Systems and methods for translating abstract and creative inputs into quantitative organoleptic profiles

Abstract

A computer-implemented method for generating a quantitative target organoleptic profile for a fermented comestible receives at least one creative input and processes the creative input to generate a geometric representation defined by at least one geometric parameter derived according to a mapping scheme. Using at least one processor, a correspondence is determined between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship. Based on the determined correspondence, the quantitative target organoleptic profile is generated, the profile comprising at least one quantitative parameter representing a sensory attribute. The method enables translation of creative inputs including musical, visual, textual, or thematic data into machine-readable specifications suitable for guiding fermentation-based product development.

Description

SYSTEMS AND METHODS FOR TRANSLATING ABSTRACT AND CREATIVE INPUTS INTO QUANTITATIVE ORGANOLEPTIC PROFILESCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 63 / 740,953, titled "SYSTEM AND METHOD FOR OPTIMIZING THEMATIC BEVERAGE FERMENTATION PROCESSES BASED ON ARTIFICIAL INTELLIGENCE MACHINE LEARNING MODELS," filed December 31, 2024, and to U.S. Provisional Application No.63 / 740,962, titled "SYSTEM AND METHOD FOR CREATING MELODY -BASED BEVERAGE FERMENTATION PROCESSES BASED ON ARTIFICIAL INTELLIGENCE MACHINE LEARNING MODELS," filed December 31, 2024 each of which is hereby incorporated by reference in its entirety. This application is related to PCT Application No. PCT / IB2025 / 059740, titled SYSTEM AND METHOD FOR PRESCRIBING FERMENTATION BASED ON MACHINE LEARNING MODELS, filed September 27, 2025, which is hereby incorporated by reference in its entirety.FIELD OF INVENTION

[0002] The present disclosure relates to computer-implemented systems and methods for generating target organoleptic profiles for fermented comestibles, and more particularly to translating creative inputs into geometric representations to generate quantitative target organoleptic profiles.BACKGROUND

[0003] The development of fermented comestibles, including beverages and other consumable products, typically involves the selection and combination of precursor substrates, microbial strains, and fermentation conditions to achieve desired sensory characteristics such as aroma, flavor, and mouthfeel. These sensory characteristics are often described collectively as an organoleptic profile. The development of fermented comestibles, including beverages and food products, has traditionally relied upon empirical approaches that involve extensive trial-and-error experimentation to achieve desired sensory characteristics. Fermentation processes involve complex biochemical transformations wherein microorganisms convert substrates into a diverse array of metabolites, including organic acids, esters, alcohols, andother flavor-associated compounds that collectively define the organoleptic profile of the final product.

[0004] In parallel with developments in fermentation science, opportunity exists for creating personalized and experiential food and beverage products that resonate with consumers on emotional or thematic levels. Creative inputs such as music, visual art, brand identities, and event themes can be explored as inspiration for product development in various consumer goods industries. The translation of such abstract creative concepts into concrete product formulations presents a distinct challenge, as it requires bridging qualitative, subjective experiences with quantitative, measurable product attributes.

[0005] The field of computational chemistry and cheminformatics has developed various methods for representing and analyzing molecular structures, including techniques for characterizing the three-dimensional geometry and topology of chemical compounds. Molecular descriptors such as principal moments of inertia, surface topology metrics, and pharmacophore arrangements provide quantitative characterizations of molecular shape and spatial properties. These computational representations have found application in drug discovery, materials science, and other domains where structure -property relationships are of interest.

[0006] Existing systems for recipe generation and fermentation optimization have demonstrated the utility of machine learning models in predicting metabolite production and organoleptic outcomes based on ingredient and process parameters. However, these systems typically operate on inputs that are already expressed in terms of chemical or process variables, rather than abstract creative concepts. The challenge of translating non -technical, creative inputs into the quantitative specifications required by such systems represents a gap in current methodologies. Consequently, there remains an opportunity to develop approaches that can systematically convert multi-modal creative inputs into actionable target profiles suitable for guiding fermentation-based product development.SUMMARY

[0007] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0008] According to an aspect of the present disclosure, a computer-implemented method for generating a quantitative target organoleptic profile for a fermented comestible is provided. The method involves receiving at least one creative input. The at least one creative input is processed to generate a geometric representation, where the geometric representation is defined by at least one geometric parameter derived according to a mapping scheme. Using at least one processor, a correspondence is determined between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship. Based on the determined correspondence, the quantitative target organoleptic profile is generated, the profile including at least one quantitative parameter representing a sensory attribute.

