BR102020011594B1 - METHODS TO IMITATE TARGET FOOD ITEMS USING ARTIFICIAL INTELLIGENCE - Google Patents

Abstract

METHODS TO IMITATE TARGET FOOD ITEMS USING ARTIFICIAL INTELLIGENCE. Systems and methods for mimicking a target food item using artificial intelligence are disclosed. The system can learn from open source and proprietary databases. A prediction model can be trained using characteristics of the source ingredients to match those of the specific target food item. A formula comprising a combination of the most relevant source ingredients and their proportions can be determined using the trained prediction model. An existing set of recipes can be used as a dataset to train a recurrent neural network (RNN) and or other suitable models. The RNN can be used to determine a recipe to mimic the target food item. The recipe may comprise a cooking process for the set of ingredients in the formula and may be cooked by a chef. The recipe can be further modified as needed based on human feedback on sensory descriptors.

Description

TECHNICAL FIELD

[001] Disclosure generally refers to food science and artificial intelligence, in particular, the use of machine learning to mimic target food items.

BRIEF DESCRIPTION OF THE FIGURES

[002] Figure 1 illustrates a high-level diagram of a system that can use machine learning algorithms to generate a recipe using source ingredients to mimic a given target food item according to certain modalities;

[003] Figure 2 illustrates a source ingredient database configured to store a respective set of characteristics for each of a plurality of source ingredients in certain embodiments;

[004] Figure 3 illustrates a target ingredient database configured to store a respective set of characteristics for each of a plurality of target ingredients in certain embodiments;

[005] Figure 4 illustrates an example of a set of existing recipes that can be stored in an existing recipe database;

[006] Figure 5 illustrates processing steps for a recipe to include atomic directions according to certain modalities;

[007] Figure 6 illustrates an example of a set of recipes that can be stored in a chef's database;

[008] Figure 7 illustrates a block diagram for a formula generator in certain embodiments;

[009] Figure 8 illustrates a block diagram for a revenue generator in certain embodiments;

[0010] Figure 9 illustrates a computer-implemented method for generating a recipe to mimic a certain target food item according to certain embodiments;

[0011] Figure 10 illustrates a computer-implemented method for determining a formula for a recipe to mimic a target food item; It is

[0012] Fig. 11 is a computer-implemented method for determining a recipe comprising a cooking process for the formula.

DETAILED DESCRIPTION

[0013] The following description of embodiments is not intended to limit the invention to those embodiments, but rather to enable anyone skilled in the art to make and use.

[0014] Today, many negative consequences of the use of animals in the food industry are known, such as deforestation, pollution, human health conditions and allergies, among others. An animal-based food item can be a food product that includes any animal-based ingredients, such as cow's milk and all dairy products, egg-based emulsions, such as mayonnaise, meat-based foods, such as beef patties, meat, sausages etc. There are several products available on the market that can provide substitutes for animal-based foods, for example chicken, meat dumplings, milk, etc. Disclosed embodiments may utilize data science, food science, and/or machine learning algorithms to find a combination of source ingredients that might taste, look like, and/or taste like a particular target food item. The source ingredients and/or the target food item may be plant-based, animal-based or synthetic. Certain embodiments may provide recipes using plant-based ingredients that can mimic animal-based foods from sensory (eg, taste and/or texture) and/or visual perspectives. In some instances, a combination of plant-based ingredients can be cooked using a certain recipe to taste, look and/or taste like cow's milk.

[0015] In certain embodiments, the machine learning algorithms may include creating a screening of a given target food item and a plurality of source ingredients, for example, a set of N source ingredients. In some examples, the given target food item may be animal-based and the source ingredients may include plant-based ingredients. In some examples, target food items can have any suitable base (e.g., animal-based, plant-based and/or synthetic, etc.) and source ingredients can include any suitable combination of animal-based ingredients, plant-based ingredients, plants, synthetic ingredients and/or any suitable type of ingredients. Screening can represent each source ingredient in a D-dimensional space that includes characteristics related to physicochemical, nutritional and/or molecular descriptors (for example, where different source ingredients may have different characteristic values for the physicochemical descriptors, nutritional and/or molecular, etc.). Various prediction algorithms can be trained to match from the hypothesis space of the N source ingredients (used as data features) to the given target food item using a feature selection process. Prediction models can use one or more supervised machine learning approaches, for example, gradient growth trees and Lasso regression (contraction and minus absolute selection operator), to make predictions.

[0016] In certain embodiments, prediction algorithms may use feature compression techniques such as kernel principal component analysis and/or auto-coding to train the prediction model. In some other embodiments, the prediction model can be trained without any feature compression technique. Once the prediction model is trained, the most important traits used for prediction can be selected as potential candidates to mimic the target food item. After selection of possible candidates, an optimization process can be performed to find a formula comprising specific proportions of the source ingredients to mimic the target food item.

[0017] Having estimated the combination of source ingredients to be used in the formula and their proportions, one can discover a cooking process that can be applied to the selected source ingredients. The cooking process can also include actions to be taken and a sequence for performing the actions. For example, actions might include grinding, mixing, boiling, frying, etc. In certain embodiments, one or more recurrent neural network (RNN) models (and/or other suitable artificial intelligence models) can be trained with already existing recipes. Additionally or alternatively, any suitable artificial intelligence approach (eg described here etc.) can be applied (eg model training etc.) based on (eg trained with) any suitable recipes. Recipes may be sourced from the web and/or other sources (eg, produced from templates described here) and may be modified to include extensive hand-processed markup that may contain oversight needed to differentiate between an ingredient and a process. The modified recipes, including the tags, can be used as a main data source to train the RNN model(s) (and/or other suitable artificial intelligence models). The RNN model(s) (and/or other suitable AI models) can be trained on the ingredients and various sequences to use those ingredients in the respective cooking processes. Given a set of ingredients based on the formula provided by the prediction model, the RNN model(s) (and/or other suitable artificial intelligence models) may be able to classify the given ingredients and predict each of the( s) ingredient process(es) (and/or other suitable models). In certain embodiments, at the end (and/or at any suitable time) a complete cooking process sequence can be produced that represents a recipe formula to mimic the given target food item.

[0018] Figure 1 illustrates a high-level diagram of a system 100 that can utilize machine learning algorithms to generate a recipe using source ingredients to mimic a given target food item in flavor, color, feel, and/or functionality. Source ingredients can include plant-based ingredients, animal-based ingredients, water-based ingredients, synthetic ingredients or a combination thereof. The target food item may also include plant-based ingredients, animal-based ingredients, water-based ingredients, synthetic ingredients or a combination thereof. The system 100 may include a source ingredient database 102, a target ingredient database 104, a formula generator 106, a recipe generator 108, an existing recipe database 110, a feedback panel from formula 112 and/or a chef's database 114.

[0019] Some non-limiting examples of herbal ingredients may include vegetables (e.g. onions, potatoes, garlic, spinach, carrots, celery, squash etc.), fruits (e.g. apples, pears, grapes etc.) , herbs (e.g. oregano, coriander, basil, etc.), spices (e.g. black pepper, turmeric, red pepper, cinnamon, etc.), oils (e.g. corn oil, olive oil), nuts (e.g. almonds, walnuts , pistachios, etc.), legumes (e.g. lentils, dried peas, soybeans, legumes, etc.), starch, protein, fiber, carbohydrates, sugars, etc. Some non-limiting examples of animal-based ingredients may include dairy products (e.g. milk, butter, cheese, yogurt, ice cream, etc.), egg-based products (e.g., mayonnaise, salad dressings, etc.), egg products meat based (eg meatballs, sausages, hot dogs, bacon etc.) and/or seafood (eg fish, crab, lobster etc.). Synthetic ingredients can include artificially produced foods, for example, artificial meats, artificial sweeteners, artificial milk, etc.

