Wine classification and evaluation method

Abstract

To quickly provide wine classification / evaluation means based only on wine ingredients without referring to price information and origin / brand information for wine classification.SOLUTION: It has been confirmed that a wide variety of wines can be properly classified by performing clustering processing for each of basic tastes, such as umami, sweetness, sourness, and bitterness, and then combining evaluations of clusters for the basic tastes.SELECTED DRAWING: None

Description

Technical Field

[0001] The present invention relates to a method for classifying and evaluating wine and a system for classifying and evaluating wine.

Background Art

[0002] The components that make up wine are diverse, such as amino acids, organic acids, nucleosides, nucleotides, etc. With the progress of analytical techniques in recent years, various analytical methods have been proposed. For example, a method for analyzing wine by food metabolomics using an analyzer that enables simultaneous analysis of hydrophilic metabolite components by ion-pair free LC / MS / MS method has been reported (for example, triple quadrupole mass spectrometer LCMS TM -8060NX brochure, manufactured by Shimadzu Corporation). However, in wine, it is said that the precisely analyzed component composition is not directly reflected in the taste of the wine. Also, gas chromatography has been used to measure the aroma components in wine, but it is also known that components with high peak intensity (alcohols, fatty acids) do not necessarily have a large contribution to the senses (see, for example, Non-Patent Document 1).

[0003] As component analysis techniques for wines in recent years, for example, there has been proposed a mobile terminal (see, for example, Patent Document 1) used in a wine taste evaluation system characterized by comprising control means configured to match the taste data of a user with the reference taste data regarding the taste of the wine stored in the wine information memory, quantify the degree of matching between the taste data of the user regarding the wine and the reference taste data regarding the taste of the wine, and display the result on a display. However, since the taste data of the user is the object of matching, there is a problem that it is difficult to objectively evaluate the wine itself. Also, a wine probe and a method for measuring an amount that is a characteristic of a wine from one such probe (see, for example, Patent Document 2) have been proposed, and measurements encoded in an RGB framework can be expressed in a framework called hue, saturation, lightness (HSL) and / or CIELAB, and provide data such as the degree of aging, body or acidity to the user, but do not present an evaluation regarding the taste of the wine itself. Further, in a computer system that proposes wines to be drunk with food (see, for example, Patent Document 3), "taste" evaluates the wine from a plurality of viewpoints regarding the characteristics of the wine such as "sweetness", "saltiness", "acidity", "astringency", "aging feeling", etc., but there is no specific teaching regarding how each of the above tastes is separated and evaluated.

[0004] Also, in recent years, while wine evaluation can be performed with sophisticated component analysis including the above metabolomics, there is an accumulation of traditional evaluations such as origin and brand, and price determination is often linked to such traditional evaluations.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Patent Document 3

Non-Patent Literature

[0006]

Non-Patent Literature 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] The problem of the present invention is to quickly provide a classification and evaluation method for wine classification that, without referring to price information, origin / brand information, evaluates only the contained components as the basis, and highly matches the evaluation of those who actually taste wine.

Means for Solving the Problems

[0008] The inventors of the present invention have continued to study means for classifying and evaluating the taste of wine. First, using the recently developed metabolomics technology, all components constituting wine such as amino acids, organic acids, nucleosides, and nucleotides were targeted for evaluation, and clustering was attempted. However, common taste characteristics could not be found in the wine groups belonging to each cluster generated by clustering. Therefore, components constituting each taste generally referred to as basic tastes such as umami, sweetness, sourness, and bitterness were selected, and clustering processing was performed for each basic taste. As a result, it was found that classification could be made that roughly matches the evaluation of the taste felt by those who are familiar with the taste of so-called wine. Furthermore, by combining evaluations such as calculating the total value of the evaluations (values) of the cluster groups for each of the above basic tastes, it was confirmed that various types of wine could be appropriately classified. Such classification was confirmed to correspond to the enhancement of taste due to the synergistic effect between the components in the wine and the components acting on the taste contained in the dish in a scene where a dish such as meat or fish is actually served with wine, and thus the present invention was completed.

[0009] That is, the present invention is as follows. [1] A wine classification method sequentially comprising the following (a) to (f). (a) A data acquisition step of acquiring component profile data of a reference wine; (b) A quantification step of quantifying the content of each component constituting the component profile data of the reference wine acquired in the data acquisition step of step (a) as a characteristic value; (c) A data extraction step of extracting the characteristic values of the components constituting the basic taste from the characteristic values of each component of the reference wine created in the quantification step of step (b); (d) A clustering step of forming n cluster groups for each basic taste by performing clustering by unsupervised learning for each characteristic value of the components constituting each basic taste extracted in the extraction step of step (c); (e) A group classification step of classifying the reference wines based on the cluster groups for each basic taste formed in step (d); (f) A comment creation step of associating taste information and creating a comment; [2] The classification method according to [1] above, characterized in that the basic taste consists of umami and sweetness. [3] The classification method according to [1] above, characterized in that the basic taste consists of umami, sweetness, and sourness. [4] The classification method according to [1] above, characterized in that the basic taste consists of umami, sweetness, sourness, and bitterness. [5] The classification method according to any one of [2] to [4] above, characterized in that the component(s) constituting umami contains one or more selected from glutamic acid, inosinic acid, aspartic acid, adenylic acid, and succinic acid. [6] The classification method according to any one of [2] to [4] above, characterized in that the components constituting umami are glutamic acid and succinic acid. [7] The classification method according to any one of [2] to [4] above, characterized in that the component(s) constituting sweetness contains one or more selected from proline, glycine, alanine, threonine, serine, ornithine, and asparagine. [8] The classification method according to any one of [2] to [4] above, characterized in that the components constituting the umami are proline and alanine. [9] The classification method according to [3] or [4] above, characterized in that the components constituting the sour taste include one or more selected from tartaric acid, malic acid, citric acid, succinic acid, lactic acid, glycyl-glutamine acetate, seryl-glutamine, and glycyl-asparagine.

[10] The classification method according to [3] or [4] above, characterized in that the components constituting the sour taste are malic acid, citric acid, succinic acid, and lactic acid.

[0010] Further, the present invention is as follows.

[11] A wine classification system that outputs a comment on wine, including a classification result using wine component profile data having the following (A) to (F). (A) A data acquisition unit that acquires data of component profile data of reference wine; (B) A quantification unit that quantifies the content of each component constituting the component profile data of the reference wine acquired by the data acquisition unit as a characteristic value; (C) A data extraction unit that extracts quantified data of components constituting the basic taste from the quantified profile data; (D) A clustering unit that forms n clusters for each basic taste by performing clustering by unsupervised learning for each quantified data of components constituting each basic taste; (E) A group classification unit that performs group classification of reference wines based on the n clusters formed for each basic taste; (F) A comment creation unit that associates taste information and creates a comment;

[12] The classification system according to

[11] above, characterized in that the basic taste consists of umami and sweetness.

[13] The classification system according to

[11] above, characterized in that the basic taste consists of umami, sweetness, and sourness.

[14] The classification system according to

[11] above, characterized in that the basic taste consists of umami, sweetness, sourness, and bitterness.

[15] The classification system according to any one of

[12] to

[14] above, characterized in that the components constituting umami include one or more selected from glutamic acid, inosinic acid, aspartic acid, adenylic acid, and succinic acid.

[16] The classification system according to any one of

[12] to

[14] above, characterized in that the components constituting umami are glutamic acid and succinic acid.

[17] The classification system according to any one of

[12] to

[14] above, characterized in that the components constituting sweetness include one or more selected from proline, glycine, alanine, threonine, serine, ornithine, and asparagine.

[18] The classification system according to any one of

[12] to

[14] above, characterized in that the components constituting sweetness are proline and alanine.

[19] The classification system according to

[13] or

[14] above, characterized in that the components constituting sourness include one or more selected from tartaric acid, malic acid, citric acid, succinic acid, lactic acid, glycyl - glutamine acetate, seryl - glutamine, and glycyl - asparagine.

[20] The classification system according to

[13] or

[14] above, characterized in that the components constituting sourness are malic acid, citric acid, succinic acid, and lactic acid.

[21] A method for collecting wine classification data, which uses the component profile data of wine presented by a user as input data, classifies it according to the wine classification method according to any one of [1] to [4] above, and records the result as reference data.