[0009] According to another aspect of the present disclosure, a system for generating a quantitative target organoleptic profile for a fermented comestible is provided. The system includes one or more processors and a non -transitory computer-readable memory storing instructions that, when executed by the one or more processors, cause the system to receive at least one creative input. The instructions further cause the system to process the at least one creative input to generate a geometric representation, where the geometric representation is defined by at least one geometric parameter derived according to a mapping scheme. The system determines, using the one or more processors, a correspondence between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship. Based on the determined correspondence, the system generates the quantitative target organoleptic profile, which includes at least one quantitative parameter representing a sensory attribute.

[0010] A non-transitory computer-readable medium is also disclosed. The medium stores instructions that, when executed by one or more processors, cause the one or more processors to perform operations. The operations include receiving at least one creative input and processing the at least one creative input to generate a geometric representation defined by at least one geometric parameter derived according to a mapping scheme. A correspondence between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation is determined based on a geometric relationship. Based on the determined correspondence, a quantitative target organoleptic profile is generated that includes at least one quantitative parameter representing a sensory attribute.

[0011] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive.BRIEF DESCRIPTION OF FIGURES

[0012] Non-limiting and non-exhaustive examples are described with reference to the following figures.

[0013] FIG. 1 is a flow diagram illustrating a method for generating a quantitative target organoleptic profile for a fermented comestible, according to an aspect of the present disclosure.

[0014] FIG. 2 is a flow diagram illustrating a method for generating a quantitative target organoleptic profile with palatability validation and novelty discovery routing, according to an aspect of the present disclosure.

[0015] FIG. 3 is a block diagram illustrating a system for generating a quantitative target organoleptic profile for a fermented comestible, according to an aspect of the present disclosure.

[0016] FIG. 4A is a block diagram illustrating creative input modalities and feature extraction, according to an aspect of the present disclosure.

[0017] FIG. 4B is a block diagram illustrating split-path processing for textual and thematic inputs, showing structural / phonetic feature extraction for geometric mapping and semantic / contextual feature extraction for profile refinement, according to an aspect of the present disclosure.

[0018] FIG. 5A is a block diagram illustrating determination of correspondence between a geometric representation and molecular structure representations, according to an aspect of the present disclosure.

[0019] FIG. 5B is a block diagram illustrating the geometric representation as a universal translation layer receiving inputs from musical, visual, and textual input modalities, according to an aspect of the present disclosure.

[0020] FIG. 6 is a flow diagram illustrating provision of a quantitative target organoleptic profile to a recipe-generation process, according to an aspect of the present disclosure.DETAILED DESCRIPTION

[0021] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary aspects described herein.

[0022] The present disclosure describes a computer-implemented method 100 for generating a quantitative target organoleptic profile 108 for a fermented comestible. Referring to FIG. 1, the method 100 receives at least one creative input 102 from a user or from an external system. The creative input 102 may comprise one or more modalities (e.g., musical, visual, textual / thematic, or multi-modal combinations among others), as described further below with reference to FIG. 4. Upon receiving the creative input 102, the method 100 processes the creative input 102 to generate a geometric representation 104. The geometric representation 104 is defined by at least one geometric parameter determined according to the mapping scheme 236. The mapping scheme 236 establishes a computational relationship between features of the creative input 102 and parameters that define the geometric representation 104.

[0023] In some embodiments, a quantitative target organoleptic profile that does not satisfy a validation criterion but satisfies a predefined novelty criterion is computationally designated as a novel candidate flavor profile and routed to a novelty discovery process distinct from validation of established flavor combinations.

[0024] With continued reference to FIG. 1, the method 100 determines using at least one processor (described further with reference to FIG. 3), a correspondence 106 between the geometric representation 104 and at least one molecular structure representation 120 corresponding to flavor-associated compounds producible via fermentation. The correspondence is established 106 based on a geometric relationship that associates geometric features of the geometric representation 104 with structural features of the molecular structure representations 120. Based on the determined correspondence 106, the method 100 generates the quantitative target organoleptic profile 108. The quantitative target organoleptic profile 108 comprises at least one quantitative parameter representing a sensory attribute of the fermented comestible, such as flavor, aroma, texture, appearance among others. In this manner, the method 100 maps creative inputs 102 to a machine -readable, quantitative target organoleptic profile 108 suitable for downstream computational use, including as an input to recipegeneration or fermentation-control systems.