[0020] Source ingredient database 102 may be configured to store a respective set of characteristics for each source ingredient in a plurality of source ingredients. The respective set of characteristics for each source ingredient may include physicochemical, biochemical, nutritional and/or molecular characteristics. Physicochemical characteristics may include data characteristics associated with physical and/or chemical characteristics of a given source ingredient. Biochemical characteristics can be associated with chemical processes that occur in living organisms, such as plants or animals. For example, near infrared (NIR) spectroscopy techniques can be used to identify physical and/or chemical characteristics of ingredients. Nutritional features can include nutritional facts for the ingredient. Molecular characteristics may include molecular formula or atomic arrangement of the ingredient. As an example, the trait set may include amino acids, vitamins, existing minerals (e.g. magnesium, manganese, zinc, sodium, potassium, selenium, copper, iron, phosphorus, calcium, ash, etc.), carotenes, ethyl alcohol alkaloids, fatty acids, carbohydrates, cholesterol, fiber, sugar, water or protein. An example source ingredient database 102 is discussed with reference to Figure 2).

[0021] As shown in Figure 2, the source ingredient database 102 can store a plurality of source ingredients, for example, a first source ingredient 102a, a second source ingredient 102b, a third source ingredient 102c and a Pth source ingredient 102p. Each of the first source ingredient 102a, the second source ingredient 102b, the third source ingredient 102c and the Pth source ingredient 102p may include a respective set of features, for example feature1, feature2, feature3 etc. feature1 can be a first type of feature, feature2 can be a second type of feature, feature3 can be a third type of feature, and so on. As an example, a feature 102a1, a feature 1 102b1,...., and a feature 1 102p1 may represent amino acids for the first source ingredient 102a, the second source ingredient 102b, and the Pth source ingredient 102p, respectively. A feature2 102a2, a feature2 102b2, and a feature2 102p2 may represent vitamins for the first source ingredient 102a, the second source ingredient 102b, and the Pth source ingredient 102p, respectively. A 3 feature 102a3, a 3 feature 102b3,...., and a 3 feature 102p3 may represent energy for the first source ingredient 102a, the second source ingredient 102b, and the Pth source ingredient 102p, respectively. Source ingredient database 102 may be implemented using memory, e.g., random access memory (RAM), electrically erasable programmable read-only memory (EEPROM), flash memory, hard disk drives, optical disc drives, solid state memory or any type of memory suitable for database storage.

[0022] Returning to Figure 1, the target ingredient database 104 can additionally or alternatively be configured to store a respective set of characteristics for each target ingredient in a plurality of target ingredients. The respective set of traits stored in the target ingredient database 104 may include the same types of traits as the source ingredient database 102, for example, physicochemical, biochemical, nutritional or molecular traits, but the target ingredient database 104 may additionally or alternatively include different types of traits as other databases (e.g. types of traits included in source ingredient database 102 etc.). In one example, the set of traits associated with a given animal-based food item, the respective set of traits for each of the plurality of food items (e.g., animal-based food items stored in a target ingredient database 104, etc. .) and the respective set of traits for each of the plurality of source ingredients (e.g., herbal ingredients stored in a source ingredient database 102) can be associated with the same set of trait types, such as where the same set of trait types comprises at least one of a nutritional descriptor trait type, a physicochemical trait type and a molecular descriptor trait type. An example target ingredient database 104 is discussed with reference to Figure 3).

[0023] As shown in Fig. 3 , the target ingredient database 104 can store a first target ingredient 104a, a second target ingredient 104b, a third target ingredient 104c and a Tth target ingredient 104t. Similar to source ingredient database 102, each of the first target ingredient 104a, the second target ingredient 104b, the third target ingredient 104c, and the tenth target ingredient 104t may include the respective set of traits feature1, feature2, feature3, and so on. on. As an example, a feature 1 104a1, a feature 1 104b1, .... and a feature 104t1 may represent amino acids for the first target ingredient 104a, the second target ingredient 104b and the Tth target ingredient 104t, respectively. A feature2 104a2, a feature2 104b2, .... and a feature2 104t2 may represent vitamins for the first target ingredient 104a, the second target ingredient 104b and the Tth target ingredient 104t respectively. A 3 feature 104a3, a 3 feature 104b3, ... and a 3 feature 104t3 may represent energy for the first target ingredient 104a, the second target ingredient 104b and the Tth ingredient 104t, respectively. The target ingredient database 104 may be implemented using memory, e.g. RAM, EEPROM, flash memory, hard disk drives, optical disc drives, solid state memory, or any type of memory suitable for database storage. .

[0024] Returning to Figure 1, the formula generator 106 can be configured to generate a formula for a given target food item using (e.g. based on etc.) source ingredients sourced from the source ingredient database 102. The formula may identify one or more source ingredients (eg two or more source ingredients, etc.) and/or their proportions. In some implementations, formula generator 106 may create a target food item screening and source ingredient set. The set of source ingredients may include part or all of the first source ingredient 102a, the second source ingredient 102b, the third source ingredient 102c, and the Pth source ingredient 102p. Sorting can represent each food item in a D-dimensional space comprising the physical-chemical, nutritional or molecular characteristics associated with the food item. For example, each food item can be represented in a vector space of various characteristics associated with physical-chemical, nutritional or molecular properties.

[0025] Formula generator 106 may be configured to determine a formula comprising a set of source ingredients and/or their proportions that correspond to a target in the feature space. In some examples, the formula generator 106 may identify a set of characteristics associated with the given target food item using the target ingredient database 104. For example, the target may be the target food item (e.g., a brownie base animal) and source ingredients can be plant-based (e.g. lettuce, carrots, peanut butter). In some examples, one or more prediction models can be trained to match the hypothesis space of the set of source ingredients (used as data features) to the given target food item using a feature selection process. Predictive model training cannot use sensory descriptors (e.g. taste, color, texture, or taste) as the data characteristics for matching with the target food item, however, matching may indirectly match imitation of the product(s). ) sensory descriptor(s) of the target food item. The prediction model can be based on a gradient boost tree for regression. The gradient enhancement tree can be trained to fit a specific target using available physicochemical, nutritional, or molecular characteristics. Each target can be a new regression problem. A formula can be determined by presenting a feature selection problem to the trained prediction model, where features can be the source ingredients. The feature selection process can include selecting the most relevant features using the prediction model as the potential set of source ingredients to be included in the formula. The formula generator 106 can be further configured to perform an optimization process to find specific ratios for the source ingredients selected to generate the formula. In certain embodiments, Lasso optimization can be performed to balance the proportions of different ingredients in the source ingredient pool. For example, ingredients with almost negligible contribution (for example, with a proportion less than 0.001%) can be excluded. In one variation, the formula generator 106 can be configured to determine a formula comprising a set of source ingredients without determining the proportions of the source ingredients.