Advantages of the Invention

[0011] According to the present invention, wine can be classified without being affected by price, brand, or circulation volume.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Figure 24

Mode for Carrying Out the Invention

[0013] As the present invention, there is no particular limitation as long as it is a wine classification method including: (a) a data acquisition step of acquiring component profile data of a reference wine; (b) a quantification step of quantifying the content of each component constituting the component profile data of the reference wine acquired in the data acquisition step of step (a) as a characteristic value; (c) a data extraction step of extracting the characteristic values of the components constituting the basic taste from the characteristic values of each component of the reference wine created in the quantification step of step (b); (d) a clustering step of forming n cluster groups for each basic taste by performing clustering by unsupervised learning for each characteristic value of the components constituting each basic taste extracted in the extraction step of step (c); and (e) a comment creation step of associating taste information with each group clustered for each basic taste and creating a comment.

[0014] Further, FIG. 8 is a diagram showing an example of the configuration of a classification system 1 according to an embodiment of the wine classification method of the present invention. The classification system 1 is not particularly limited as long as it is a classification system including a data acquisition unit 10, a quantification unit 20, a data extraction unit 30, a clustering unit 40, a group classification unit 50, and a comment creation unit 60. If necessary, it can further include a component determination unit for determining components constituting the basic taste from the quantified profile data.

[0015] The data acquisition step of (a) above is not particularly limited as long as it is a step of acquiring component profile data of a reference wine. The reference wine can be any wine that can be selected from all wines available on the market. The number of reference wines is not particularly limited as long as it is a number of wines that can be classified into n clusters, but examples include 50 or more types, preferably 100 or more types, more preferably 200 or more types, still more preferably 1000 or more types, and even more preferably 2000 or more types of wines. The maximum number can be all types of wines that may be available on the market.

[0016] In the data acquisition step of (a) above, as a method for acquiring the component profile data of the reference wine, known component analysis methods can be cited. As the component detection unit 5 capable of detecting and analyzing the components constituting each reference wine, gas chromatography / mass spectrometry (GC / MS), liquid chromatography / mass spectrometry (LC / MS), capillary electrophoresis / mass spectrometry (CE / MS), etc., mass spectrometry methods such as MALDI-TOF-MS, GC-TOF / MS analysis, metabolome analysis (metabolomics analysis), etc. Equipment for performing known chemical methods can be cited. Specifically, commercially available metabolomics analysis software Profiler TM AM + , UPLC (registered trademark) / Synapt for metabonomics TM HDMS TM System (manufactured by Waters Japan Ltd.), triple quadrupole mass spectrometer LCMS TM -8050, 8060NX, GCMSNX series, GCMS-QPTM2020, GCMS-TQTM series, GCMS-QP2010 SE (all manufactured by Shimadzu Corporation), in the solid-phase microextraction (SPME) method, Pegasus(R) HT GC-TOFMS (manufactured by LECO Corporation), Agilent 1290 Infinity II LC system (manufactured by Agilent Technologies), etc. can be used. The data acquisition unit 10 of the present invention can acquire data on profile data of the components constituting the wine using the so-called external component detection and analysis system described above. In addition, by using self-developed software or commercially available multivariate analysis software, it can be directly acquired by the data acquisition unit 10 including the component detection unit 5 that detects and analyzes the components constituting the reference wine.

[0017] The components included in the component profile data of the reference wine are not particularly limited as long as they are components constituting each reference wine. However, amino acids such as histidine, arginine, valine, methionine, tryptophan, phenylalanine, isoleucine, leucine, lysine, cysteine, tyrosine, proline, glycine, alanine, threonine, serine, asparagine, aspartic acid, glutamic acid, ornithine, etc.; organic acids such as succinic acid, citric acid, malic acid, lactic acid, tartaric acid, gluconic acid, glyceric acid, threonic acid, etc.; nucleic acids and nucleobases such as inosinic acid, adenyl acid, hypoxanthine, adenine, guanine, thymine, cytosine, uracil, purine, pyrimidine, etc.; vitamins such as nicotinic acid, pantothenic acid, nicotinamide, riboflavin, 4-aminobenzoic acid, etc.; pigments such as anthocyanin, tannin, etc.; saccharides such as glucose, fructose, etc., dipeptides such as glycyl-glutamine, alanyl-glutamine, etc. can be exemplified. However, it is desirable to include components constituting each basic taste in the present invention.

[0018] Such component profile data of the reference wine can be stored in the component profile data storage unit 71 after being obtained through processes such as detection, analysis, and identification in the data acquisition unit 10. Also, such stored data can be output to the quantification unit 20 to extract characteristic values of the components constituting the basic taste.

[0019] In addition, in order to obtain the profile data of the wine, the necessary pre-treatment of the wine can be performed for each of the plurality of reference wines as needed in the pre-treatment step (a 0 ’). Specifically, for example, before applying the reference wine to LCMS, processes such as passing it through an ultrafiltration filter to remove starch, sediment, etc., or diluting it with water can be mentioned. After performing such processes in the pre-treatment unit 4, processes such as detection, analysis, and identification can also be performed.

[0020] As the quantification process in (b) above, there is no particular limitation as long as it is a process of quantifying the content of each component constituting the component profile data of the reference wine as a characteristic value. Examples thereof include a process of setting the peak height of a chromatogram specific to each component as a relative quantification value, and a process of treating the concentration specific to metabolites (compounds) such as each amino acid as an absolute value. However, since the peak height measured by a single chromatogram measurement may vary for each measurement, a process of measuring the specific concentration of each metabolite as a control value and converting the concentration from the peak value is preferably mentioned. The above characteristic value (peak value) is not particularly limited as long as it is a numerical value representing the characteristics of each component, and examples thereof include 1) an identity characteristic value, 2) a Mean characteristic value, 3) a Max characteristic value, 4) a Min characteristic value, etc.

[0021] Also, for example, for the peaks measured by LC-MS, the height or area of the peaks can be quantified by known software, and the quantified values can be used as peak values. Alternatively, for the content (concentration) of each component, it can also be converted to a corresponding value by creating a calibration curve.

[0022] When using the above identity characteristic value, the identity characteristic value indicates a characteristic value indicating the content itself.

[0023] The process of quantifying the content of each component constituting the component profile data of the reference wine obtained in the above data acquisition unit as a characteristic value can be executed in the quantification unit 20. The peak height of a chromatogram specific to each component can be quantified as a relative quantification value, and the concentration of each component in each wine specific to metabolites (compounds) such as each amino acid can be quantified. Further, the data regarding the characteristic value of the reference wine obtained as a result of the process can also be stored in the characteristic value data storage unit 72, and information can also be presented in the information presentation unit 80.

[0024] The extraction process in (c) is not particularly limited as long as it is an extraction process for extracting the characteristic values of the components that make up the basic taste from the characteristic values of each component of the reference wine created in the quantification process in (b). The basic taste is generally defined as sweetness, umami, bitterness, sourness, and saltiness, which are received by taste cells and are defined as tastes that can be clearly distinguished from each other. However, the basic tastes in the present invention can include umami, sweetness, sourness, bitterness, and unpleasant taste, and can also be limited to umami, sweetness, sourness, and bitterness, or umami, sweetness, sourness, and unpleasant taste, or umami, sweetness, and sourness. However, it is necessary to include at least umami and sweetness.

[0025] The above extraction process (c) for extracting the characteristic values of the components performed in the data extraction unit 30 can further include a process (c0) for determining the components of each basic taste to be used in the subsequent clustering process and the like for each of the above basic tastes. As a method for determining the components that make up such a basic taste, there is a method of picking up the components known from literature and the like that make up each basic taste. For example, when there is a strong correlation (multicollinearity) between the variables of a plurality of components, when the correlation coefficient of the peak or concentration data is equal to or greater than a predetermined value, the effect can also be predicted by other components that are strongly correlated with the variable. Therefore, since there is a possibility that the importance of the variables inherent in a specific component cannot be appropriately reflected, the data of one or two or more components can be deleted, and the information constituted by the remaining data can be used as the quantification data of the components that make up each basic taste, and can also be used as the input data for generating the learning model.

