingredients and a preparation process can be specified. Furthermore, it can be specified by what extent the specifications can be deviated from without departing from the definition of the pound cake. In a third exemplary category, the size 315 of a baking pan can be specified. In a fourth exemplary category, a distinction can be made between a loaf pan 320 and a Bundt pan 325.

[0082] Figure 4 shows an exemplary first taxonomy 140 with additional information. For example, from the user's perspective, the pound cake 210 can be assigned an image 405 of a finished cake, images 410 of different baking pans, descriptions 415 of a desired result, or a distinguishing feature 420 or characteristic of a pound cake 210.

[0083] In the exemplary representation of Figure 4, images 405 of a chocolate cake 425, a lemon cake 430 and an unglazed marble cake 435 are possible.

[0084] Images 410 can, for example, be intended for a loaf pan 440 or a Bundt pan 445. Possible descriptions 415 of the result to be achieved include, for example, heavy and moist 450, with a dense consistency 455, or an indication 460 that the cake should be cut into slices and a slice should be held in the hand.

[0085] As a distinguishing feature, the 420 recipe uses a ratio of 465, which requires one part butter, one part sugar, one part egg, and one part flour. The name "Pound Cake" comes from the fact that each part traditionally amounts to one pound.

[0086] Figure 5 shows an exemplary second taxonomy 145 with information and cooking recommendations. By way of example, a common preparation method 505, for example, in the form of kneading 510, is specified for pound cake dough 310. Also by way of example, a fixed ratio 515 of ingredient groups is assigned to pound cake dough 310. For example, a ratio 520 of one part each of butter, sugar, egg, and flour can be specified. Furthermore, a permitted deviation 525 can be specified. Suitable replacement ingredients 530 can be specified for ratio 520. Furthermore, categories 535 of similar ingredients can be determined.

[0087] With respect to the loaf pan 320 and the Bundt pan 325, a size or dimension 540 or a suitable material 545 can be specified. The loaf pan 320 and the Bundt pan 325 can each be assigned a dimension 540 and a material 545.

[0088] A first cooking recommendation 550 can be assigned to the loaf pan 320. This recommendation can include, for example, an oven size 555, the use of a steam cooking function 560, or a microwave function 565. A second cooking recommendation 570, each with associated individual information 555 to 565, can be assigned to the Bundt pan 325. Cooking recommendations 550 and 570 are shown with thick lines in Figure 5. Cooking recommendations 550, 570 are special versions of a cooking recommendation 110.

[0089] The second taxonomy 145 may contain a range of information that may be helpful in compiling process parameters such as optimal furnace settings (e.g., heating method, time, and temperature parameters in a furnace). This information may include, but is not limited to:

[0090] - Proportions of ingredients

[0091] - Ratios of ingredient groups

[0092] - Ratios of ingredients according to physico-chemical properties

[0093] - Common cooking or baking techniques

[0094] - Common dimensions and materials of cookware and the optimal process parameters themselves.

[0095] The information may also include a list of different process parameters depending on the main appliance characteristics, for example, different parameters for ovens with and without steam function.

[0096] Figure 6 shows an exemplary mapping 600 between a first taxonomy 140 and a second taxonomy 145. The first taxonomy 140 from Figure 2 is shown in a left-hand area, and the second taxonomy 145 from Figure 3 is shown in a right-hand area. The mapping 600 maps all twelve variants of the first taxonomy 140 onto only two different elements of the second taxonomy 145. In the illustrated case, it may be sufficient to prepare no more than two fundamental compositions of dishes from the second taxonomy 145 in order to be able to provide suitable recommendations 550, 570 for all dishes classifiable in the first taxonomy 140.

[0097] Figure 7 shows a device 100 in a preferred embodiment in a side view. A mobile device 710 can be stored on a base device 705 and connected to it for data purposes, preferably via a wireless interface such as Bluetooth LE. Both the base device 705 and the mobile device 710 can each have an input device 120, an output device 125, or a processing device 130. Preferably, the base device 705 is connected via a further interface to an external device that can implement computationally intensive processes such as recognizing an image or converting speech into text or vice versa. A taxonomy 140, 145 can be provided in a memory 135 of the base device 705 or the mobile device 710. In a further embodiment, one of the taxonomies 140, 145 can also exist on the external site.

[0098] It should be noted that a technique presented herein may also run on a platform other than the illustrated device 700, for example, in an app on a mobile device (smartphone). In another embodiment, the technique may be implemented as an extension (skill) running on a smart speaker or a device with voice recognition (Alexa).