[0026] In some implementations, the prediction model(s) may (m) use a feature compression technique such as kernel principal component analysis (KPCA) or autocoding; however, prediction models can be trained without feature compression steps. KPCA or autocoding can provide techniques for non-linear feature extraction to reduce the size of the feature dataset. Note that any techniques for dimensionality reduction can be used without deviating from the scope of the disclosed technologies.

[0027] In certain embodiments, forecasting model(s) (and/or other suitable models, such as revenue generator model(s) etc.), appropriate components of the system 100 embodiments and/or appropriate portions of the modalities of the methods described here may include, apply, employ, execute, use, build on, and/or associate with artificial intelligence approaches (e.g., machine learning approaches, etc.), including any one or more of the following: learning supervised (e.g., using gradient-gaining trees, using logistic regression, using reverse propagation neural networks, using random forests, decision trees, etc.), unsupervised learning (e.g., using an Apriori algorithm, using means clustering K), semi-supervised learning, a deep learning algorithm (e.g., neural networks, a constrained Boltzmann machine, a deep belief network method, a convolutional neural network method, a recurrent neural network method, autocoder method stacked, etc.), reinforcement learning (e.g., using a Q learning algorithm, using time difference learning), regression algorithm (e.g., ordinary least squares, logistic regression, stepwise regression, multivariate adaptive regression streaks , locally estimated scatterplot smoothing, etc.), an instance-based method (e.g., k-nearest neighbor, learning vector quantization, self-organizing map, etc.), a regularization method (e.g., ridge regression , minus absolute contraction and selection operator, elastic network, etc.), a decision tree learning method (e.g. classification tree and regression, iterative dichotomizer 3, C4.5, diquadratic automatic interaction detection, decision stub , random forest, multivariate adaptive regression streaks, gradient augmentation machines, etc.), a Bayesian method (e.g. naive Bayes, single dependency mean estimation, Bayesian belief network etc.), a kernel method (e.g. , a support vector machine, a radial basis function, a linear discriminant analysis, etc.), a clustering method (e.g. k-means clustering, expectation maximization, etc.), an associated rule learning algorithm ( for example, an Apriori algorithm, an Eclat algorithm, etc.), an artificial neural network model (for example, a Perceptron method, a Backpropagation method, a Hopfield network method, a self-organization method), a learning vector quantization, etc.), a dimensionality reduction method (e.g., principal component analysis, partial least squares regression, Sammon mapping, multidimensional scaling, projection search, etc.), an ensemble method (e.g., (eg, boosting, bootstrap aggregation, AdaBoost, stacked generalization, gradient-boosting machine method, random forest method, etc.) and/or any suitable artificial intelligence approach.

[0028] The models described here (eg, formula prediction models, revenue generator models, etc.) can be run or updated: once; at a predetermined frequency; every time a certain process runs; whenever a trigger condition is satisfied and/or at any other suitable time and frequency. Templates may be run or updated concurrently with one or more templates, serially, at varying frequencies and/or at any other suitable time. Each model can be validated, verified, strengthened, calibrated or updated based on newly received updated data; historical or updated data based on other appropriate data.

[0029] The formula may be a combination of selected source ingredients and/or their respective proportions intended to mimic the particular target food item. For example, a formula fi can be defined as a set of pairs of Mi, each comprising source ingredients and their respective proportions: fi = {(pi, Ii); (p2, I2); , (pMi, IMI)}. Once the combination of the source ingredients is achieved, the formula fi can be fed into the recipe generator 108 which can predict the order in which the source ingredients are to be used when cooking and the actions (e.g. chopping, boiling, frying, stirring, etc.) associated with one or more source ingredients.

[0030] The recipe generator 108 may be configured to generate a recipe comprising a cooking process for the set of source ingredients in the formula to mimic the target food item. The cooking process may include a set of actions, tools, steps and/or the respective proportions of the source ingredients based on the formula generated by the formula generator 106 (and/or any suitable source). As an example, a recipe ri = {fi, Qi} can be defined as a 2D tuple comprising the formula fi and a sequence of processes Qi. The sequence of Qi processes can be defined as {{(a 1, t1, h 1, ηi); (a2, t2, h2, η2); '" ; (aui, tui, hui, nui)}, where each tuple ((am, tm, hm ηm) (m = [1 • • .Ui]) can be composed of an am action (for example, to boil), a tool tm (e.g. a pan), a step hm in the recipe sequence (e.g. a third step to be performed), and the source ingredient set nm {I1, I2, . . . , IMi } involved in step hm Note that ui can be greater than or equal to Mi because an ingredient can be used in more than one step associated with ri.

[0031] The recipe generator 108 may include a deep recurrent neural network that can be trained using existing recipes that have been previously generated. Existing recipes can be obtained from various websites or other sources and can be processed to provide a structured dataset with a given format. In some cases, existing revenue can be collected by capturing multiple websites. Existing recipes can be stored in the existing recipe database 110. Each existing recipe can include raw text. As an example, each existing recipe might include an ingredient list that might specify a name of each ingredient, quantity of each ingredient, and a state of each ingredient (for example, four avocados, halved and pitted). Each existing recipe may also include instructions to describe a list of instructions for cooking the ingredients. A set of examples of existing recipes is discussed with reference to Figure 4. The set of existing recipes may form part of the existing recipe database 110.

[0032] Figure 4 illustrates an example of a set of existing recipes 400 that can be stored in the existing recipe database 110. The existing set of recipes 400 may include a first recipe 402a, a second recipe 402b, ... and a recipe Hth 402h. Recipes 402a, 402b, .... 402h may each include a name, ingredients and/or instructions. For example, the first recipe 402a may include a name 404, an ingredient list 406, and instructions 408. The ingredient list 406 may include a first ingredient 406a, a second ingredient 406b, ... and a Fth ingredient 406f. Each ingredient can include an ingredient name, quantity and/or state associated with it. For example, the first ingredient 406a may include an ingredient name 406a1, an amount 406a2, and a state 406a3. Likewise, instructions 408 may include a set of instructions, for example, a first instruction 408a, ... and a Gth instruction 408g. Existing recipe set 400 can be processed by recipe generator 108 to be used as a training set for training one or more RNN models (and/or other suitable artificial intelligence models etc.). For example, each recipe in the existing 400 recipe pool can be processed to have a specific format to prepare for RNN model training. Processing an example recipe is discussed with reference to figure 5).

[0033] Figure 5 illustrates the processing of an existing 500 recipe for a vegan chunky stew. Recipe 500 can belong to the set of existing recipes 400. For example, recipe 500 can be first recipe 402a, second recipe 402b, ... or Hth recipe 402h. The 500 recipe may have been obtained from an online website or another source. Recipe 500 may include a list of ingredients 502 and/or directions 504. Recipe 500 may be processed by recipe generator 108 to include a specific format to prepare for training one or more RNN model(s) (and/or or other suitable artificial intelligence models, etc.). In some implementations, processing may include a combination of automatic and manual processing. For example, the raw text of the ingredients list 502 can be automatically processed to identify ingredient names, amounts and/or states as shown in a modified ingredients list 506. The modified ingredients list 506 can include ingredients from the ingredients list 502 represented as one or more words (for example, oil, onion, salt, etc.). The amount of each ingredient in Modified Ingredients List 506 may or may not be specified and the units of measurement may vary between ingredients (eg 5 tbsp olive oil, 1 yellow onion). Each ingredient can have zero or more states associated with it (eg, sliced, thinly sliced, chopped, etc.).