[0026] Regarding the presence or absence of the above-mentioned multicollinearity, in the component determination unit 35, the examined content can be confirmed by graphically displaying the correlation coefficients between the components or by examining the content confirmed by displaying the correlation matrix consisting of the respective correlation coefficients in the form of a heat map, and the components to be used for clustering, etc. can be determined. When determining the components having the variables to be used for classification, for example, even if the content of the components in wine is below the threshold in taste, the knowledge regarding the taste enhanced by the synergistic effect of, for example, glutamic acid and inosinic acid, and / or succinic acid and inosinic acid, etc. due to the marriage when ingested with food can be further taken into consideration. By means of such processing, the data of the components with overlapping features are unified, so that the accuracy when the learning model is generated is increased. The creation of the above-mentioned correlation matrix and heat map can be performed using, for example, publicly known software such as the Python library, the pandas library, matplotlib, and seaborn. The graph displaying the correlation coefficients between the components and the heat map displaying the correlation matrix consisting of the correlation coefficients can be displayed on the display 810, etc. in the information presentation unit 80.

[0027] As a method for determining the components constituting each taste by using the above-mentioned heat map, for example, if the correlation coefficient of each component on the heat map is 0.7 or less, 0.5 or less, preferably 0.4 or less, and more preferably 0.3 or less, it can be picked up as the component constituting each taste. Also, even if the correlation coefficient between specific two components is 0.7 or more, 0.5 or more, 0.4 or more, or 0.3 or more, when the correlation relationship with a plurality of other components on the heat map satisfies, for example, 0.3 or less, it can also be picked up as the component constituting each taste.

[0028] The display in the above heatmap format and the graph in which the correlation coefficients are graphed can also be displayed on the display or the like of the information presentation unit 80. For example, the display 810 may graphically display the correlation coefficients, or may display the correlation matrix composed of the respective correlation coefficients in the heatmap format. The memory 700 of the data storage unit can also store findings regarding the correlation coefficients and the thresholds that humans feel for the components constituting each taste, and by referring to such data, the operator can also determine the components constituting each taste.

[0029] As described above, for the data of one peak, when duplicate data information is deleted and the remaining peak data is used to form the input data, unnecessary peak (feature) data is deleted, so the peak information and concentration information indicating the results of mass spectrometry for each wine sample, and the characteristic information indicating predetermined characteristics such as the taste of the sample are used. The accuracy of the generated learning model is improved.

[0030] The above umami is one of the basic tastes that has been confirmed to be perceivable since a receptor that reacts to glutamic acid was confirmed in the sensory cells in the taste buds of the tongue in 2002. Examples of the components constituting umami include one or more selected from the group consisting of glutamic acid, inosinic acid, aspartic acid, adenilic acid, succinic acid, glutathione, histidine, carnosine, and lactic acid, but it is preferable to select at least glutamic acid and succinic acid.

[0031] Examples of the components that make up the sweetness include proline, glycine, alanine, threonine, serine, ornithine, glycerol, etc. Yeast in grape juice consumes sugars such as glucose and fructose contained in grapes through fermentation. When the alcohol content reaches about 15% and the sugars are consumed, the yeast is considered to die. Therefore, in sweet wines and ultra-sweet wines, there may be residual sugars such as glucose and fructose. However, as components that exhibit sweetness common to all wines including dry wines, proline, glycine, alanine, threonine, and serine are selected, and the evaluation of sweetness can be carried out based on their contents and content ratios. In addition, after defining the sweetness containing one or more of the above proline, glycine, alanine, threonine, and serine as sweetness by amino acids, as the basic taste in the present invention, in addition to the above (amino acid) sweetness, sweetness having sugars such as glucose and fructose as constituent components is further defined as saccharide sweetness, and such saccharide sweetness can be added to the basic taste to implement the wine classification method of the present invention. However, saccharide sweetness can be included or not included. It is preferable to set alanine and proline as components that make up the sweetness and implement the wine classification method of the present invention.

[0032] Examples of the components that make up the sourness in the present invention include acids derived from grape fruits such as tartaric acid, malic acid, and citric acid, acids generated by yeast fermentation during the brewing process such as succinic acid, lactic acid, and acetic acid, and dipeptides having taste properties such as glycyl-glutamine, seryl-glutamine, and glycyl-aspartic acid. However, as components that commonly exhibit sourness in all wines, the evaluation of sourness can be carried out based on the contents and content ratios of tartaric acid, citric acid, malic acid, lactic acid, gluconic acid, and glycyl-glutamine. However, it is preferable to set malic acid, lactic acid, citric acid, and succinic acid as components that make up the sourness and implement the wine classification method of the present invention.

[0033] Among the components that make up the above acidity, malic acid and lactic acid can also be used as indicators for judging the presence or absence of malolactic fermentation (MLF) in wine. Clusters with a high content of malic acid and a low content of lactic acid belong to the clusters of wines in which malolactic fermentation has not occurred. Clusters with a high content of lactic acid and a low content of malic acid can be judged to belong to the clusters of wines in which malolactic fermentation has occurred. The presence or absence of malolactic fermentation can be included in the comments created by the comment creation section.

[0034] The above malolactic fermentation is a fermentation carried out after the main fermentation (alcohol fermentation) by yeast in the wine production process, by the fermentation of lactic acid bacteria, to convert malic acid in the wine into lactic acid, suppress the acidity of the wine, and give it a mellow flavor. The lactic acid bacteria that carry out the above malolactic fermentation may be lactic acid bacteria attached to grapes or the like, or lactic acid bacteria artificially added.

[0035] Examples of the components that make up the above astringency include histidine, arginine, valine, methionine, tryptophan, phenylalanine, isoleucine, leucine, lysine, cysteine, tyrosine, nicotinic acid, hypoxanthine, pantothenic acid, nicotinamide, riboflavin, 4-aminobenzoic acid, anthocyanins, catechins (such as tannins), magnesium salts, calcium salts, etc. Although there are literature descriptions that tannins contribute to the astringency of wine, rather than tannins that give a characteristic astringency to red wines and some white wines, selecting the above components as components that exhibit an astringency common to all wines and evaluating the astringency based on their content and content ratio are the characteristics of the present invention. In addition, it is preferable to set proline, histidine, tryptophan, arginine, valine, and phenylalanine as components that make up the astringency and implement the wine classification method of the present invention. Briefly, the presence or absence of bitterness can also be determined by the amount of arginine.

[0036] The above basic taste can further include unpleasant components. Examples of unpleasant components include bioamines such as histamine, putrescine, cadaverine, cadabaline, beta-phenylethylamine, tryptamine, spermine, and spermidine. Also, 2-aminoethanol, guaiacol, ethylphenol, vinylphenol, ethyl acetate, acetaldehyde, acetone, sulfur dioxide, hydrogen sulfide, mercaptan, methionol, etc., which are known as off-flavors of wine, and 2,4,6-trichloroanisole (TCA). When drinking wine that contains more bioamine components than other standard wines, it may cause headaches, asthma, or allergic symptoms. Components known as the above off-flavors, except for TCA, may be evaluated as adding depth to the taste if present in small amounts.

[0037] In the data extraction step of step (c) above, a process of extracting characteristic values of components constituting the basic taste from the characteristic values of each component of the reference wine created in the quantification step of step (b) is performed. Such a process can be executed in the data extraction unit 30 that extracts quantification data of components constituting the basic taste from the quantification profile data. Also, data regarding the characteristic values of each component constituting the component profile data of the reference wine, obtained as a result of the process, can be stored in the characteristic value data storage unit 72, and information can also be presented in the information presentation unit 80.

[0038] The clustering step of step (d) above is not particularly limited as long as it is a process of creating n clusters for each basic taste by performing clustering by unsupervised learning for each characteristic value of the components constituting each basic taste extracted in the extraction step of step (c). Clustering is performed by unsupervised learning based on characteristic values for each basic taste such as umami, sweetness, sourness, and / or astringency, unpleasant taste, etc.

[0039] The above-mentioned unsupervised learning is a method that, by providing a large amount of input data to a learning system, learns the distribution of the input data and can perform compression, classification, shaping, etc. on the input data without providing teacher data corresponding to the input data. Since the above-mentioned unsupervised learning does not require a correct answer, it is used in many fields. However, instead of having a correct answer, it may be difficult to interpret the results, or completely different results may be obtained due to a slight difference in the initial parameters, and it is not always possible to reach the optimal solution. Sometimes, it may reach an intermediate solution (local solution) and the learning may end there. Therefore, trials can also be conducted with different parameters multiple times.