[0099] The technology described here can provide an improved electronic cooking assistant that can, in particular, support a wide variety of recipes and keep the testing effort required by the manufacturer of the cooking appliances within a reasonable and efficient framework. Ultimately, this effect can be achieved by reducing the wide variety of recipes offered to a much smaller number of technical cooking variants.

[0100] It is proposed to apply combinatorics mechanisms based on vectors that allow a model to weight the relevance of cooking terms in the context of an ontology or a chemical process. An ontology can be represented as a knowledge graph and typically includes categories such as ingredients, cooking techniques and equipment, as well as the chemical reactions that occur during cooking, for example, caramelization, a Maillard reaction, or fermentation. From a combination of such information, the ontology can determine a cooking recommendation. It should be noted that properties of a cooking appliance can also be modeled by the ontology. This allows parameters for the cooking appliance to be determined with regard to the cooking process of a dish.

[0101] The model can implement a form of machine learning. In particular, the ontology can be made to improve its definitions over time using supervised learning. For example, the model can be trained to understand the meaning of a term depending on its position and surrounding terms in the human-readable description. This can significantly improve the correct interpretation of a culinary text. The properties and grouping of words can be adjusted during the training phase to reflect not only the isolated meaning of a word but also its contextual relationships. By incorporating ontological structures and an understanding of chemical processes, the invention enables more precise modeling and prediction of culinary outcomes.

[0102] The model can be fine-tuned using a specially curated dataset. Such a dataset preferably includes a large number of recipes, descriptions, and culinary texts, preferably from different regions or languages. Through targeted training, the model can learn to recognize and account for subtle nuances and regional differences in the use of culinary terms.

[0103] To determine a cooking recommendation, an ontology can be used that encompasses formalized expert knowledge about food preparation. The ontology can, in particular, contain knowledge about the following:

[0104] - Foods and their properties, whereby foods with similar properties can also be grouped into food groups;

[0105] - Initial conditions of the food;

[0106] - Target states depending on the food;

[0107] - Approximations of reaction kinetics of chemical processes in food;

[0108] - parameters of the cooking climate; and

[0109] - Optimal process parameters for a food product.

[0110] Furthermore, a technical component can be provided that analyses similarities between foods, as well as a technical component that generates food-relevant details and a technical component that produces a cooking recommendation in the form of appliance settings or usage instructions using the data from the other components. The component for generating food-relevant details can in particular comprise an interface to a database so that, for example, a cooking-chemical parameter can be looked up for a dish or an ingredient. Such a parameter can comprise lexicographical knowledge or cooking-chemical knowledge. Furthermore, a usual use or treatment of the food can be determined, for example, from a collection of recipes.

[0111] Similar foods can be identified based on similar cooking chemical properties. Similarity can also be determined based on similar common uses, especially in other cooking recipes.

[0112] For example, the process for creating a device setting and usage instructions for a new food item might be as follows:

[0113] - Determining similarities between the food to be cooked and a food modeled in the ontology;

[0114] - Sorting the new food into the ontology;

[0115] - Create an instance for the food to be cooked in the ontology

[0116] - Enriching the data and connections of the new instance of the food in the ontology by inference on existing data of the ontology;

[0117] - Providing missing data from an external source and associating the data to the new instance for the food to be cooked in the ontology

[0118] - Determine process parameters of the new instance; and

[0119] - Determining appliance parameters of a predetermined cooking appliance for cooking the food.

[0120] The ontology can include reaction kinetics of chemical processes according to food groups. Molecular reaction kinetics describes the temporal course of chemical reactions at the molecular level and includes microkinetics, which deals with the kinetics of elementary reactions. Macrokinetics considers the influence of macroscopic heat and mass transport processes on the kinetics of chemical reactions and thus represents the link between reaction kinetics and chemical reaction engineering. Reaction kinetics can be specified in various orders (1, ..., n), which are approximated, for example, using Arrhenius equations. For example, vitamin degradation can be modeled as a function of temperature and time. Reaction kinetics can include an approximation for a threshold value above which a predetermined effect occurs, for example the temperature at which a Maillard reaction can be triggered.Furthermore, target parameters may be specified that must be met for certain chemical effects.

[0121] In general, it is not absolutely necessary to provide exact or 100% empirically confirmed formulas or values ​​for reaction kinetics; rather, estimates or approximate calculations are sufficient in most cases. An approximation is preferably designed in such a way that

[0122] - the chemical process itself depends on the cooking climate

[0123] - the dependence on the initial state

[0124] - and, above all, the effect on the target state depending on the cooking climate can be quantified in a simplified manner. These approximations are intended to facilitate well-founded extrapolations or interpolations based on verified reference values.

[0125] These relationships will be explained below using an example. Figure 8 shows an example ontology 800, represented in the form of a knowledge graph.