[0034] Instructions 504 may describe a set of instructions for cooking the ingredients listed in ingredient list 502. In certain embodiments, instructions 504 may be manually processed (and/or automatically processed) to include tags to differentiate between words that correspond to an ingredient and the words correspond to a cooking process. For example, the set of instructions in directions 504 may be manually rendered (and/or automatically rendered) to include steps for cooking the ingredients, as shown in a cooking process 508. In some implementations, each step may be a paragraph of text including one or more atomic statements with only one verb (or any suitable number of verbs and/or other types of words) and each atomic statement can indicate an action to be performed on one or more ingredients. Cooking process 508 may include steps 508a for respective actions 508b to be performed on respective ingredients 508c. For example, as shown in figure 5, a step 1 can indicate a "heat" action to be performed on the ingredients "olive oil" and "soy sauce". The ingredients 508c listed in the cooking process 508 may include all of the ingredients in the ingredients list 502. In some examples, the cooking process 508 may also include tools (not shown) used in the process, for example, a large pot, a frying pan , a measuring spoon, etc.

[0035] Returning to Figure 1, the recipe generator 108 can use a set of existing recipes stored in the existing recipe database 110 as a training set to train one or more RNN models (and/or other suitable models of artificial intelligence, etc.). The RNN model(s) (and/or other suitable neural networks) can be a type of artificial neural network in which connections between nodes can form a directed graph along a temporal sequence. The RNN model(s) (and/or other suitable neural networks) can use its internal states to process the input sequence. The one or more RNN models (and/or other suitable artificial intelligence models etc.) can be trained to receive a formula fi generated by the formula generator 106 and provide a list of actions and/or the respective source ingredients used to the cooking process to mimic a particular target food item. In certain embodiments, to initiate generation of a new recipe for the target food item using formula fi, the source ingredients specified in formula fi may be encoded as a heat vector. To start the process, the heat vector can be fed into the RNN model (and/or other suitable models) with a reserved first label (eg START) and a first action and one or more ingredients can be obtained. To generate the next action and its associated ingredients, the previous action can be added to a list of previously generated actions. The new action list and ingredient list can now be used as inputs to one or more RNN models (and/or other suitable artificial intelligence models, etc.). Generation of the new recipe can end when a second reserved label (eg END) is returned.

[0036] A plurality of such recipes can be obtained for each target food item. For example, for a given target food item T, a set of recipes R = {r1, r2, . . . , rR} can be generated. A score s(fi)l based on the difference between the formula fi and the target food item T can be assigned to each recipe ri (i = [ 1...R]). In some examples, the s(fi) score for each recipe can be calculated as where the formula fi can be extracted from its respective recipe ri using any suitable function G(r). Here, Pk is a proportion of the ingredients Ik included in the formula fi, Ik j is the value of variable j in ingredient Ik, and Tj is the value of the target food item in characteristic j. However, an s(fi) score can be determined in any suitable way, based on any suitable variables.

[0037] In one example, a set of recipes (e.g. including any generated recipes etc.) may be generated, where each recipe in the recipe set may include a respective cooking process for a respective formula and where a score may be determined for each recipe in the recipe set based on match to the specified target food item (eg, match of the formula to the given target food item, such as degree of difference between the formula and the given target food item; etc.).

[0038] Based on the s(fi) scores (and/or other suitable data), one or more recipes can be harvested and/or cooked by one or more people (and/or machines, such as one or more robots, appliances by smart cooking and/or by any suitable entities and/or components, etc.). For example, the set of recipes R generated by the recipe generator 108 may be provided to the formula feedback panel 112. The formula feedback panel 112 may include a person to cook one or more recipes and/or a group of people to taste the cooked food item. For example, a chef can cook many recipes for one target, or he can cook one recipe in many different ways. The chef may cook the recipe as is, or may vary proportions or ingredients (e.g., a little; a lot; by a percentage amount; by an absolute amount; based on a chef's manual actions; based on recommendations and/or in other outputs of one or more models, etc.), as specified by the recipe formula. Feedback on the cooked food item can be provided by the chef and/or the group of people. For example, feedback can include comments on sensory descriptors (eg color, taste, taste, mouthfeel, etc.) as well as visual appearances. In some cases, the feedback provided by the formula 112 feedback panel may include a modified recipe formula. The modified recipe formula, its preparation, one or more images, the sensory feedback and/or any other relevant information can be saved in the chef's database 114. In some instances, the recipes generated by the recipe generator 108 can be in the similar format to the modified ingredients list 506 and/or the cooking process 508. In another example, the recipes may have been modified by the recipe generator 108 to include free text, similar to the ingredients 502 and the directions 504. Text Based Recipes free may be generated manually, automatically (eg using any suitable types of templates described herein, etc.) and/or through any suitable means. A set of sample recipe formulas, which can be stored in the chef's database 114, is discussed with reference to figure 6.

[0039] Figure 6 illustrates an example set of recipes 600 that can be stored in the chef's database 114. Set of recipes 600 can be generated using the system 100 in Figure 1. Set of recipe formulas 600 can be include a first recipe formula 602a, a second recipe formula 602b,... and a Wth recipe formula 602w. The first recipe formula 602a may include an ingredient list 604, a cooking process 606, photos 608, and human feedback 610. The ingredient list 604 may include two or more source ingredients from the plurality of source ingredients 102a to 102p, as discussed with reference to Figure 2. For example, ingredients list 604 may include first source ingredient 102a. Each ingredient can include a respective name and/or quantity. For example, the first source ingredient 102a might include a name 604a1 and an amount 604a2.

[0040] The cooking process 606 may include a set of instructions for cooking the ingredients 604. For example, the cooking process 606 may include a first instruction 606a, ... and an nth instruction 606n. Each statement can include multiple steps. For example, the first instruction 606a may include a first step 606a1,.... and a Jth step 606aj. In some examples, each instruction 606a to 606n may be in the same format as every instruction in directions 504.

[0041] The photos 608 may include images of the ingredients, cooked food or intermediate stages of the cooking process 606. The human feedback 610 may include a taste 610a, a color 610b and/or any other sensory feedback. Flavor 610a may include a cooked food flavor and color 610b may include a cooked food color. As an example, human feedback 610 can be provided by formula feedback panel 112.

[0042] In certain embodiments, recipe set 600 may include multiple versions of the same recipe to mimic a specified target food item. As an example, the first recipe formula 602a may correspond to an original recipe generated by the recipe generator 108 for a target food item, and the second recipe formula 602b may correspond to a modified recipe based on the formula feedback panel 112. Likewise, multiple versions of a recipe can be stored with different variations and/or corresponding photos and/or human feedback.

[0043] Figure 7 illustrates a block diagram 700 for the formula generator 106 of Figure 1. In some implementations, the formula generator 106 may include a sorting process 702, a prediction model 704, a compression model of feature 706 and/or an optimizer 708. Note that the formula generator components 106 may be implemented using software, hardware, firmware, or a combination thereof. In some implementations, one or more components of formula generator 106 may include a processor configured to execute instructions stored on a non-transient computer-readable medium.