[0040] The above-mentioned clustering belongs to the above-mentioned unsupervised learning and is a method of classification by grouping a group of data whose correct answer is not known in advance into one group (cluster) with similar properties. Specifically, the k-means method, which is a method of obtaining a center for each cluster using the algorithm of non-hierarchical clustering and classifying each data into the cluster with the closest cluster center, can be exemplified.

[0041] Since the above-mentioned k-means method is an algorithm for creating an arbitrarily specified k number of clusters, it is a method named the k-point average method (k-means). For example, first, the data is divided into appropriate clusters, and then the sum of squared errors within the clusters (SSE) is repeatedly minimized to adjust the data to be better separated.

[0042] The above clustering is an exploratory data analysis method. Although the division can be said to be based on some subjectivity or perspective, methods such as the elbow method and the silhouette method can be used as methods for automatically estimating the number of clusters (n). The above elbow method is, for example, a method of trying the K-means method, etc. with various numbers of clusters, exploring the point where the accuracy does not increase much even if the number of clusters is further increased, and specifying the number of clusters obtained as a result of the exploration as the optimal number of clusters. It is a method of plotting the sum of squared errors (SSE) for each cluster and regarding the number of clusters at the point where the graph showing the change in SSE bends like an "elbow" as the optimal number of clusters.

[0043] Regarding the results obtained by the above clustering, it is also possible to determine whether each cluster formed by clustering is appropriate as necessary. As a method for determining whether each cluster is appropriate, an example is a method of setting a model wine and selecting a classification in which clustering is performed so that each model wine belongs to a different cluster. The selected clustering is prevented or suppressed from falling into a local solution and can be determined to be an appropriate clustering. Such a determination can also be made by an operator.

[0044] Examples of the above model wine include setting a wine with a determined evaluation, a wine having a characteristic taste and / or different tastes from each other. For example, a plurality of people proficient in the taste of wine can identify two or more, preferably three or more, more preferably four or more, and even more preferably n or more wines having a characteristic taste of different types. Examples of the people proficient in the taste of the above wine include sommeliers, wine producers, wine traders, wine enthusiasts, etc. can be given.

[0045] The above-mentioned n number is not particularly limited and can be determined according to the number of reference wines. For example, when the number of reference wines is 50 to 500, n = 3 or 4; when the number of reference wines is 100 to 1000, n = 3 or 4 or 5; when the number of reference wines is 1000 to 5000, n = 4 or 5; when the number of reference wines is 3000 to 15000, 4 or 5 or 6; when the number of reference wines is 5000 to 30000, 5 or 6; when it is 10000 to 100000, 6 or 7 or 8 can be cited. However, due to the bias of the origin to which the reference wines belong and the diversity of wine types, the above numerical values can be increased or decreased.

[0046] The data of each basic taste cluster generated by the above clustering can also be stored by the cluster data storage unit 73, and a graph such as a bar graph or a pie chart can be created to make the composition of the components in each cluster visually clear, and the screen can be further configured to be displayed on the above display 810.

[0047] The process of forming n clusters for each basic taste by performing clustering by unsupervised learning for each characteristic value of the components constituting each basic taste can be executed in the clustering unit 40. Also, based on the characteristics of the amount and blending ratio of each component in the cluster obtained as a result of the process, the information indicating the characteristics of the cluster obtained by interpreting the characteristics for each cluster can be stored in the cluster data storage unit 73. However, by an information presentation device integrated with or separate from the cluster data storage unit 73, the n clusters for each basic taste can also be presented with profile data information indicating the characteristics of the cluster in the form of a graph or the like.

[0048] Examples of the information stored in the characteristic value data storage unit 72 include the center points of the components and blending amounts constituting the cluster in each cluster, the range covered by the components and blending constituting the cluster in the cluster, and the explanatory comments about the basic taste in the cluster to which each wine belongs.

[0049] Further, by visually displaying the result of the above clustering in a graph, the composition ratio of each component can be immediately recognized visually. For example, a bar graph is created by stacking the characteristic values of the components in each cluster for each basic taste, and it is possible to determine whether the clustering is appropriate based on this graph. Such a graph can also be displayed on the display 810 in the above information presentation device.

[0050] The above step (e) is not particularly limited as long as it is a step of classifying the reference wine based on the clustering generated in step (d). As a classification method, a method of classifying the reference wine by combining the labeling numbers assigned to each of the above clusters can be mentioned. As the above labeling method, for example, from the larger total value of the characteristic values of the components constituting each basic taste, it can be labeled as n, n - 1, ··· 2, 1. For example, when the number of clusters is 3, it can be labeled as 3, 2, 1. Note that the numerical values of the above labeling can also be determined by the operator of the classification system, but for example, the number G of the cluster groups can also be used.

[0051] When the number of reference wines is small, for example, when it is 50 or less, 100 or less, 150 or less, 200 or less, etc., in each basic taste, a plurality of clusters can be grouped together.

[0052] In the present invention, the wine group classification step of step (e) may or may not be performed after the clustering step of (d) above. However, it is a step in which group classification is performed using the labeling in each basic taste for which the above clustering has been performed. It is not necessarily required to perform group classification based on the labeling of clusters using all of umami, sweetness, sourness, bitterness, and unpleasant taste. It is preferable to perform group classification based on the combination of numerical values of labeling based on at least umami and sweetness. It is more preferable to perform group classification based on umami, sweetness, and sourness. It is even more preferable to perform group classification based on umami, sweetness, sourness, and bitterness. Also, group classification can be performed based on umami, sweetness, sourness, bitterness, and unpleasant taste.

[0053] As a specific method of the above group classification, data on the labeling of each cluster of umami and sweetness; data on the labeling of each cluster of umami, sweetness, and sourness; data on the labeling of each cluster of umami, sweetness, sourness, and bitterness; or data on umami, sweetness, sourness, bitterness, and unpleasant components are integrated, sorted by value in Excel, and arranged and classified in order. This can be cited as the simplest method. Such group classification can be performed in the group classification unit 50, and the information on group classification can also be stored in the group classification data storage unit 74.

[0054] As the comment creation step of step (f), there is no particular limitation as long as it is a step of creating a comment by associating the characteristics of each cluster group for which clustering has been performed and the taste information of each group for which group classification has been performed for each taste in the wine for which the comment is to be created. The comment can also be created in the comment creation unit 60, and the created comment can also be stored in the comment data storage unit 75. In addition, price information about each wine can also be added.

[0055] Also, for each of the labeled clusters, factor analysis can be performed to determine in what respects the reference samples classified within the same cluster are similar. Such factor analysis is not particularly limited as long as it is a known factor analysis method. For example, for the classified clusters, the commonality of the samples belonging to each cluster can be analyzed by human intuition or a statistical analysis method based on human intuition, and for example, an evaluation comment can be created by AI.

[0056] Also, the evaluation comments created in the comment creation step can be stored in the comment data storage unit 75, and explanations can be presented via the comment creation unit regarding the overall evaluation of each wine.

[0057] In the wine classification method of the present invention, using the data related to the clustering for each basic taste and / or the data of the grouping of reference wines performed based on the cluster groups for each basic taste as teacher data, a prediction model is constructed using a random forest, a support vector machine (SVM) which is one of the pattern recognition models using supervised learning, and / or a neural network. Using the component profile data of a new wine presented by the user as input data, based on the component data of the new wine, according to the above wine classification method, clustering is performed, and based on the result, it is possible to predict to which wine group or group the new wine belongs, thereby performing an evaluation.

[0058] Also, the teacher data can be constructed as a feedback-enabled (recursively inputtable) regression-type machine learning model. For example, the predicted component profile data and classification results of the new wine are accompanied by the creation of comments, and additionally recorded as data of new reference wines, whereby data of new reference wines in the present invention can be created. As a result, by generating machine learning data that improves the classification accuracy, more evolved comprehensive wine data can be collected.

[0059] Hereinafter, the present invention will be described more specifically by way of examples. However, the technical scope of the present invention is not limited to these examples.