[0025] In some embodiments, the method 100 further comprises computationally evaluating palatability 110 of candidate components associated with the quantitative target organoleptic profile 108 and, based on one or more configurable outcomes of the palatability evaluation 110, selectively routing the quantitative target organoleptic profile 108 for validation (validated profile 112) or novelty processing (novelty discovery path 114). In some embodiments, the method 100 further comprises providing the quantitative target organoleptic profile 108 in a machine-readable format as an input to a downstream or external recipegeneration process 130 that generates a recipe 132 specifying ingredients, fermentation agents, and process parameters. In some embodiments, the quantitative target organoleptic profile 108 may serve as a target phenotype input for an inverse-design fermentation system, such as that described in the related PCT application incorporated by reference herein.

[0026] Creative Inputs

[0027] Referring to FIG. 1 and FIG. 4A, the method 100 receives at least one creative input 102. The creative input 102 may comprise any data modality suitable for computational processing, including musical inputs, visual inputs, textual or thematic inputs, or multi-modal combinations thereof. In some embodiments, the creative input 102 comprises a musical input 102a, which may be provided in an audio format or in a symbolic representation such as MIDI orMusicXML. In other embodiments, the creative input 102 comprises a visual input 102b, such as a digital image. In further embodiments, the creative input 102 comprises textual data or a thematic description represented in a machine-interpretable format, referred to herein as a textual or thematic input 102c. Multi-modal creative inputs may comprise two or more of the foregoing input types provided together.

[0028] In some embodiments, the method 100 comprises extracting one or more features 105 from the creative input 102 using one or more feature extraction processes, such as a feature extraction engine 103. The particular features extracted, and the techniques used to extract them, may vary depending on the modality of the creative input 102 and are implementation-dependent. In some embodiments, the method employs a split-path processing architecture in which the creative input 102 is processed along two or more parallel computational paths that extract different categories of features, such as structural, temporal, perceptual, semantic, or contextual features, with outputs from the respective paths applied at different stages of geometric representation generation, correspondence determination, or profile refinement. As illustrated in FIG. 4B, extracted features 105 may be separated into structural or phonetic features 105a and semantic or contextual features 105b, wherein thestructural or phonetic features 105a are routed to geometric mapping and the semantic or contextual features 105b are routed to a refinement step applied to the quantitative target organoleptic profile.

[0029] The split-path processing architecture illustrated in FIG. 4B is not limited to a particular creative input modality. In some embodiments, the split-path architecture is applicable to musical inputs 102a, visual inputs 102b, textual or thematic inputs 102c, or multimodal combinations thereof. In such embodiments, features encoding structure, form, rhythm, geometry, or perceptual pattern are routed to geometric mapping, while features encoding meaning, affect, intent, or contextual nuance are routed to a refinement step performed after correspondence determination.

[0030] For example, for visual inputs 102b, the feature extraction process may extract visual features relating to color, shape, texture, or spatial composition of the visual input. Non-limiting examples of visual features include color distributions, dominant color values, color harmony relationships, edge density, contour curvature, geometric shape prevalence, symmetry, or spatial arrangement of visual elements. In some embodiments, computer vision, image processing, and pattern recognition techniques may be employed to detect edges, contours, or regions of interest, and to derive feature representations characterizing the visual structure of the input, as will be appreciated by those skilled in the art.

[0031] In another example, for textual or thematic inputs 102c, the feature extraction process may extract semantic, affective, or conceptual features from the textual or thematic data. Non-limiting examples of such features include semantic embeddings, sentiment or affect scores, thematic categories, conceptual clusters, or vectors representing mood, intensity, or descriptive attributes expressed in the input. In some embodiments, the feature extraction process may additionally or alternatively extract structural and phonetic features 105a from textual inputs, including cadence, phoneme patterns, syllabic structure, and sound symbolism characteristics. For example, textual inputs containing sharp consonants (such as "k," "t," or "p" sounds) may be associated with angular geometric primitives, while textual inputs containing round vowels (such as "o" or "u" sounds) may be associated with curved geometric primitives. This phonetic-to-geometric mapping provides one example by which textual inputs such as poems, brand names, or thematic phrases may be converted into geometric representations suitable for correspondence determination with molecular structures. Semantic / contextual features 105b extracted from textual inputs may be applied in a profile refinement step as described below. In some embodiments, Natural Language Processingtechniques may be employed to convert textual or thematic inputs into machine -interpretable feature representations, such as numerical vectors in a semantic embedding space, suitable for subsequent processing by the mapping scheme 236, as will be appreciated by those skilled in the art.