[0044] Screening process 702 may be used to screen the target food item and/or source ingredients into a plurality of source ingredients. Screening can represent each food item in a D-dimensional space that can contain characteristics related to physicochemical, nutritional and/or molecular descriptors (for example, where different source ingredients may have different characteristic values for the physicochemical descriptors, nutritional and/or molecular, etc.). For example, each food item can be represented in a vector space of various characteristics associated with physical-chemical, nutritional or molecular descriptors. As discussed with reference to Figure 2 and Figure 3, characteristics associated with source ingredients may be stored in source ingredient database 102 and/or characteristics associated with the target food item may be stored in target ingredient database. 104. The screening process 702 may be configured to identify a set of characteristics associated with the given target food item, using the target ingredient database 104, creating a screening of the target food item and the plurality of source ingredients.

[0045] The prediction model 704 can be implemented using a machine learning technique for regression, such as gradient augmentation trees. Prediction model 704 can be trained to match the hypothesis space of the plurality of source ingredients (used as data features) to a target. For example, gradient-enhancing trees can be trained to fit a specific target using available physicochemical, nutritional, or molecular characteristics. A formula can be determined by presenting a feature selection problem to the trained prediction model, where features can be the source ingredients. The most relevant characteristics selected using prediction model 704 may be the potential set of source ingredients to be included in the formula. Training the predictive model 704 to match the set of characteristics of each of the plurality of source ingredients to the identified set of characteristics associated with the target food items may indirectly match the imitation of the sensory descriptors (e.g., flavor, color, texture, or taste). ) of the target food item; however, the set of characteristics used for matching may not directly include the sensory descriptors of the target food item. Additionally or alternatively, any suitable artificial intelligence approach (e.g. described herein etc.) can be used for prediction model 704.

[0046] In certain embodiments, feature compression template 706 may be used to perform feature compression, for example, kernel principal component analysis (KPCA) or autocoding. In some other embodiments, the prediction model 704 can be trained without any feature compression. For example, feature compression can be performed to reduce the size of the feature dataset.

[0047] Optimizer 810 can be used to perform an optimization process to determine specific proportions of source ingredients selected to generate the formula. In some examples, Lasso optimization can be performed to determine the proportions of different source ingredients in the source ingredient pool. Note that other techniques for finding optimal proportions of source ingredients in the formula are possible within the scope of the disclosed technologies.

[0048] Figure 8 illustrates a block diagram 800 for the recipe generator 108 of Figure 1. In some implementations, the recipe generator 108 may include an existing recipe collector 802, a recipe processor 804, a training model 806, a recipe selector 808, and/or a recipe formatter 810. Note that the recipe generator components 108 may be implemented using software, hardware, firmware, or a combination thereof. In some implementations, one or more components of recipe generator 108 may include a processor configured to execute instructions stored on a non-transient computer-readable medium.

[0049] Existing Recipe Collector 802 can be configured to retrieve existing recipes from online websites, manual entries, or other suitable sources. Existing recipe collector 802 can be configured to retrieve existing recipes automatically (e.g. via APIs and/or other data requests for recipe retrieval, capture, etc.), manually (e.g., through manual entries, etc.) and/or through any suitable means. Existing recipes can be processed to include a specific format and/or can be used as training data to determine a cooking process from the set of source ingredients provided in the formula to mimic a target food item. Existing recipes may be stored in existing recipe database 110, as discussed with reference to figure 4. Existing recipes may include recipes similar to recipe 500 discussed with reference to figure 5).

[0050] The recipe processor 804 can be configured to perform automatic and/or manual processing of existing recipes stored in the existing recipe database 110. Automatic and/or manual processing can be used to differentiate words corresponding to ingredients and the words corresponding to a process by marking the corresponding words. As discussed with reference to Figure 5, the recipe 500 may be processed to provide the modified ingredients list 506 comprising the name, quantity and/or states of the ingredients and/or the cooking process 508 comprising the atomic instructions for the respective ingredients.

[0051] The 806 training model may implement one or more RNN models (and/or other suitable artificial intelligence models, etc.) that may be used to determine a cooking process for the set of ingredients in the formula. One or more RNN models (and/or other suitable artificial intelligence models, etc.) can be trained on a sequence of ingredients and/or their respective processes from existing recipes. The 806 training model may be able to classify a given set of source ingredients and predict each of the source ingredient processes, ending up in a complete sequence that may represent a complete food recipe to mimic a given target food item. In certain embodiments, to initiate generation of a new recipe for the target food item using formula fi, the source ingredients specified in formula fi may be encoded as a heat vector. To start the process, the single heat carrier can be fed to the 806 training model, and a first action and/or one or more ingredients can be obtained. To generate the next action and its associated ingredients, the previous action can be added to a list of previously generated actions. This process can be repeated until a complete sequence of source ingredients representing a complete food recipe is obtained.

[0052] Recipe selector 808 may select one or more recipes provided by training model 806. In some examples, training model 806 may provide a plurality of recipes for each target food item. For example, for a given target food item, a set of recipes can be generated. In certain embodiments, recipe selector 808 may select one or more recipes based on an s(fi) score. The s(fi) score can indicate how similar the formula is to the target food item in the feature space. As an example, a lower score (eg based on a smaller difference between a formula and the characteristics of a target food item, etc.) may indicate that the recipe is closer in the characteristic space of the target food item. As another example, a higher score (eg, based on proper calculations, etc.) can be used to indicate preferable recipes (eg, recipes closer in the trait space to the target food item, etc.). The one or more selected recipes can be provided to the formula 112 feedback panel for feedback. The one or more recipes can be prepared by a chef and human feedback on any suitable combination of sensory descriptors (eg color, taste, flavor, mouthfeel, etc.) and/or visual appearances can be collected. In some instances, the one or more recipes may be modified based on feedback. In some cases, the feedback provided by the formula 112 feedback panel may include a modified recipe formula. The modified recipe formula, its preparation, one or more images, the sensory feedback and/or any other relevant information can be saved in the chef's database 114, as discussed with reference to figure 6.

[0053] The recipe formatter 810 can be configured to format the recipes in a certain format for storing in the chef's database 114. For example, each respective recipe formula can be stored in a certain format in the chef's database 114 as discussed with reference to FIG. 6. In some examples, the recipes generated by the recipe generator 108 may be in a format similar to the modified ingredients list 506 and/or the cooking process 508. In another example, the recipes may have been modified by recipe generator 108 to include free text, similar to ingredients 502 and/or directions 504.

[0054] However, modalities of a system 100 may be configured in any suitable manner.

[0055] Figure 9 illustrates a computer-implemented method 900 for generating a recipe to mimic a particular target food item. Method 900 can be performed by any suitable embodiments of system 100 (eg, Figure 1).

[0056] In step 902, a set of characteristics associated with the given target food item may be identified using a database of target ingredients. The target ingredient database can be configured to store a respective set of characteristics for each of a plurality of target food items. For example, the set of characteristics associated with the specified target food item may be identified using the target ingredient database 104. The set of characteristics associated with the target food item may include one or more of characteristic1, characteristic2, characteristic3, etc. for the respective target ingredient. In some examples, the target ingredient database 104 may store a respective set of characteristics for each of a plurality of animal-based food items (and/or any suitable type of food item, etc.). The respective set of characteristics may include amino acids, vitamins, carbohydrates, fiber, color, smell and/or texture, etc.