Example

[0060] [Reference Example 1] Component analysis of 53 types of reference wines was performed using a triple quadrupole mass spectrometer LCMS TM -8050 (manufactured by Shimadzu Corporation) and LC / MS / MS method package. Quantification was performed by cell culture profiling, and profile data for each wine was created. Classification was performed using unsupervised machine learning with all the components analyzed. Hierarchical cluster analysis was performed, but the classification was complex and no clear features could be grasped (Figure 1). Since the number of samples and component items were also large, the k-means method, which is a non-hierarchical clustering method, was used to instruct classification into 6 clusters, and clustering was performed. The results are shown in the graph of Figure 2.

[0061] Note that the above 53 types of reference wines were prepared by removing the coagulated matter in the sample with an ultrafiltration filter and diluting it 20 times with water, and then analyzed using LCMS TM -8050. When the above reference wines were stored at 4°C in a sealed container, no significant difference was observed in the detection results depending on the measurement time. On the other hand, when analyzing those left in the bottle and stored at room temperature, a difference was observed in the measured values depending on the measurement time. Therefore, a wine stored at 4°C was used, which was filled to the brim in a 15 mL polypropylene tube and sealed with parafilm.

[0062] As is clear from Figure 2, the average values of the analysis values of each component for 53 types of reference wines were calculated and visualized in a stacked graph, but no classification that could grasp the characteristics of the wines was obtained. The vertical axis of the graph represents the value obtained by numerically converting the peak value of each component analyzed by the above LCMS and expressing it in natural logarithm (ln).

[0063] [Example 1] Next, from the profile data of the reference wines, the components that make up the umami, sweetness, acidity, and bitterness of each wine were picked up, and clustering was performed by unsupervised learning for each of umami, sweetness, acidity, and bitterness.

[0064] From the components constituting the profile data for each of the 53 types of reference wines created in Reference Example 1 above, the quantified values of the components that make up the basic tastes of umami, sweetness, astringency, and acidity were extracted.

[0065] As the components that make up the umami, glutamic acid, aspartic acid, and succinic acid were picked up. As the components that make up the sweetness, threonine, alanine, proline, serine, ornithine, glycine, and hexose were picked up. As the components that make up the acidity, lactic acid, gluconic acid, malic acid, citric acid, and tartaric acid were picked up. As the components that make up the bitterness, arginine, phenylalanine, tyrosine, histidine, lysine, leucine, isoleucine, methionine, tryptophan, valine, nicotinic acid, nicotinamide, pantothenic acid, hypoxanthine, 4-aminobenzoic acid, and riboflavin were picked up.

[0066] Figure 3 is a graph showing the results when clustering was performed with the number of clusters n specified as 3 for the umami components. Cluster 3 is a group with a high total content of umami components, a high content of aspartic acid and glutamic acid, and a low content of succinic acid. Cluster 2 is a group with an intermediate content of umami components and a good balance of aspartic acid, glutamic acid, and succinic acid. Cluster 1 has a high content of succinic acid and contains approximately twice the content of each of aspartic acid and glutamic acid.

[0067] Figure 4 is a graph showing the results of clustering with the number of clusters n specified as 3 for the sweet components. Cluster 3 has a high total content of sweet components, contains similar amounts of threonine, alanine, and proline, has the next highest amount of serine, and contains small amounts of ornithine and glycine. Cluster 2 contains the most proline, followed by threonine, alanine, serine, ornithine, and glycine. Cluster 1 contains similar amounts of proline and threonine, followed by alanine, serine, ornithine, and glycine. Also, in any of the groups, the content of hexose sugars was trace amounts and negligible here.

[0068] Figure 5 is a graph showing the results of clustering with the number of clusters n specified as 3 for the sour components. Cluster 3 has a high total content of sour components, with a very high content of lactic acid, followed by gluconic acid, malic acid, and citric acid. Cluster 2 has a lower total content of sour components than Cluster 3, but the proportion of sour components it constitutes is almost the same as that of Cluster 3. It was judged that there is a high possibility that MLF has been carried out for Cluster 2 and Cluster 3, and they may be classified as the same group. However, as the number of reference wines increases, the distinction between Cluster 2 and Cluster 3 may become clearer, so they are described separately. On the other hand, Cluster 1 had an equivalent content of sour components to Cluster 2, but malic acid accounted for the majority.

[0069] Also, Cluster 3 of the above sour components has a mellow and creamy taste, Cluster 2 has a slightly higher citric acid content and a slightly sour taste, and Cluster 1 has a high citric acid content and a sharp sour taste.

[0070] Figure 6 is a graph showing the results of clustering with the number of clusters n specified as 3 for bitter components. Cluster 3 had a high content of bitter components. When looking at the components separately, it had a very high content of arginine, followed by large amounts of phenylalanine, tyrosine, and histidine, and small amounts of other components such as lysine, leucine, isoleucine, methionine, tryptophan, nicotinamide, and valine. Cluster 2 had a high content of phenylalanine, followed by large amounts of tyrosine and histidine, and contained arginine, lysine, and leucine, and was characterized by small amounts of isoleucine, nicotinamide, pantothenic acid, hypoxanthine, 4-aminobenzoic acid, valine, etc. Cluster 1 had a low content of bitter components, but had a high content of phenylalanine, followed by large amounts of tyrosine and histidine, and was characterized by small amounts of arginine, isoleucine, nicotinamide, pantothenic acid, hypoxanthine, 4-aminobenzoic acid, valine, etc.

[0071] (Group Classification) Based on the clustering results for each basic taste up to this point, the data was integrated to perform group classification of wines. Examples of group classification methods include sorting by values in Excel, arranging them in order, and classifying by total values. Since the number of reference wines was 53 this time and not very large, for convenience, they were divided into 11 groups (classified as Group in Tables 1 and 2), but if the number of reference wines increases, the number of groups can be increased. An example of the sorted table is shown in Table 1 below. Also, a list of the prices of each wine and their average values is shown in Table 2.

[0072]

Table 1

[0073]

Table 2

[0074] As can be seen from Table 2, in each of Groups 1 to 11, there was not much difference in the average price, but in Group 1 and Group 2, there were wines with high prices. For example, 2021013 and 2021014 also received high evaluations in terms of taste evaluation, and this is reflected in the high price. Therefore, it can be determined that wines belonging to Group 1 and Group 2, which have high scores in each basic taste, are likely to be highly evaluated in terms of taste evaluation, and the wines whose evaluation should be reflected in price.

[0075] Subsequently, using the group classification of this reference wine as the teacher value, prediction models for groups were constructed using three types: random forest, SVM, and neural network, and the group of the test sample wine was predicted. The test sample was also subjected to component analysis by triple quadrupole mass spectrometer LCMS TM -8050 (manufactured by Shimadzu Corporation) and LC / MS / MS method package cell culture profiling to perform quantification and create profile data for each wine. Although prediction was possible with any of the above prediction models, Table 3 shows the results of classification using the prediction model using SVM.

[0076]

Table 3

[0077] (Results) The above G2021-0031 to 0036 are commercially available wines, but the others are non-commercially available wines. By classifying the taste into groups, it was possible to confirm which commercially available wines have a similar taste. It was also confirmed that it is possible to compare the prices of wines with similar characteristics and use them as reference values. For example, it can be proposed to traders that the wine of G2021-0027, although not yet on the market, may be priced slightly higher.

[0078] [Example 2] As described above, the taste of wine has been evaluated based on its components. However, since the taste of wine is also important in the so-called "marriage" of the taste of the food itself in a meal, in addition to the thresholds of each component, while taking into account the synergistic effect of taste when combined with a meal, an attempt was made to further improve the accuracy of the generated learning model by incorporating additional indicators.

[0079] [Umami] Regarding the reference wine that was analyzed using LCMSTM-8050 and stored at 4°C as described in the above reference example, an examination was conducted. In Example 1 above, as components constituting umami, glutamic acid, aspartic acid, and succinic acid were picked up and examined. However, this time, first, it was decided to create a correlation matrix, which is used to confirm the strength of the correlation between variables, to determine whether there is a strong correlation (multicollinearity) between the variables of the components. Using the pandas library, which is a library for data manipulation and analysis, the correlation was calculated and displayed in the form of a heatmap (Figure 9(a)) using matplotlib and seaborn, and the correlation relationship for each cluster of glutamic acid and aspartic acid was graphed. (Figure 9(b)).