[0032] In yet another example, where the creative input 102 comprises a musical input 102a, the feature extraction engine 103 may extract one or more musical features characterizing rhythmic, harmonic, melodic, or spectral aspects of the musical input, including, by way of non-limiting example, tempo, chord progression, interval relationships, timbre, or spectral characteristics. Such features may include time-domain features (e.g., amplitude envelope, zero-crossing rate) and frequency-domain features (e.g., spectrograms or Mel-Frequency Cepstral Coefficients (MFCCs), using known digital signal processing techniques, as will be appreciated by those skilled in the art.

[0033] Feature extraction is optional. In some embodiments, the creative input 102 may be processed directly by the mapping scheme 236 without an explicit feature extraction step. The optional, modality -agnostic, and split-path nature of feature extraction enables the method 100 to accommodate a wide range of creative inputs 102 while preserving flexibility in how extracted information influences geometric mapping and subsequent refinement, without departing from the scope of the present disclosure.

[0034] Geometric Representation and Mapping Scheme

[0035] Next, the method 100 processes the at least one creative input 102 to generate a geometric representation 104. The geometric representation 104 is defined by at least one geometric parameter derived according to a mapping scheme 236 and serves as an intermediate computational structure encoding information from the creative input 102 in a form suitable for correspondence determination with molecular structure representations 120.

[0036] The geometric representation 104 may take a variety of forms depending on the embodiment and the nature of the creative input 102. In some embodiments, the geometric representation 104 comprises one or more geometric primitives, including, by way of non-limiting example, triangles, polygons, curves, or other abstract shapes that encode relationships derived from the creative input 102. In other embodiments, the geometric representation 104 comprises geometric parameters expressed as numerical values, vectors, matrices, or tensors characterizing spatial, angular, dimensional, or relational properties.

[0037] In further embodiments, the geometric representation 104 comprises a multidimensional geometric space, wherein the creative input 102 is mapped to coordinates, regions, or trajectories within a space defined by multiple geometric dimensions. Such a multidimensional geometric space may be static or dynamic, may evolve over time, and may comprise a learned or defined manifold, embedding space, or topological space. In still further embodiments, the geometric representation 104 comprises a learned geometric embedding generated by a machine-learning model trained to represent creative inputs in a geometric or vector space.

[0038] As used herein, the term “geometric” includes abstract mathematical and representational spaces and does not require any physical or visual instantiation. The geometric representation 104 may exist entirely as a computational construct, including high-dimensional vector spaces or topological representations in which distances, neighborhoods, or relational properties encode similarities or structural relationships among creative inputs, as will be appreciated by those skilled in the art.

[0039] The mapping scheme 236 comprises a computational function that maps the creative input 102, or features 105 extracted therefrom, to geometric parameters defining the geometric representation 104. The mapping scheme 236 may establish a deterministic, probabilistic, learned, or hybrid relationship between inputs and output geometric parameters. In some embodiments, the mapping scheme 236 comprises a rule-based mapping that applies predefined rules or logical conditions encoding domain knowledge to transform creative input features into geometric parameters. In other embodiments, the mapping scheme 236 comprises a machine-learned mapping implemented using a trained model that has learned associations between creative input features and geometric representations from training data. In further embodiments, the mapping scheme 236 comprises an optimization -based mapping that determines geometric parameters by solving an optimization problem subject to one or more objectives or constraints (e.g., minimizing a distance metric or maximizing a similarity measure). In still further embodiments, the mapping scheme 236 comprises a similarity -based mapping that retrieves, interpolates, or infers geometric parameters based on similarity relationships between the creative input 102 and reference examples stored in a database or knowledge base. Hybrid mapping schemes may combine two or more of the foregoing approaches.

[0040] In some embodiments, the mapping scheme 236 comprises a machine -learned model trained on a dataset comprising pairs of creative input features and correspondinggeometric representations. Suitable model architectures may include, by way of non-limiting example, neural networks, decision trees, support vector machines, or ensemble models. The particular architecture, training process, and parameterization of the mapping scheme 236 are implementation-dependent and do not limit the scope of the present disclosure.

[0041] In embodiments where the creative input 102 comprises a musical input 102a, the method 100 may further comprise extracting one or more musical features from the musical input 102a. The musical input 102a may be provided in an audio format, a symbolic format, or a hybrid representation, and the extracted musical features may characterize rhythmic, harmonic, melodic, spectral, or temporal aspects of the musical input.

[0042] In some embodiments, the method 100 may further comprise decomposing the musical input 102a into constituent structural components prior to or during feature extraction. For example, the musical input 102a may be decomposed into (i) a base component corresponding to harmonic or rhythmic structure and (ii) one or more signature components corresponding to melodic, thematic, or expressive elements. Different structural components may be mapped to different aspects of the geometric representation 104, enabling distinct musical characteristics to influence different portions of the resulting quantitative target organoleptic profile 108.