[0126] To determine appliance parameters for a given cooking appliance, a context can be determined that applies to the cooking of a given food. For example, such a context can include one or more of the following parameters:

[0127] - a goal to be achieved through the cooking process

[0128] - an energy input

[0129] - a cooking time

[0130] - Properties of food components

[0131] - Initial states of components

[0132] - a dietary adjustment or restriction

[0133] - Capabilities of an available cooking appliance

[0134] - a cooking method to be used

[0135] - further wishes or restrictions In further embodiments, the context can also be based on other or additional parameters.

[0136] Based on the context, a cooking method suitable for cooking the food can be determined. From the cooking method, a device that supports this cooking method can be determined. Furthermore, device parameters can be determined for this cooking method that best reflect the given context parameters. The device parameters can, in particular, include a device configuration or device setting, which can be provided to a user or controlled directly on the device.

[0137] In this example, a user wants to prepare Romanesco for a salad. The Romanesco should have a crisp texture and retain as many vitamins as possible. The user wants to use a steam oven for preparation. The ontology 800 does not contain a suitable setting for the user's steam oven, so a question about suitable parameters for the appliance cannot be answered directly. Therefore, the user asks for suitable settings, e.g., within an oven agent 826.

[0138] Within the ontology 800, as shown in Figure 2, no suitable instance for "Romanesco" is found. Therefore, parameters representing a device setting cannot be accessed directly.

[0139] Subsequently, instances of food in the ontology 800 can be compared with the food to be cooked. A comparison could also be performed using image recognition techniques, for example. Figure 9 shows an example instance 900 of an ontology 800.

[0140] In the chosen example, a similarity between the food Romanesco 916 and cauliflower 918 and broccoli 920 can be recognized. Therefore, the new food 916 is assigned to the group of "brassic vegetables" 922.

[0141] For example, using inference, new data and connections can be generated for the newly created instance 916. The system now knows that Romanesco, for example, can be prepared whole 906 or in florets 908 (initial state 914), or can be cooked both soft 928 and crisp 930 (target state 924). Romanesco 916 can be connected in the ontology 900, like cauliflower 918 or broccoli 920, to the initial state 914 and the brassica 922. The desired query can now be understood as an instance 916 within the ontology.

[0142] Within ontology 900, the target state "crisp" within the vegetable category 920 is linked to the degradation of pectin 946. The associated first reaction kinetics 954 describe a dependence on the thickness of the food. A significant effect on the degradation of pectin only occurs above a temperature threshold in the core of the food. Therefore, the size and thickness of the food are relevant parameters for approximating the heat input into the interior. For this reason, its size is estimated within component 916, for example, using generative kl.

[0143] In this way it is determined that:

[0144] - Romanesco is usually prepared in florets; and

[0145] - the usual size of the florets of Romanesco is estimated as in the similar food broccoli

[0146] Reaction kinetics 960 for vitamin degradation describes the average effect of temperature and time on degradation. The effect of average vitamin degradation could be described using averaged values ​​of the activation energy and rate constant in an Arrhenius equation and approximately estimated over the oven temperature range. Since vitamin preservation is set as the target state, the relationship with the cooking climate determines that steaming is optimal for vitamin preservation.

[0147] Thus, the data of the instances within the ontology 900 can be extended, for example like this:

[0148] Cooking method: Steaming

[0149] Vegetable Broccoli

[0150] Initial conditions: florets with an average diameter of 2.5 to 5 cm

[0151] Target states crisp, nutrients preserved

[0152] Cooking parameters can be determined from this:

[0153] Heating type: steam

[0154] Temperature -100 °C

[0155] Time 7 to 8 min

[0156] Accessories steam container

[0157] The identified reference (“Known settings for broccoli florets”) of the process parameters and the reaction kinetics can now be used to estimate optimal parameters for Romanesco 916, e.g. using regression or another optimization algorithm.

[0158] The "nutrient retention" effect is quantified as a function of time and temperature as input variables using the approximation of the activation energy and rate constant. Heat transfer into the food's interior is quantified using the new characteristic variable of Romanesco florets, depending on the cooking climate parameters (depending on temperature, steam content, turbulence, etc.). The minimal effect on the pectin content for the target state "crisp" is taken as a minimum condition, since otherwise the Romanesco 916 is still raw. This minimum threshold value of an effect, e.g., calculated from a specific activation energy and rate constant and core temperature and holding time as input parameters, can be estimated as a boundary condition.