[0080] As is clear from Figures 9(a) and (b), the correlation between glutamic acid and aspartic acid is high, and as the concentration of aspartic acid increases, glutamic acid also tends to increase. While the taste threshold of glutamic acid is said to be 30 mg / 100 mL, the content of glutamic acid in wine is reported to be 0 mg to 16.8 mg / 100 ml (Food Chem. 2021 Oct 30;360:128971). Therefore, it is considered that it is rare to feel the umami of glutamic acid in wine alone. However, it has been reported that when combined with inosinic acid, its umami becomes 8 to 100 times due to the synergistic effect (Advances in Biochemistry and Physiology (eds Filer, L. J., Jr. et al.) 35-54)

[0081] As described above, the correlation between glutamic acid and aspartic acid is quite large. However, the threshold value of aspartic acid has been reported to be 160 mg / 100 mL (Journal of the Japanese Society for Agricultural Chemistry 65 (2), 163 - 169, 1991), and there is a report that umami is only 8% of glutamic acid (Advances in Biochemistry and Physiology (eds Filer, L. J., Jr. et al.) 35 - 54). Here, from the perspective of the marriage in diet, aspartic acid was excluded, and glutamic acid was picked up as the umami component in the present invention.

[0082] On the other hand, as is clear from Fig. 9(a), it was confirmed that the correlation between glutamic acid and succinic acid is small. From the above, for the analysis of the components constituting the umami in wine, the concentrations of glutamic acid and succinic acid were determined based on the peak values.

[0083] (Determination of the number of umami clusters) When performing cluster analysis, as a method for predicting appropriate clusters, analysis was performed by the elbow method (Fig. 10(a)) and the silhouette method (Fig. 10(b)), and it was predicted that seven were appropriate. Therefore, for umami, it was decided to classify it into seven clusters.

[0084] (Characteristics of each umami cluster) When the clusters are divided into seven by the K-means method and confirmed in the scatter plot of Fig. 11(a), it can be seen that the plots of each cluster are color-coded without overlapping. Regarding how separated each data point is from its own cluster and other clusters, based on the graph of the silhouette method in Fig. 10(b), when calculated using the silhouette score, which is an indicator for quantitatively evaluating, the average silhouette coefficient was calculated to be 0.403. If the average value is 0.4 or more, the result of clustering is judged to be "good". Actually, when calculating the characteristic values for each cluster, it can be seen that each cluster is characterized as shown in Figs. 12(a) and (b). Note that the values on the X-axis and Y-axis use the data of the peak values digitized in the component analysis of the reference wine performed in Reference Example 1. Also, it was confirmed in the case of only red wine (Fig. 11(b)) and in the case of white wine and sparkling wine (Fig. 11(c)). In red wine, it was shown that glutamic acid and succinic acid are inversely correlated. It was confirmed that those with a large amount of glutamic acid have a small amount of succinic acid, and those with a large amount of succinic acid have a small amount of glutamic acid.

[0085] Note that in the above cluster classification, since the No. of the cluster was automatically assigned, using the data of the peak values digitized in the component analysis of the reference wine performed in the above Reference Example 1, the amounts of glutamic acid and succinic acid for each cluster were confirmed, and the assignment of each cluster group (G) was performed again. Graphs showing the peak values for each cluster group (G) were created for glutamic acid (Fig. 12(a)) and succinic acid (Fig. 12(b)). Characteristics of each cluster G7: Cluster 3: Glutamic acid is very high, succinic acid is medium G6: Cluster 1: Glutamic acid is high, succinic acid is low G5: Cluster 5: Glutamic acid is high, succinic acid is medium G4: Cluster 0: Glutamic acid is medium, succinic acid is high G3: Cluster 4: Glutamic acid is very low, succinic acid is very high G2: Cluster 2: Low glutamic acid, medium succinic acid G1: Cluster 6: Low glutamic acid, low succinic acid Regarding the scatter plot in Fig. 11(c), the redrawn graph when the assignment of each cluster group (G) is performed is shown in Fig. 11(d). Through such grouping, it was confirmed that the clusters of G4 and G3 do not exist in white wine or sparkling wine. Regarding the scatter plots of sweetness, acidity, and bitterness hereafter, the scatter plot with the cluster No. automatically assigned is modified, and the scatter plot with the assignment of each cluster group (G) is displayed.

[0086] [Sweetness] In the above Example 1, glutamic acid, aspartic acid, and succinic acid were picked up and examined as components constituting sweetness. This time, first, whether there is a strong correlation (multicollinearity) between the variables of the components was confirmed by calculating the correlation using the correlation matrix, which is used to confirm the strength of the correlation between variables, with pandas, and displaying it in the form of a heat map using matplotlib and seaborn (Fig. 13).

[0087] It has been reported that when inosinic acid (IMP) is added to sweet amino acids such as L-alanine and L-serine by psychophysical methods, taste enhancement occurs (Chemical Senses, Volume 27, Issue 8, October 2002, Pages 739-745). Therefore, based on the peak values of the concentrations of sweet components in wine, a correlation matrix was plotted for proline, alanine, glycine, threonine, and serine by representing it as a heat map. As is clear from Figure 13, proline had no correlation with the others. For alanine, glycine, threonine, and serine, the correlation coefficients were found to be 0.81 each, indicating a high correlation. Therefore, alanine, which is known to have a strong sweet taste in wine (Physiol Behav. 1981 Jul;27(1):51-9), was picked up from the group consisting of alanine, glycine, threonine, and serine, and furthermore, together with proline, which has a low correlation with the above group, cluster analysis was performed on the sweetness based on the peak values of the concentrations of sweet components in wine.

[0088] (Determination of the number of sweetness clusters) When predicted by the elbow method, the result was that the number of clusters being 3 was appropriate. (Data not shown) In the above cluster classification, since the cluster No. was automatically assigned, the data of the peak values digitized in the component analysis of the reference wine performed in Reference Example 1 were used to reassign each cluster group (G).

[0089] (Characteristics of each sweetness cluster) The clusters were divided into three by the K-means method, and in order to check whether the clusters were separated, the pair plots of proline and alanine were checked by changing the X-axis and Y-axis. As can be seen from the scatter plots of the pair plots in Figs. 14(a-1), (a-2) and (b), it can be seen that the plots of each cluster are color-coded without overlapping. When calculated using the silhouette score, the average silhouette coefficient was calculated to be 0.503. If the average value is 0.4 or more, the result of clustering is judged to be "good". In order to check how various wines are distributed, each group was labeled, and as is clear from Fig. 14(b), which is a graph showing the correlation between proline and alanine, it was confirmed that each cluster is distributed with characteristics.

[0090] Also, the amounts of proline (Fig. 15(a)) and alanine (Fig. 15(b)) for each cluster were checked. The characteristics of each group for which clustering was performed are shown below for the components that make up the umami. Characteristics of each cluster G3: High in proline and low in alanine G2: Slightly high in proline and high in alanine G1: Low in both proline and alanine

[0091] [Sourness] As is clear from the heat map in Fig. 16, a high correlation was shown between malic acid and lactic acid, and between malic acid and citric acid, and a somewhat high correlation was also shown between lactic acid and citric acid. However, in the acidity of wine, malic acid, lactic acid, and citric acid are the main acids. In particular, malic acid and lactic acid increase or decrease under the influence of malolactic fermentation, and the results are also very important. Therefore, acids were targeted as analysis items. Succinic acid is not only a umami component but also a component felt as acidity, and it was decided to add it to the analysis items for this acidity. Based on the peak values of the concentrations of the acidity components of malic acid, lactic acid, citric acid, and succinic acid in wine, a study was conducted using these four components. Note that glucuronic acid is mainly contained in the juice of noble rot grapes invaded by Botrytis cinerea, so it is considered inappropriate to include it in the evaluation of the whole wine as the above-mentioned glycoside sweetness, and it was not included in the study this time.