[0043] The mapping scheme 236 may include a music-theoretic mapping that converts extracted musical features into geometric parameters. In some embodiments, the music -theoretic mapping employs established theoretical frameworks that represent musical relationships in geometric or spatial form, including, by way of non-limiting example, tonal or interval-based geometric models. Additionally or alternatively, the mapping scheme 236 may incorporate cross-modal correspondence relationships, including psychoacoustic or perceptual associations between auditory characteristics and other sensory modalities, such that certain sound characteristics influence parameters associated with sharp, bright, acidic sensory attributes while other sound characteristics influence parameters associated with round, smooth, creamy sensory attributes.

[0044] Molecular Structure Representations

[0045] The method 100 determines a correspondence between the geometric representation 104 and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation. Molecular structure representations may comprise two-dimensional molecular representations, three-dimensional molecularconformations, graph-based molecular representations, learned molecular embeddings, or a combination thereof. Two-dimensional molecular representations may encode molecular connectivity and bonding patterns in a planar format. Three-dimensional molecular conformations capture the spatial arrangement of atoms within a molecule, including low-energy conformers that represent stable molecular geometries. Graph-based molecular representations encode molecules as nodes and edges representing atoms and bonds, respectively, enabling computational analysis of molecular topology. Learned molecular embeddings position molecules within a vector space based on structural or functional similarities learned from training data, as will be appreciated by those skilled in the art.

[0046] In some embodiments, molecular structure representations are characterized by at least one molecular geometric descriptor. The at least one molecular geometric descriptor may comprise at least one of principal moments of inertia coordinates, molecular surface topology, or pharmacophore point spatial arrangement. The molecular structure representations correspond to flavor-associated compounds producible via fermentation, including, by way of non-limiting example, volatile organic compounds, esters, terpenes, aldehydes, ketones, lactones, phenols, or functional metabolites that contribute to sensory attributes of fermented comestibles. In some embodiments, a database links chemical compounds to geometric descriptors, enabling retrieval of molecular structure representations based on geometric criteria, and cheminformatics toolkits may be employed to generate conformers and calculate descriptors used in correspondence determination.

[0047] Correspondence Determination

[0048] Referring to FIG. 5A, determining the correspondence comprises associating at least one geometric feature of the geometric representation with at least one structural feature of the molecular structure representation to define the geometric relationship. The correspondence may comprise a one-to-one association, a one-to-many association, a probabilistic association, a ranked association, an association defined in a representation space, or a combination thereof. In some embodiments, correspondences are determined using embedding-based associations, distance-based associations, manifold-based associations, or topological associations. For example, in some embodiments the geometric representation and molecular structure representations are positioned within a shared representation space and correspondences are determined based on proximity under one or more distance or similarity metrics. In other embodiments, correspondences are determined based on shared topological properties between the geometric representation and molecular structure representations.

[0049] For example, in some embodiments, the correspondence may be determined by computing molecular geometric descriptors for compounds stored in a molecular structure database 234 and identifying molecular structures whose descriptors satisfy criteria associated with the geometric representation 104. In one illustrative example, a triangular geometric primitive derived from a musical interval of a third may be matched to molecular structures having disk-like principal moments of inertia coordinates, such as planar esters. In another illustrative example, a hexagonal or cyclic geometric primitive derived from a musical scale or visual swirl may be matched to molecular structures having ring -based geometries, such as terpenes or lactones. In a further illustrative example, a starburst or spiky geometric primitive derived from high-frequency sounds or angular visual elements may be matched to molecular structures having rod-like or linear geometries, such as aldehydes. These examples are illustrative and do not limit the correspondence determination to any particular geometric primitives, molecular structures, or matching criteria.

[0050] By way of further illustration, the following examples demonstrate how different creative inputs may be processed through the geometric mapping framework to generate candidate correspondences with molecular structures. In a first example involving a pop music input, a pop song characterized by staccato rhythms and bright major-key harmonies may generate angular and symmetric geometric primitives. These geometric primitives may be associated with molecular structures such as short-chain aldehydes and fruity esters, which are associated with bright, crisp sensory attributes in the resulting quantitative target organoleptic profile.

[0051] In a second example involving a meditative musical composition, a healing or meditative composition featuring slow tempos, triplet rhythms, and consonant intervals may generate smooth, curved geometric primitives. These geometric primitives may be associated with molecular structure representations corresponding to calming or low-reactivity metaoblites, such as compounds related to GABA precursors or L-theanine pathways, enabling the generation of a quantitative target organoleptic profile oriented toward relaxation or wellness applications.