[0057] In step 904, two or more source ingredients may be identified using a database of source ingredients based on matching the set of identified characteristics associated with the target food item. The source ingredient database can be configured to store the respective set of characteristics for each of a plurality of source ingredients. As an example, two or more source ingredients can be identified using source ingredient database 102. In some examples, source ingredient database 102 can store a respective set of characteristics for each of a plurality of plant-based food items (and/or any suitable types of food items, etc.). The characteristic types of the respective characteristic set can be the same as the characteristic types of the characteristic set associated with the target food item, e.g. amino acids, vitamins, carbohydrates, fiber, color, smell and/or texture, other types of features etc. In certain embodiments, the formula generator 106 can identify the two or more source ingredients by training the prediction model 704 to match the respective set of characteristics for each of the plurality of source ingredients to the identified set of characteristics associated with the target food items. .

[0058] In step 906, a formula may be generated for combining the two or more source ingredients from the plurality of source ingredients in specific ratios based on matching the set of characteristics associated with the given target food item. As discussed with reference to Figure 1, the formula generator 106 can be used to generate the formula to combine the two or more source ingredients in specific proportions based on matching the set of characteristics associated with the given target food item T, using a feature selection process. Formula generator 106 may use optimizer 708 to perform an optimization process to determine specific ratios for combining the two or more source ingredients in the formula.

[0059] In step 908, a recipe including a cooking process for the formula is generated based on a set of existing recipes. The existing recipe set can be similar to the existing recipe set 400. The existing recipe set can be obtained from online features and stored in the existing recipe database 110. The cooking process can be determined by the recipe generator 108 using the formula. The cooking process may include actions to be taken on the two or more source ingredients and/or a sequence for performing the actions. For example, actions can include cutting, boiling, mixing, etc.

[0060] The recipe generator 108 can use the training model 806 to generate one or more recipes based on the training dataset and/or the formula. The training dataset may be prepared using the existing recipe set collected by the existing recipe collector 802 and modified by the recipe processor 804 to support a specific format. Recipe selector 808 may select the recipe to mimic the given target food item from one or more recipes generated by recipe generator 108 based on the respective s(fi) scores associated with each of the one or more recipes ri. In some instances, the recipe may be provided to the formula 112 feedback panel for feedback. The recipe can be prepared by one person and feedback on flavor and/or other sensory descriptors can be provided by the chef and/or a group of people who have tasted the cooked food item. Based on the feedback, the recipe can be modified and/or stored in the Chef 114 database.

[0061] Figure 10 illustrates a computer-implemented method 1000 for determining a formula for a recipe to mimic a target food item. Computer-implemented method 1000 can be performed using any suitable embodiments of system 100 of Figure 1 to determine a recipe formula to mimic the target food item using source ingredients. The target food item or source ingredients can be plant-based, animal-based, or artificially generated (eg, synthetic) food items.

[0062] In step 1002, a database of target ingredients and/or a database of source ingredients may be prepared. Each of the target ingredient databases and the source ingredient database can be described using the same types of traits, eg physicochemical, nutritional and molecular descriptors (and/or different types of traits). As discussed with reference to Figures 1 to 3, the target ingredient database 104 and the source ingredient database 102 may include the same types of traits, e.g., feature1, feature2, feature3, etc. (and/or different characteristics).

[0063] In step 1004, a screening of a particular target food item and source ingredient set may be created. Screening can represent each source ingredient in a D-dimensional space based on (e.g., what it contains) its respective set of characteristics, including physicochemical, nutritional, and/or molecular descriptors (e.g., where different source ingredients may have different values for the physical-chemical, nutritional and/or molecular characteristics of the descriptor, etc.). As discussed with reference to Figure 7, formula generator 106 may use screening process 702 to create a screening of the given target food item and set of source ingredients.

[0064] In an optional step 1006, a feature compression method for determining a more compact representation of the feature space can be performed. As discussed with reference to Figure 7, formula generator 106 can use feature compression template 706 to perform feature compression in order to reduce the size of the feature data set, if desired. Feature compression model 706 may use kernel principal component analysis (KPCA), autocoding, or other suitable method for feature compression.

[0065] In step 1008, a prediction model may be trained using the set of characteristics of the source ingredients to match those of the target food item based on a feature selection process. As discussed with reference to Figure 7, the formula generator 106 may use the prediction model 704 implemented using the gradient-gaining tree regression model. Prediction model 704 can be trained to match the hypothesis space of the plurality of source ingredients (used as data features) to a target. For example, gradient enhancement trees can be trained to fit a specific target, using available physicochemical, nutritional, or molecular characteristics.

[0066] Step 1010 may include selecting the most relevant source ingredients using the prediction model to be included in a formula for a recipe to mimic the target food item. The formula generator 106 can use the prediction model 704 to determine a formula phi, presenting a feature selection problem to the trained prediction model 704, where the features can be the source ingredients. The most relevant characteristics selected using prediction model 704 may be the potential set of source ingredients to be included in the formula.

[0067] Step 1012 may include determining the formula using an optimization process to determine a respective proportion of each of the most relevant source ingredients in the formula fi. Formula generator 106 may use optimizer 810 to perform an optimization process to determine specific proportions of the most relevant ingredients to generate formula phi.

[0068] Fig. 11 illustrates a computer-implemented method 1100 for determining a recipe comprising a cooking process for the formula generated by performing any suitable embodiments of the method 1000. The computer-implemented method 1100 may be performed using any suitable embodiments of the method. system 100 of figure 1.

[0069] In step 1102, a set of existing recipes can be obtained for training a deep RNN (and/or any suitable artificial intelligence model). Every existing recipe can include a list of ingredients and a set of instructions. As discussed with reference to FIG. 8, existing recipe collector 802 may obtain a set of existing recipes from online features. For example, the existing recipe set may include recipe set 400.

[0070] In step 1104, the set of existing recipes can be edited to modify each existing recipe to only include atomic directions. As discussed with reference to FIG. 5 for example recipe 500, ingredient list 502 and/or directions 504 can be modified using automatic and manual processing to include atomic instructions for the respective ingredients, as shown by cooking process 508 For example, atomic instructions may include "heat", "braise", "stir", "cook", etc. In some examples, the existing set of recipes may be stored in the existing recipe database 110.

[0071] In step 1106, one or more RNN models (and/or other suitable artificial intelligence models etc.) can be trained using the existing formula and/or set of recipes. The one or more RNN models (and/or other suitable artificial intelligence models, etc.) can provide a cooking process comprising a list of actions and/or respective ingredients used to generate one or more recipes to mimic the target food item . As discussed with reference to Figure 8, the training model 806 that implements one or more RNN models (and/or other suitable artificial intelligence models, etc.) can be trained using the formula fi generated by the formula generator 106 and the set of existing recipes modified by recipe processor 804. Training model 806 may provide a cooking process similar to cooking process 508 comprising a list of actions and respective ingredients used to generate one or more recipes to mimic the target food item.

[0072] In step 1108, one or more recipes can be generated using the trained RNN model(s) and/or other suitable artificial intelligence models, etc.) for each target food item. Each recipe can include a respective score indicating a difference between the fi formula and the target food item. As discussed with reference to FIG. 8, training template 806 can be used to generate one or more recipes. Each recipe can include a respective s(fi) score. The s(fi) score can indicate how similar the formula is to the target food item in the feature space. As an example, a lower score might indicate that the recipe is closer in the trait space to the target food item.