[0092] (Determination of the number of clusters of acidity) When predicting the number of clusters using the elbow method, the result was obtained that the number of clusters being 5 is appropriate. On the other hand, the silhouette method showed that the number of clusters being 6 is appropriate. In order to verify in detail, the silhouette scores for each number of clusters were calculated. As a result, when the number of clusters is 5, the average silhouette score is 0.39, which is close to 0.4, but there are clusters where the silhouette score becomes negative for some data points. This suggests that the separation between clusters is not sufficient and the classification may be inappropriate. Furthermore, when comparing by changing the number of clusters to 4, 5, and 6, when the number of clusters is set to 6, the average silhouette score becomes 0.41, and the silhouette score is 0 or more for all data points. From this, it was confirmed that the separation between clusters was improved and a more appropriate classification was being performed. For the above reasons, it was judged that it is appropriate to set the number of clusters to 6.

[0093] (Characteristics of each cluster of acidity) When the clusters were divided into six by the K-means method and confirmed by the scatter plot of the pair plot in Fig. 17, etc., combinations in which each cluster was mixed were also seen in the pair plot. However, as a result of checking the box-and-whisker plot of the feature amounts, differences were found in the distribution of the feature amounts for each cluster, and it was confirmed that certain characterization could be done between the clusters. For this reason, it was judged that the clustering was appropriately performed.

[0094] As is clear from Fig. 18, in cluster G1 (the upper left group) where lactic acid is low and malic acid is high, none of the red wines among the reference wines measured this time are included, and the wines in cluster G1 are considered not to have undergone lactic acid fermentation.

[0095] In the above cluster classification, since the No. of the clusters was automatically assigned, as shown in Figs. 19(a) to (d), using the data of the peak values quantified in the component analysis of the reference wines performed in Reference Example 1, the abundance was confirmed, and the assignment of each cluster group (G) was performed again. The characteristics of each group for which clustering was performed for the components constituting the acidity are shown below. Characteristics of each cluster G6: Malic acid is low, lactic acid is high, citric acid is low, and succinic acid is medium. G5: Malic acid is low, lactic acid is medium, citric acid is medium, and succinic acid is medium. G4: Malic acid is low, lactic acid is medium, citric acid is low, and succinic acid is medium. G3: Malic acid is low, lactic acid is medium to high, citric acid is medium, and succinic acid is low. G2: Malic acid is low, lactic acid is medium, citric acid is low, and succinic acid is medium to high. G1: Malic acid is high, lactic acid is low, citric acid is medium to high, and succinic acid is low to medium.

[0096] [Bitterness] Regarding bitterness, a correlation matrix used to confirm the strength of the correlation between variables using ten amino acids, namely proline, histidine, arginine, tryptophan, arginine, valine, methionine, tryptophan, phenylalanine, isoleucine, leucine, and lysine, was plotted in Fig. 20 by representing it as a heatmap.

[0097] As is clear from Fig. 20, proline, histidine, tryptophan, and arginine have a low correlation with each other, but a correlation was observed in the group consisting of valine, methionine, phenylalanine, isoleucine, leucine, and lysine. On the other hand, arginine, phenylalanine, valine, and leucine are reported to be mainly associated with bitterness at a threshold in the range of 40 - 90 mg / 100 mL (Food Sci Biotechnol. 2022 May 19;31(7):767 - 785). Therefore, in addition to proline, histidine, tryptophan, and arginine, valine and phenylalanine were picked up, and cluster analysis was performed using the peak values of six amino acid components for the concentration of bitter components in wine.

[0098] It was predicted that 3 was appropriate by the elbow method (data not shown).

[0099] (Characteristics of each cluster of bitterness) In the above cluster classification, since the No. of the cluster was automatically assigned, when the abundance of each component for each cluster was confirmed using the data of the peak values quantified in the component analysis of the reference wine performed in Reference Example 1, except for arginine, there were no major characteristics in clusters 0 and 1. Therefore, these clusters were grouped as the same group, and the assignment of each cluster group (G) was performed again. The characteristics of each group for which clustering was performed for the components constituting bitterness are shown below. Characteristics of each cluster G2: Cluster 2: High in arginine G1: Cluster 0: Low in arginine G1: Cluster 1: Low in arginine

[0100] (Comprehensive Evaluation) From the above evaluations, summarizing the results of the group classification, it was as follows. They were sorted in descending order from the total score. There were only two groups for bitterness, and there were few wines corresponding to Group 2. Bitterness components such as histidine, arginine, valine, methionine, phenylalanine, leucine, and lysine are contained in the wines, but all of them were originally below the threshold of 40 - 90 mg / 100 mL, and it was confirmed that they were below the threshold, so they were not included in the comprehensive evaluation.

[0101] The group points of umami, sweetness, and acidity were totaled and divided into 12 groups from 16 points to 3 points. It is shown in Table 4 below.

[0102] When the umami is 5 points or more, the glutamic acid is high. In addition, wines with high sweetness and acidity are found among wines with a total of 13 points or more. The price range is from 2,800 yen to 16,400 yen, but the average price is 8,750 yen, and there is a tendency to be evaluated as high - priced. For those with umami of 5 points or more, the price range was from 1,345 yen to 16,400 yen, and the average was 6,523 yen. Wines with umami of 4 points and 3 points are wines with a lot of succinic acid. Among 15 bottles, 8 bottles were Japanese wines. Among 52 bottles, there were 13 Japanese wines, and 8 out of 13 had a lot of succinic acid. Succinic acid goes well with a diet of fish and shellfish, and it was confirmed that Japanese wines are suitable for Japanese cuisine. On the other hand, there were few Japanese wines with a lot of glutamic acid. Of course, the number of reference wines was not sufficient, and there may be a bias, but the fact that Japanese wines are often said to be suitable for Japanese cuisine is likely due to these components.

[0103]

Table 4

[0104] When checking the relationship between the total points and the market price (Table 5), there were no wines with low or high total points. On the other hand, there were wines with high total points and low market prices. Since the price of wine is determined not only by taste but also by factors such as brand, there is often no correlation with taste. However, it is suggested that by using the evaluation method of the present invention, it may be possible to set a high price based on taste. To confirm this, for wines without a market price, points were predicted using the above grouping as a model. The results are shown in Table 6.

[0105]

Table 5

[0106] Table 6 includes wines that are not currently on the market. After performing component analysis, it was confirmed that there are also wines with high total points. It can be utilized as useful data for predicting the market price.

[0107]

Table 6

[0108] The following Table 7 describes the taste characteristics of each group by a panel group of wine experts. This evaluation is not only to strictly evaluate the aroma, texture, and aftertaste of the wine itself according to strict rules such as rinsing the mouth with water and refreshing it before moving to another sample as is done in wine tastings, but also to evaluate the effect of the marriage with dishes in each taste by tasting the wine while tasting meat and fish dishes. The results are shown in Table 7 below.

[0109]

Table 7

[0110] (Results) As is clear from Table 7, the content of glutamic acid in wine is below the threshold, but it has been clarified that when tasted with dishes containing inosinic acid such as meat and cheese, the umami is enhanced by the synergistic effect. Also, it was confirmed that Groups 3 and 2, which contain a large amount of succinic acid, enhance umami when tasted with seafood dishes. Except for sweet wines and champagne, the sugar in wine should be consumed during alcoholic fermentation and the sweetness of sugar should not be felt, but there are wines known to have a sweet taste. As a result of measurement, it was confirmed that when there is a large amount of proline or alanine, it tastes sweet and has a richness. Also, proline has an effect of enhancing the umami of glutamic acid, and it was confirmed that umami is enhanced when tasted with kombu dashi, seafood, and meat. It is well known that when malolactic fermentation occurs, the malic acid decreases while the lactic acid increases and other amino acids also increase, resulting in an overall mellow flavor. In fact, when measured, it was clearly divided into whether there is a large amount or a small amount of lactic acid. However, although the malic acid was small, there was variation in the concentration of lactic acid. When proline, glutamic acid, or succinic acid is high in wine, it feels rich, but when the acidity is low, the balance is bad and it does not match light dishes. However, there were many wines that could be enjoyed sufficiently even alone. Wines with a high measured value of malic acid had a sharp taste. The bitter components correspond to amino acids produced during the fermentation process, but as in Additional Example 2, the measured values of each component were below the threshold, and it was considered difficult to evaluate the individual components.

[0111] [Example 3] [Analysis by Component Concentration] So far, the measurement has been performed based on the peak value, but the evaluation based on the peak value is a relative evaluation and tends to lack reproducibility when the number of samples increases. Therefore, calculations were performed based on the concentration of each component in wine using a different sample from before.