[0052] In a third example involving a visual art input, a visual artwork such as a painting characterized by swirling brushstrokes and high color saturation may generate spiral and cyclic geometric primitives. These geometric primitives may be associated with molecular structure representations such as terpenes and lactones, which are associated with complex, textured sensory attributes including bitterness and aromatic depth.

[0053] FIG. 5B illustrates how the geometric representation serves as a universal translation layer, receiving geometric features derived from musical, visual, and textual inputs and enabling correspondence determination with molecular structure representations across all input modalities.

[0054] As an example, in embodiments employing a complementary relationship, geometric features are matched to molecular structures that complement rather than mirror the geometric characteristics. Inverted relationships may associate geometric features with molecular structures having opposing or inverse geometric properties. Transformed relationships may apply mathematical transformations to geometric features before matching to molecular structures. Proximity-based relationships may identify molecular structures whose geometric descriptors fall within a defined proximity threshold of the geometric features of the geometric representation. Topological relationships may match geometric representations to molecular structures based on shared topological properties such as connectivity patterns or neighborhood relationships that are preserved under continuous transformations.

[0055] Referring to FIG. 4B, in some embodiments involving textual or thematic inputs, the method 100 employs a split-path processing approach. In such embodiments, a first processing path extracts structural and phonetic features 105a from the textual input, which are mapped to geometric primitives for correspondence determination with molecular structures as described above. A second processing path extracts semantic and contextual features 105b representing the meaning or conceptual content of the textual input. The semantic features 105b from the second path may be applied in a profile refinement step to adjust quantitative parameters of the target organoleptic profile after the initial geometric correspondence is determined. For example, semantic features associated with concepts such as "joy" or "calm" may be used to bias, weight, or adjust quantitative parameters of the target organoleptic profile, such as relative proportions or emphasis among candidate flavor-associated compounds, while preserving the geometric mapping framework as the primary mechanism for correspondence determination.

[0056] Quantitative Target Organoleptic Profile

[0057] Based on the determined correspondence, the method 100 generates the quantitative target organoleptic profile 108 comprising at least one quantitative parameter representing a sensory attribute. The quantitative target organoleptic profile 108 provides acomputational specification that characterizes the sensory properties of a fermented comestible in numerical or parametric form and is machine-readable, enabling storage, transmission, and downstream computational processing.

[0058] The quantitative target organoleptic profile 108 may be represented in various nonlimiting computational forms, including a multidimensional sensory space, a vector representation, or a parameter set associating named sensory attributes with corresponding quantitative values. The quantitative parameters within the quantitative target organoleptic profile 108 derive from the correspondence determined between the geometric representation 104 and the molecular structure representations 120, and may reflect estimated or relative concentrations, ratios, or other measurable characteristics associated with flavor-relevant compounds or derived sensory attributes.

[0059] Palatability Evaluation and Novelty Routing

[0060] In some embodiments, the method 100 may further comprise computationally evaluating palatability 110 of candidate components associated with the quantitative target organoleptic profile 108. Palatability may be evaluated using a structured knowledge representation 238 encoding at least one relationship type among flavor-relevant components. The structured knowledge representation 238 may comprise a flavor harmony database, a knowledge base, or a knowledge graph wherein nodes represent molecular compounds or ingredient classes and edges represent relationships between those components.

[0061] In some embodiments, a quantitative target organoleptic profile that does not satisfy a validation criterion but satisfies a predefined novelty criterion is computationally designated as a novel candidate flavor profile and routed to a novelty discovery process distinct from validation of established flavor combinations.

[0062] The at least one relationship type encoded within the structured knowledge representation 238 may comprise compatibility relationships, antagonistic relationships, synergistic relationships, co-occurrence relationships, or a combination thereof. Compatibility relationships indicate that two or more flavor-relevant components are likely to produce a harmonious sensory experience when combined. Antagonistic relationships indicate that certain component combinations may produce undesirable sensory interactions or clashes. Synergistic relationships indicate that certain combinations may enhance or amplify sensory attributes beyond what would be expected from the individual components alone. Co-occurrence relationships indicate that certain components are frequently found together in successful formulations, providing empirical evidence of compatibility.

[0063] Palatability may be computed as a palatability score determined from pairwise or multi-component relationships among flavor-relevant components, for example using one or more aggregation functions applied to relationship weights. Based on the palatability score, the method 100 may selectively determine a downstream processing path, including designating the quantitative target organoleptic profile 108 as a validated target profile 112 or routing the quantitative target organoleptic profile 108 to a novelty discovery process 114. Thresholds, scoring functions, and criteria are configurable and implementation-dependent.