[0073] In step 1110, a recipe can be selected from one or more recipes based on the score. The selected recipe can be cooked or modified by a chef (and/or robot, smart kitchen appliance and/or any suitable entity). Afterwards, a sensory panel can taste the cooked product and give feedback on sensory descriptors. The one or more selected recipes can be provided to the formula 112 feedback panel for feedback. As discussed with reference to Figure 6, the one or more recipes can be prepared by a chef (and/or other suitable entity) and human feedback on sensory descriptors (e.g. color, taste, taste, palate, etc.) how visual appearances can be collected. In some instances, the one or more recipes may be modified based on feedback. In some cases, the feedback provided by the formula 112 feedback panel may include a modified recipe formula. The modified recipe formula, its preparation, one or more images, the sensory feedback and/or any other relevant information can be saved in the Chef 114 database.

[0074] One or more instances and/or parts of embodiments of the methods and/or processes described in this document can be executed asynchronously (for example, sequentially), simultaneously (for example, in parallel; simultaneously in different threads for parallel computation to improve system processing, etc.), in temporal relation to a triggering event (e.g., performance of a part of an embodiment of a method described herein) and/or in any other suitable order at any suitable time and frequency by and/or using one or more instances of system 100 embodiments, components and/or entities described herein.

[0075] As discussed with reference to Figures 1 to 11, the disclosed embodiments may utilize various machine learning algorithms and/or proprietary databases to generate recipes for certain target food items using different types of source ingredients. Certain embodiments may provide flexibility in generating recipes for any type of target food item (e.g., plant-based, animal-based, or synthetic) using any type of source ingredient (e.g., plant-based, animal-based, or synthetic). synthetic).

[0076] Portions of embodiments of methods and/or systems described herein are preferably performed by a first party, but may additionally or alternatively be performed by one or more third parties, users and/or any suitable entities.

[0077] Additionally or alternatively, the data described here can be associated with any suitable time indicators (e.g. seconds, minutes, hours, days, days, weeks, periods, time points, timestamps, etc.), including one or more: time indicators indicating when data was collected, determined (eg, output by a model described herein), transmitted, received and/or processed; temporal indicators that provide context to the content described by the data; changes in temporal indicators (eg, data over time; change in data; data patterns; data trends; data extrapolation and/or other prediction; etc.); and/or any other suitable time-related indicators.

[0078] Additionally or alternatively, parameters, metrics, inputs, outputs and/or other suitable data can be associated with value types, including any one or more of: scores (eg, revenue scores, etc.), text values (e.g., indicating ingredients, actions, etc.), numerical values (e.g., indicating ratios for ingredients; indicating aspects of atomic directions, etc.), binary values, classifications, confidence levels, identifiers, values along a spectrum, and /or any other suitable type of values. Any suitable types of data described herein can be used as inputs (eg, for different models described herein; for components of a system 100; etc.), generated as outputs (eg, of models; of components of a system 100; etc.) and/or manipulated in any manner suitable for any suitable components.

[0079] Additionally or alternatively, suitable portions of embodiments of methods and/or systems described in this document may include, apply, employ, perform, use, rely on and/or associate with one or more processing operations, including any one or more of: feature extraction, performing pattern recognition on data, merging data from multiple sources, combining values (e.g. mean values, etc.), compression, conversion (e.g. digital-to-analog conversion, analogue-to-digital conversion), performing statistical estimation of data (e.g., ordinary least squares regression, non-negative least squares regression, principal component analysis, ridge regression, etc.), normalization, updating, ranking, weighting, validation , filtering (e.g. for baseline correction, data clipping, etc.), noise reduction, smoothing, filling (e.g. gap filling), alignment, model fit, winding, windowing, cropping, transforms, mathematical operations (e.g., derivatives, moving averages, addition, subtraction, multiplication, division, etc.), data association, interpolation, extrapolation, grouping, image processing techniques, other signal processing operations, other signal processing operations image, preview and/or any other suitable processing operations.

[0080] System embodiments 100 and/or parts of system embodiments 100 may be fully or partially implemented by, hosted, communicated and/or otherwise include one or more: remote computing systems (for example, one or more servers , at least one network computing system, stateless, solemn; etc.), local computing systems, cell phone devices, other mobile devices, personal computing devices, tablets, databases, application programming interfaces ( APIs) (e.g. to access the data described here) etc.) and/or any suitable components. Communication by and/or between any component of system 100 and/or other suitable components may include wireless communication (e.g., WiFi, Bluetooth, radio frequency, Zigbee, Z wave, etc.), wired communication, and/or any other types of proper communication.

[0081] The embodiment components of the system 100 may be physically and/or logically integrated in any manner (e.g., with any suitable distributions of functionality between the components, as with respect to portions of the embodiments of the described methods.

[0082] Embodiments of the method 900, system 100 and/or any suitable systems and/or methods described in this document and/or variants thereof may include all combinations and permutations of the various components of the system and the various processes of the method, including any variants (e.g., embodiments, variations, examples, specific examples, figures, etc.), in which parts of method 900 and/or processes described herein can be performed asynchronously (e.g., sequentially), concurrently (e.g., in parallel) or in any other suitable order by and/or using one or more instances, elements, components of and/or other aspects of system 100 and/or other entities described herein.

[0083] Any of the variants described here (e.g. embodiments, variations, examples, specific examples, figures, etc.) and/or any part of the variants described here can be additionally or alternatively combined, added, deleted, used, performed in series, performed in parallel and/or otherwise applied.

[0084] The system 100, method 900, and/or any suitable systems and/or methods described in this document and/or variants thereof may be incorporated and/or implemented at least in part as a machine configured to receive a readable medium by computer that stores computer-readable instructions. Instructions can be executed by computer-executable components that can be integrated into the system. The computer-readable medium may be stored on any computer-readable media, such as RAMs, ROMs, flash memory, EEPROMs, optical devices (CD or DVD), hard disks, floppy disk drives or any suitable device. The computer-executable component can be a general or application-specific processor, but any dedicated hardware device or suitable hardware/firmware combination can alternatively or additionally execute the instructions.

[0085] As one skilled in the art will recognize from the foregoing detailed description and figures and Claims, modifications and changes may be made to system 100, method 900 and/or variants without departing from the scope defined in the following Claims.