[0112] [Umami] As is clear from FIGS. 22(a) and (b), succinic acid was 200 mg / L or more, i.e., above the threshold, in both red wine and white wine.

[0113] As described above, while the taste threshold of glutamic acid is said to be 30 mg / 100 mL, the content of glutamic acid in wine has been reported to be 0 mg to 16.8 mg / 100 ml. However, when glutamic acid is combined with inosinic acid, if the umami effect of glutamic acid is enhanced at least eightfold due to the synergistic effect, the umami effect of glutamic acid can be expected if the concentration of glutamic acid is at least 37.5 mg / L (30 mg / 100 mL ÷ 8). Therefore, a line was drawn at a glutamic acid concentration of 37.5 mg / L in the graphs of FIGS. 22(a) and (b). In red wine, in particular, wines belonging to groups G7 and G6 were confirmed to exhibit an umami effect due to the synergistic effect when consumed with dishes rich in inosinic acid.

[0114] For example, wine No. 306 in the graph shows that both glutamic acid and succinic acid are present at a certain level. On the other hand, No. 305 also shows a high amount of succinic acid. Both No. 306 and No. 305 are produced by the same winery. Wines made from this winery have been shown to be wines with high umami. On the other hand, white wines and sparkling wines generally have a glutamic acid content of 37.5 mg / L or less, with little synergistic umami effect, and can be expected to have a refreshing throat feel even when paired with dishes containing inosinic acid.

[0115] Neither the WSET, the world's largest wine education institution, nor the Japan Sommelier Association includes the evaluation of wine umami as an evaluation item. However, the marriage of the taste of alcoholic beverages and food is an important factor and is considered to be a particularly useful indicator when evaluating the umami potential.

[0116] (Sweetness) Cluster analysis of proline and alanine was performed as described above. The number of clusters was set to 3 and shown as scatter plots in Figs. 23(a) and (b). Although there are various reports on the sweetness threshold of proline, a red line was drawn on the graph assuming it to be 690 mg / L.

[0117] The amounts of proline (Fig. 24(a)) and alanine (Fig. 24(b)) for each cluster were confirmed, and the assignment of each cluster group (G) was carried out. G1: Cluster 0: Both alanine and proline were low, and proline was below the threshold. G2: Cluster 2: Proline was very high and above the threshold. G3: Cluster 1: Proline was above the threshold and alanine was at a medium level. This result was roughly consistent with the case of analysis at the peak value.

[0118] (Summary) - By calculating based on concentration instead of peak value, comparison with the threshold became possible. Also, since it can be stored in the database as an absolute value, it became clear that it is desirable to measure by concentration. - As the components to be measured, in particular, by measuring glutamic acid, succinic acid, proline, alanine, malic acid, lactic acid, and citric acid, it was found that it is possible to classify the taste of wine. In particular, conventionally, there has been no index for evaluating umami in wine or sweetness other than in sweet wines. However, in reality, umami was present and sweet components other than sugar could be detected. Glutamic acid contained in wine is below the threshold and cannot be evaluated by normal tasting, but can be detected by measurement. Considering enjoying wine with food, measurement of umami is considered very useful. - Of course, as a taste through the throat, glycerin, tannin, etc. are also considered to affect the taste, and including aroma, etc., it is necessary to include many elements. However, since complex elements are difficult for actual end-users to understand, simple classification is useful.

Industrial Applicability

[0119] Based on a classification method that emphasizes taste for wine, the present invention can propose to wine traders and consumers regarding the taste and price setting of individual wines, including the compatibility with dishes.

Explanation of Signs

[0120] 1 Classification system 4 Pretreatment unit 5 Component detection unit 10 Data acquisition unit 20 Quantification unit 30 Data extraction unit 35 Component determination unit 40 Clustering unit 50 Group classification unit 60 Comment creation unit 70 Data storage unit 71 Profile data storage unit 72 Feature value data storage unit 73 Cluster data storage unit 74 Group classification data storage unit 75 Comment data storage unit 80 Information presentation unit 810 Display

Claims

1. A method for classifying wines, sequentially comprising the following (a) to (f). (a) A data acquisition step of acquiring component profile data of a reference wine; (b) A quantification step of quantifying the content of each component constituting the component profile data of the reference wine acquired in the data acquisition step of step (a) as a characteristic value; (c) A data extraction step of extracting the characteristic values of the components constituting the basic taste from the characteristic values of each component of the reference wine created in the quantification step of step (b); (d) A clustering step of forming n cluster groups for each basic taste by performing clustering by unsupervised learning for each characteristic value of the components constituting each basic taste extracted in the extraction step of step (c); (e) A group classification step of classifying reference wines based on the cluster groups for each basic taste formed in step (d); (f) A comment creation step of associating taste information and creating a comment;

2. The classification method according to claim 1, characterized in that the basic taste consists of umami and sweetness.

3. The classification method according to claim 1, characterized in that the basic taste consists of umami, sweetness, and sourness.

4. The classification method according to claim 1, characterized in that the basic taste consists of umami, sweetness, sourness, and bitterness.

5. The classification method according to any one of claims 2 to 4, characterized in that the component(s) constituting umami contains one or more selected from glutamic acid, inosinic acid, aspartic acid, adenylic acid, and succinic acid.

6. The classification method according to any one of claims 2 to 4, characterized in that the component(s) constituting umami is / are glutamic acid and succinic acid.

7. The classification method according to any one of claims 2 to 4, characterized in that the component(s) constituting sweetness contains one or more selected from proline, glycine, alanine, threonine, serine, ornithine, and asparagine.

8. The classification method according to any one of claims 2 to 4, characterized in that the component(s) constituting sweetness is / are proline and alanine.

9. The classification method according to claim 3 or 4, characterized in that the component(s) constituting sourness contains one or more selected from tartaric acid, malic acid, citric acid, succinic acid, lactic acid, glycyl-glutamine acetate, seryl-glutamine, and glycyl-asparagine.

10. The classification method according to claim 3 or 4, characterized in that the components constituting the sour taste are malic acid, citric acid, succinic acid, and lactic acid.

11. A wine classification system that outputs comments on wine, including classification results using component profile data of wine, comprising the following (A) to (F). (A) A data acquisition unit that acquires data of component profile data of reference wine; (B) A quantification unit that quantifies the content of each component constituting the component profile data of the reference wine acquired by the data acquisition unit as a characteristic value; (C) A data extraction unit that extracts quantified data of components constituting the basic taste from the quantified profile data; (D) A clustering unit that forms n clusters for each basic taste by performing clustering by unsupervised learning for each quantified data of components constituting each basic taste; (E) A group classification unit that classifies reference wines based on the n clusters formed for each basic taste; (F) A comment creation unit that associates taste information and creates a comment.

12. The classification system according to claim 11, characterized in that the basic taste consists of umami and sweetness.

13. The classification system according to claim 11, characterized in that the basic taste consists of umami, sweetness, and sourness.

14. The classification system according to claim 11, characterized in that the basic taste consists of umami, sweetness, sourness, and bitterness.

15. The classification system according to any one of claims 12 to 14, characterized in that the components constituting umami include one or more selected from glutamic acid, inosinic acid, aspartic acid, adenylic acid, and succinic acid.

16. The classification system according to any one of claims 12 to 14, characterized in that the components constituting umami are glutamic acid and succinic acid.

17. The classification system according to any one of claims 12 to 14, characterized in that the components constituting sweetness include one or more selected from proline, glycine, alanine, threonine, serine, ornithine, and asparagine.

18. The classification system according to any one of claims 12 to 14, characterized in that the components constituting sweetness are proline and alanine.

19. The classification system according to claim 13 or 14, characterized in that the component(s) constituting the sour taste comprises one or more selected from tartaric acid, malic acid, citric acid, succinic acid, lactic acid, glycyl-glutamine acetate, seryl-glutamine, and glycyl-asparagine.

20. The classification system according to claim 13 or 14, characterized in that the component(s) constituting the sour taste is / are malic acid, citric acid, succinic acid, and lactic acid.

21. A method for collecting wine classification data, which classifies, according to the wine classification method according to any one of claims 1 to 4, using the component profile data of wine presented by a user as input data, and records the result as reference data.

Images