[0064] System

[0065] Referring now to FIG. 3, a system 200 for generating a quantitative target organoleptic profile 108 for a fermented comestible is described. The system 200 comprises one or more processors 222 and a non -transitory computer-readable memory 226 storing instructions 232. When executed by the one or more processors 222, the instructions 232 cause the system 200 to computationally perform operations for generating the quantitative target organoleptic profile 108 from creative inputs 102.

[0066] The system 200 includes a processing unit 220 that houses the one or more processors 222. The processors may include general-purpose processors, specialized accelerators, or combinations thereof, and may interface with one or more input / output devices.

[0067] A storage 230 within the system 200 stores the instructions 232 along with data structures that support the operations of the system 200. The storage 230 may contain a molecular structure database 234 that stores molecular structure representations 120 corresponding to flavor-associated compounds producible via fermentation. The storage 230 may also contain mapping scheme data 236 that defines the computational relationships used to transform creative inputs 102 into geometric representations 104. Additionally, the storage 230 may contain a knowledge representation 238 encoding relationships among flavor-relevant components for use in palatability evaluation 110.

[0068] The instructions cause the system to receive at least one creative input, generate the geometric representation according to the mapping scheme, determine the correspondence to molecular structure representations based on a geometric relationship, and generate the quantitative target organoleptic profile comprising at least one quantitative parameter representing a sensory attribute. In some embodiments, the system 200 provides thequantitative target organoleptic profile 108 in a machine-readable format as an input to a downstream recipe-generation process 130.

[0069] The present disclosure also encompasses a non-transitory computer-readable medium storing instructions 232 that, when executed by one or more processors 222, cause the system to perform the operations described above with reference to method 100, including receiving the creative input 102, generating the geometric representation 104 according to the mapping scheme 236, determining the correspondence 106 to molecular structure representations 120 based on a geometric relationship, and generating the quantitative target organoleptic profile 108.

[0070] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims.

Claims

CLAIMS1. A computer-implemented method for generating a quantitative target organoleptic profile for a fermented comestible, the method comprising:(a) receiving at least one creative input;(b) processing the at least one creative input to generate a geometric representation, the geometric representation defined by at least one geometric parameter derived according to a mapping scheme;(c) determining, using at least one processor, a correspondence between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship; and (d) generating, based on the determined correspondence, the quantitative target organoleptic profile comprising at least one quantitative parameter representing a sensory attribute.

2. The method of claim 1 , wherein the at least one creative input comprises at least one of: a musical input, visual image, textual data, a thematic description represented in a machine-interpretable format, or a multi-modal combination thereof.

3. The method of claim 1 , wherein the geometric representation comprises at least one of geometric primitives, geometric parameters, a multidimensional geometric space, or a learned geometric embedding.

4. The method of claim 3, wherein the multidimensional geometric space comprises at least one of a temporal geometric representation, a dynamic geometric representation, a manifold, or a topological space.

5. The method of claim 1, wherein the mapping scheme comprises a computational function that maps features extracted from the at least one creative input to parameters of the geometric representation.

6. The method of claim 1, wherein the mapping scheme comprises at least one of a rule-based mapping, a machine-learned mapping, an optimization-based mapping, a similarity -based mapping, or a hybrid thereof.

7. The method of claim 6, wherein the mapping scheme comprises a machine-learned model trained to associate features of the creative input with geometric features.

8. The method of claim 2, wherein the at least one creative input comprises a musical input, and wherein the method further comprises extracting at least one musical feature from the musical input, the at least one musical feature comprising at least one of: tempo, key signature, chord progression, interval set, spectral centroid, or rhythmic density.

9. The method of claim 8, wherein the mapping scheme comprises a music-theoretic mapping that converts the at least one musical feature into at least one geometric parameter, the at least one geometric parameter defining a geometric primitive.

10. The method of claim 1, wherein the molecular structure representations comprise two-dimensional molecular representations, three-dimensional molecular conformations, graph-based molecular representations, learned molecular embeddings, or a combination thereof.

11. The method of claim 10, wherein the molecular structure representations are characterized by at least one molecular geometric descriptor comprising at least one of principal moments of inertia coordinates, molecular surface topology, or pharmacophore point spatial arrangement.

12. The method of claim 1, wherein determining the correspondence comprises associating at least one geometric feature of the geometric representation with at least one structural feature of the molecular structure representation to define the geometric relationship.