[0159] Using these or other approximations, a mathematical optimum for the optimal process parameters can now be determined:

[0160] Cooking method: Steaming

[0161] Vegetable Romanesco

[0162] Initial conditions: florets, fresh

[0163] Target states crisp, nutrients preserved

[0164] From this, cooking parameters can be determined: Heating type steam

[0165] Temperature -100 °C

[0166] Time 10 min

[0167] Accessories steam container

[0168] The specific cooking parameters can be provided to the user so that they can set them on the cooking appliance. In another embodiment, the cooking parameters can also be transmitted directly to the cooking appliance.

[0169] Using the described technology, optimal appliance settings can be created without manual effort. Certain appliance settings can optimally consider specific heating types or technologies of a given cooking appliance. The appliance settings can also be created and controlled dynamically (see example ontology 800 in Figure 8). Cooking results can thus be optimal from the customer's perspective, or an available technology can be optimally used according to customer requirements. Generated settings can therefore go beyond "generic settings," such as those generated using a large language model such as Chat GPT, because device- or manufacturer-specific knowledge is also taken into account.

[0170] From a customer perspective, this can result in advantages:

[0171] - more assistance with optimal cooking settings, as more food can be covered without additional testing effort

[0172] - more specific assistance for optimal cooking settings, as the variance of food (initial conditions) can be covered

[0173] - more precise assistance on optimal cooking settings, which can be precisely mapped to customer requirements (target states)

[0174] - better cooking results, as individual customer wishes and relevant context are optimally taken into account Reference symbols

[0175] 100 device

[0176] 105 human-understandable description

[0177] 110 Cooking recommendation

[0178] 115 Cooking appliance

[0179] 120 input device

[0180] 125 Dispensing device

[0181] 130 processing facility

[0182] 135 storage

[0183] 140 first taxonomy

[0184] 145 second taxonomy

[0185] 150 procedures

[0186] 155 Understanding a human-understandable description of a dish

[0187] 160 Determining a first classification

[0188] 165 Deriving a second classification

[0189] 170 Determining a cooking recommendation

[0190] 175 Displaying the cooking recommendation

[0191] 205 Cakes from the User's Perspective

[0192] 210 Pound Cake

[0193] 215 Chocolate Cakes

[0194] 220 Lemon Cake

[0195] 225 Marble cake without icing

[0196] 230 without cast

[0197] 235 with chocolate icing after baking

[0198] 240 with chocolate pieces in the dough

[0199] 245 Lemon glaze after baking

[0200] 250 loaf pan

[0201] 255 Bundt pan 305 Cake from a preparation perspective

[0202] 310 Pound Cake Dough

[0203] 315 Size of a baking pan

[0204] 320 loaf pan

[0205] 325 Bundt cake pan

[0206] 405 Picture of a finished cake

[0207] 410 pictures of different baking tins

[0208] 415 Description of a result

[0209] 420 Distinguishing feature

[0210] 425 Chocolate Cakes

[0211] 430 Lemon Cake

[0212] 435 unglazed marble cake

[0213] 440 loaf pan

[0214] 445 Bundt cake pan

[0215] 450 Heavy and humid

[0216] 455 dense consistency

[0217] 460 sliced ​​and held in hand

[0218] 465 Quantity ratio

[0219] 505 usual preparation method

[0220] 510 kneading

[0221] 515 Ratio of ingredient groups

[0222] 520 ratio

[0223] 525 permitted deviation

[0224] 530 exchange ingredients

[0225] 535 categories of similar ingredients

[0226] 540 Size / Dimensions

[0227] 545 materials

[0228] 550 first cooking recommendation

[0229] 555 oven size

[0230] 560 steamer

[0231] 565 Microwave

[0232] 570 second cooking recommendation 600 illustration

[0233] 705 base unit

[0234] 710 mobile device

[0235] 800 Ontology

[0236] 802 Customer

[0237] 804 initial states

[0238] 806 Target Desired State

[0239] 808 Food Properties

[0240] 810 chemical and physical processes during cooking

[0241] 812 Cooking method

[0242] 814 Reaction kinetics

[0243] 816 Dependence on the cooking climate

[0244] 818 Effect on target states

[0245] 820 cooking climate

[0246] 822 sensors

[0247] 824 actuators

[0248] 826 Oven Agent

[0249] 828 Model of the cooking climate

[0250] 850 has

[0251] 852 plus

[0252] 854 influenced

[0253] 856 takes place during cooking

[0254] 858 influence

[0255] 860 requires the

[0256] 864 described by

[0257] 866 Part of

[0258] 868 Part of

[0259] 870 influenced

[0260] 872 influenced

[0261] 874 is included 876 determines optimal parameters for the respective condition