Claims (19)

1. Computer-Implemented Method for generating a recipe using plant-based ingredients to mimic a specific food item of animal origin, characterized in that the method comprises: retrieving, from a target database, a set of characteristics of data associated with the target food item of animal origin, the set of data characteristics for the target food item of animal origin comprising at least one of: physicochemical data characteristics of the target food item of animal origin, nutritional data characteristics of the item target food of animal origin or molecular data characteristics of the target food item of animal origin; retrieving, from a source ingredient database, a respective set of data characteristics for each of the herbal source ingredients comprising at least one of: physicochemical data characteristics of the source ingredient to the base of plants, nutritional data characteristics of the herbal source ingredient, or molecular data characteristics of the plant source ingredient; perform a computer-implemented feature compression method on the respective data feature set for each of the plant-based source ingredients to determine the compact representation of the respective data feature set for each of the source-based ingredients of plants; creating a first training set comprising the compact representation of the respective feature set of data for each of the herbal ingredients; train a machine learning prediction model using the first training set to generate a feature space of plant-based ingredients; create a second training set based on the set of existing processed recipes stored in a database of existing recipes and processed to a specific data format to train a recurrent neural network (RNN), each existing recipe in the set of existing processed recipes including a ingredient list and a list of instructions for cooking the ingredients in the ingredient list; train the RNN using the second training set; use the trained machine learning prediction model to select matching plant-based ingredients from the plant-based ingredients that match the target animal-based food item in the feature space, where selecting the ingredients of plant origin is based on matching data traits from the respective data trait set for each of the herbal origin ingredients to the retrieved data trait set associated with the target food item of animal origin in the feature space; determining a formula and a corresponding similarity score, where the formula combines the corresponding plant-based ingredients in specific proportions, and the similarity score indicates how similar the formula is to the target animal-based food item in the trait space; and generating, from the formula and using the trained RNN, a recipe that uses the corresponding plant-based ingredients to mimic the target food item of animal origin, the recipe comprising a cooking process. 2. Computer-Implemented Method, according to Claim 1, characterized in that the set of data characteristics associated with the target food item of animal origin and the respective set of data characteristics for each of the ingredients of plant origin are associated to the same set of data trait types, and wherein the same set of data trait types comprises at least one of: a nutritional descriptor trait type, a physicochemical descriptor trait type, or a trait type molecular descriptor. 3. Computer-Implemented Method according to Claim 1, characterized in that the cooking process includes a list of actions to be carried out using the corresponding herbal ingredients. 4. Computer-Implemented Method, according to Claim 1, characterized in that the machine learning prediction model is based on gradient-increasing trees. 5. Computer-Implemented Method, according to Claim 1, characterized in that the method further comprises: representing each source ingredient of plant-based source ingredients in a D-dimensional space based on the respective identified data feature set for parent ingredient, representation of the parent ingredient in D-dimensional space comprising at least one of: physicochemical, nutritional or molecular descriptors of the parent ingredient. 6. Computer-Implemented Method, according to Claim 1, characterized in that determining the formula includes performing an optimization process to determine the specific proportions to combine corresponding herbal ingredients in the formula using minus absolute regression operator shrinking and sorting (Lasso). 7. Computer-Implemented Method, according to Claim 1, characterized in that the correspondence of the set of data characteristics retrieved indirectly corresponds to the imitation of at least one aroma, color, texture and flavor of the target food item of animal origin. 8. Computer-Implemented Method, according to Claim 1, characterized in that creating the second training set includes: obtaining a set of existing recipes, each existing recipe comprising a respective list of ingredients, respective amounts and respective instructions for cooking the respective list of ingredients; and getting the set of existing processed recipes based on modifying each of the existing recipes, in the existing recipe set, to include the respective atomic directions for cooking the respective ingredient list of the existing recipe. 9. Computer-Implemented Method, according to Claim 8, characterized in that modifying the set of existing recipes includes manually marking each of the respective instructions and the respective list of ingredients to include the respective atomic directions for cooking the respective list of ingredients. 10. Computer-Implemented Method, according to Claim 1, characterized in that a set of recipes comprising the recipe is generated, in which each recipe of the set of recipes comprises a respective cooking process for a respective formula and in which the method computer-implemented further comprises determining a similarity score for each recipe in the recipe set based on matching the target food item of animal origin. 11. Computer-Implemented Method according to Claim 10, characterized in that an individual follows the cooking process of the recipe to produce a cooked food item and feedback on the sensory descriptors of the cooked recipe is collected. 12. Computer-Implemented Method according to Claim 11, characterized in that the recipe is modified based on feedback and the modified recipe is stored in a chef's database. 13. Computer-Implemented Method, according to Claim 1, characterized in that creating the first training set based on the respective set of data characteristics for each of the ingredients of herbal origin is based on: executing a method computer-implemented feature compression on the respective set of trait data for each of the herbal source ingredients to determine the compact representation of the respective set of trait data for each of the herbal source ingredients; and including the compact representation of the respective data feature set for each of the herbal ingredients in the first training set. 14. Computer-implemented method according to Claim 1, characterized in that the trained recurrent neural network is a trained deep recurrent neural network. 15. Recipe Generating System, to mimic a specific target food item, characterized in that the system comprises: a source ingredient database configured to store a respective set of data characteristics for each source ingredient in source ingredients, wherein the respective set of data characteristics for each source ingredient comprises at least one of: source ingredient physicochemical data characteristics, source ingredient nutritional data characteristics, or source ingredient molecular data characteristics; a target database configured to store a respective set of data characteristics for each target food item in target food items, wherein the respective set of data characteristics for each target food item comprises at least one of: physical- target food item chemical, target food item nutritional data characteristics, or target food item molecular data characteristics; a trained machine learning prediction model that is trained using a first training set, the first training set generated based on a compact representation of the respective set of data characteristics for each source ingredient of the source ingredients; a recipe database configured to store a set of existing recipes, each existing recipe in the existing recipe set comprising a list of ingredients and a list of instructions for cooking the ingredients in the ingredient list; a trained recurrent neural network that is trained using the second training set, the second training set generated based on the existing recipe set; one or more non-transient computer-readable media; and computer-executable instructions stored on one or more non-transient computer-readable media, the computer-executable instructions, when executed by one or more computer processors, cause one or more computer processors to perform: use the trained model of machine learning prediction for selecting matching source ingredients from source ingredients, wherein selection of matching source ingredients is based on matching data features from the respective set of data features associated with a specific target food item of target food items; determine the formula that combines the corresponding source ingredients in specific proportions; and generating, from the formula and using the trained recurrent neural network, a recipe that uses the corresponding source ingredients to mimic the specific target food item, the recipe comprising the cooking process. 16. System to Generate Revenue, according to Claim 15, characterized in that it further comprises: computer-executable instructions stored on one or more non-transient computer-readable media that, when executed by one or more computer processors, cause one or more computer processors represent each source ingredient of the source ingredients in a D-dimensional space based on the respective data feature set associated with the source ingredient, the representation of the source ingredient in the D-dimensional space comprising at least one of: physical-chemical, nutritional or molecular descriptors of the source ingredient. 17. System to Generate Revenue, according to Claim 15, characterized in that it further comprises: computer-executable instructions stored on one or more non-transient computer-readable media that, when executed by one or more computer processors, cause one or more computer processors generate the second training set based on the existing recipe set based on: obtaining the existing recipe set from the recipe database; generate a set of existing processed recipes based on modifying each of the existing recipes, in the existing recipe set, to include the respective atomic direction for cooking ingredients list of the existing recipe; and include the set of existing processed recipes in the second training set. 18. System to Generate Revenue, according to Claim 15, characterized in that it further comprises: computer-executable instructions stored on one or more non-transient computer-readable media that, when executed by one or more computer processors, cause one or more computer processors create the first training set based on the respective data feature set for each of the source ingredients is based on: performing a computer-implemented feature compression method on the respective data feature set to each of the source ingredients to determine the compact representation of the respective set of data characteristics for each of the source ingredients; and include the compact representation of the respective set of data features for each of the source ingredients in the first training set. 19. Revenue Generating System, according to Claim 15, characterized in that the trained recurrent neural network is a trained deep recurrent neural network.