13. The method of claim 12, wherein the correspondence comprises one or more of a similarity relationship, a complementary relationship, an inverted relationship, a transformed relationship, a proximity -based relationship, or a topological relationship.

14. The method of claim 12, wherein the correspondence comprises a one-to-one association, a one-to-many association, a probabilistic association, a ranked association, an association defined in a representation space, or a combination thereof.

15. The method of claim 14, wherein the association defined in the representation space comprises at least one of: an embedding-based association, a distance-based association, a manifold-based association, or topological association, or a combination thereof.

16. The method of claim 1, further comprising computationally evaluating palatability of candidate components associated with the quantitative target organoleptic profile.

17. The method of claim 16, wherein palatability is evaluated using a structured knowledge representation encoding at least one relationship type among flavor -relevant components.

18. The method of claim 17, wherein the at least one relationship type comprises compatibility relationships, antagonistic relationships, synergistic relationships, co-occurrence relationships, or a combination thereof.

19. The method of claim 17, wherein the structured knowledge representation comprises a flavor harmony database or knowledge base encoding compatibility relationships among flavor-relevant components.

20. The method of claim 19, wherein palatability is computed based on a palatability score determined from pairwise or multi-component compatibility relationships among flavor-relevant components.

21. The method of claim 20, further comprising selectively determining a downstream processing path based on the palatability score.

22. The method of claim 21, wherein selectively determining the downstream processing path comprises: (i) when the palatability score satisfies at least one validation criterion, designating the quantitative target organoleptic profile as a validated target profile; and (ii) when the palatability score satisfies at least one novelty criterion, routing the quantitative target organoleptic profile to a novelty discovery process.

23. The method of claim 1, further comprising providing the quantitative target organoleptic profile, in a machine-readable format, as an input to a recipe-generation process for producing a fermented comestible.

24. A system for generating a quantitative target organoleptic profile for a fermented comestible, the system comprising:- one or more processors; and- a non-transitory computer-readable memory storing instructions that, when executed by the one or more processors, cause the system to:(a) receive at least one creative input;(b) process the at least one creative input to generate a geometric representation, the geometric representation defined by at least one geometric parameter derived according to a mapping scheme;(c) determine, using the one or more processors, a correspondence between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship; and(d) generate, based on the determined correspondence, the quantitative target organoleptic profile comprising at least one quantitative parameter representing a sensory attribute.

25. The system of claim 24, wherein the instructions further cause the system to provide the quantitative target organoleptic profile, in a machine-readable format, as an input to a recipe-generation process for producing a fermented comestible.

26. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:(a) receiving at least one creative input;(b) processing the at least one creative input to generate a geometric representation, the geometric representation defined by at least one geometric parameter derived according to a mapping scheme;(c) determining a correspondence between the geometric representation and at least one molecular structure representation corresponding to flavor-associated compounds producible via fermentation, based on a geometric relationship; and(d) generating, based on the determined correspondence, a quantitative target organoleptic profile comprising at least one quantitative parameter representing a sensory attribute.

27. The non-transitory computer-readable medium of claim 26, wherein the at least one creative input comprises a musical input, and wherein the operations further comprise extracting at least one musical feature from the musical input.

28. The method of claim 1, wherein wherein the geometric representation comprises one or more vectors in a multidimensional embedding space, wherein the mapping scheme maps features of the creative input to corresponding vector parameters.

29. The method of claim 1, further comprising:- computationally evaluating a palatability score associated with the quantitative target organoleptic profile based on relationships among flavor-associated compounds; and- designating the quantitative target organoleptic profile as a novel candidate profile when the palatability score satisfies a novelty criterion that is distinct from a validation criterion.

30. The method of claim 29, further comprising routing the quantitative target organoleptic profile to a novelty discovery process when the palatability score falls within a predefined novelty range bounded by an acceptance threshold and a rejection threshold.

31. The system of claim 24, wherein the instructions further cause the one or more processors to:- compute a palatability score for a quantitative target organoleptic profile based on compatibility relationships among flavor-associated compounds; and- identify the quantitative target organoleptic profile as a novel flavor profile when the palatability score satisfies a novelty criterion distinct from a validation criterion.

32. The method of claim 1 , wherein determining the quantitative target organoleptic profile comprises:(a) generating an initial quantitative target organoleptic profile based on geometric correspondence between structural or phonetic features of the creative input and molecular structure representations; and(b) applying a profile refinement step in which semantic or contextual features of the creative input are used to adjust one or more quantitative parameters of the initial quantitative target organoleptic profile.

33. The method of claim 32, wherein the semantic or contextual features comprise affective, emotional, or conceptual attributes extracted from the creative input.

Images