[0262] 878 influence

[0263] 880 quantifies impacts approximately

[0264] 882 influenced

[0265] 884 predicts

[0266] 886 measures

[0267] 888 influence

[0268] 890 regulatory adjustment

[0269] 892 defines control / regulation

[0270] 894 reads values

[0271] 896 is part of

[0272] 898 influenced

[0273] 900 Ontology

[0274] 902 fresh

[0275] 904 frozen

[0276] 906 whole

[0277] 908 in Röschen

[0278] 910 fresh frozen

[0279] 912 transition sizes

[0280] 914 initial states

[0281] 916 Romanesco

[0282] 918 Cauliflower

[0283] 920 Broccoli

[0284] 922 Brassicas

[0285] 924 Target state

[0286] 926 Textur

[0287] 928 soft

[0288] 930 crisp

[0289] 932 Taste

[0290] 934 natural flavor

[0291] 936 roasted aroma

[0292] 938 Color

[0293] 940 nutrients support 942 chemical processes by food group

[0294] 944 Processes during heating

[0295] 946 Pectin degradation

[0296] 948 Chlorophyll degradation

[0297] 950 Maillard reaction

[0298] 952 Vitamin degradation

[0299] 954 Reaction Kinetics 1

[0300] 956 Reaction Kinetics 2

[0301] 958 Reaction Kinetics 3

[0302] 960 Reaction Kinetics 4

[0303] 962 Dependence of reaction kinetics on parameters of the cooking climate

[0304] 964 Effect on Texture

[0305] 966 Dependence of reaction kinetics on parameters of the cooking climate

[0306] 968 Effect on Color

[0307] 970 Dependence of reaction kinetics on parameters of the cooking climate

[0308] 972 Effect on taste

[0309] 974 Dependence of reaction kinetics on parameters of the cooking climate

[0310] 976 Effect on nutrients

[0311] 978 Model of the cooking climate

[0312] 980 temperature

[0313] 982 heat transfer coefficients

[0314] 984 Modeling of energy transfer using microwaves

[0315] 986 Humidity

[0316] 988 Flow behavior

[0317] 990 direct energy input

Claims

PATENT CLAIMS 1. A method (150) for providing a cooking recommendation (110) for a food to be cooked, the method (150) comprising the following steps: Capturing (155) a human-understandable description (105) of the food; determining (160) a first classification of the description (105) with respect to a first predetermined taxonomy (140); - deriving (165) a second classification of the food from the first classification and with respect to a second predetermined taxonomy (145); - the second taxonomy (145) describes the cooking properties of the food; and Providing (175) a cooking recommendation (110) for the food.

2. The method (150) according to claim 1, wherein the second taxonomy is formed on the basis of an ontology that models general cooking-chemical relationships.

3. The method (150) according to claim 1 or 2, wherein there is a predetermined association between the first and the second taxonomy.

4. Method (150) according to claims 2 and 3, wherein the assignment is formed by the ontology.

5. The method (150) according to any one of claims 2 to 4, wherein providing a cooking recommendation comprises the following steps: Determining a food modeled in the ontology that is similar to the food to be cooked; Sorting the food into the ontology; Determining a process parameter for the food to be cooked; and providing the cooking recommendation based on the process parameter.

6. The method (150) according to claim 5, wherein data or connections of the representation of the food to be cooked in the ontology are enriched by inference.

7. The method (150) according to claim 5 or 6, wherein missing data of the food to be cooked are generated in the ontology and linked to the representation of the food to be cooked.

8. The method (150) of any one of claims 2 to 7, wherein the ontology comprises reaction kinetics of food.

9. The method (150) according to any one of claims 2 to 8, wherein the ontology comprises threshold values ​​at which a chemical effect occurs.

10. The method (150) of any one of claims 2 to 9, wherein the ontology comprises indicators for achieving predetermined cooking chemical effects.

11. The method (150) of claim 2, wherein the cooking chemical relationship is modeled as an approximation.

12. Device (100) for providing a cooking recommendation (110) for a food to be cooked, the device (100) comprising the following elements: - an input device for capturing a human-understandable description (105) of the food; - an output device for providing a cooking recommendation (110); - a memory with a first (140) and a second taxonomy (145); wherein the first taxonomy (140) describes a food in a human-understandable manner; and the second taxonomy (145) describes cooking properties of the food; - a processing device configured to determine a first classification of the description (105) with respect to the first taxonomy (140); to derive a second classification therefrom with respect to the second taxonomy (145); and to provide a cooking recommendation (110) for the food on the basis of the classification.

13. The device (100) according to claim 12, wherein the processing device is further configured to determine a second taxonomy with respect to an ontology; wherein the ontology models general cooking-chemical relationships.

Images