Cigarette flavor regulation and control method

Abstract

The invention discloses a regulation and control method for cigarette flavor. The regulation and control method comprises the following steps: constructing a flavor standardized database comprising material characteristics, experimental conditions and sensory evaluation results; according to the flavor standardization database, based on a tobacco text mining strategy, identifying a core flavor-causing substance; carrying out heterogeneity test and effect quantity combination on the core flavor-causing substances through Meta analysis, and determining quantitative correlation between each substance and a sensory effect by adopting a random effect model so as to determine an optimal regulation and control range; based on single-factor ANOVA and clustering analysis, carrying out production area raw material type selection; the sensory quality is accurately regulated and controlled. According to the regulation and control method of the cigarette flavor, a complete regulation and control system is constructed by integrating multidisciplinary technologies, directional, quantitative and efficient regulation and control of the flavor are achieved, a scientific basis is provided for accurate regulation and control of sweet fragrance, cool feeling and fruit sweet flavor of cigarettes, and a technical support is also provided for diversified innovation of flavor in the tobacco industry.

Description

Technical Field

[0001] This invention relates to the field of tobacco processing technology, and more specifically, to a method for controlling the flavor of cigarettes. Background Technology

[0002] The ongoing global movement to control tobacco use and the growing health awareness of consumers have prompted the tobacco industry to shift towards flavor diversification. Sweetness, coolness, and fruitiness, as core flavor dimensions of cigarettes, have become crucial for companies to achieve product differentiation. While existing research has identified some flavor-causing substances, several technical shortcomings remain: First, there is a lack of a systematic screening system for core flavor-causing substances. Most studies rely on single database searches, leading to an underestimation of the role of low-threshold flavor-causing substances. Furthermore, general text mining tools are not adapted to tobacco-specific terminology, resulting in biased extraction of key information. Second, the dose-response relationship between flavor-causing substances and sensory effects is unclear. Existing data is fragmented, and no directly applicable quantitative control standards have been established. Third, research on the correlation between regional differences and raw material selection is insufficient, making it impossible to achieve precise matching between raw materials and target flavors. Fourth, the synergistic control mechanism of multiple substances is unclear, making it difficult to achieve coordinated enhancement of the three flavor categories.

[0003] For example, current technologies for sweetness control largely rely on empirical sugar addition or the addition of sweeteners, without considering the balancing effect of the sugar-to-alkali ratio, leading to an imbalance between sweetness and irritation. Cooling sensation control suffers from problems such as insufficient spiciness or cooling due to inappropriate cooling agent concentrations, and fails to fully utilize the synergistic effect between novel and traditional cooling agents. Fruity sweetness, due to the difficulty in creating a realistic fruity flavor from a single substance, lacks a clear understanding of the application value of aroma-generating pathways such as microbial production. Furthermore, the chemical composition of tobacco leaves varies significantly across different production areas, but existing methods have not established a correspondence between production area classification and flavor adaptation, resulting in low raw material utilization and poor product quality stability.

[0004] Therefore, there is an urgent need for a method to control the flavor of cigarettes. Summary of the Invention

[0005] The purpose of this invention is to provide a method for controlling cigarette flavor, so as to solve the problems in the prior art, and to accurately control cigarette flavor in a scientific, systematic and quantitative manner.

[0006] This invention provides a method for controlling the flavor of cigarettes, comprising:

[0007] A standardized flavor database is constructed, which includes material characteristics, experimental conditions, and sensory evaluation results.

[0008] Based on the flavor standardization database, and using a tobacco text mining strategy, the core flavor compounds corresponding to each flavor are identified, including sweetness, coolness, and fruitiness.

[0009] Meta-analysis was used to test the heterogeneity and combine the effect sizes of the identified core flavoring substances. A random effects model was used to determine the quantitative correlation between each substance and sensory effects in order to determine the optimal range of regulation.

[0010] Based on univariate ANOVA and cluster analysis, the raw material selection of the production area is carried out;

[0011] Based on the identification results of the core flavor compounds, the optimal control range of the core flavor compounds, and the selection results of raw materials from the production area, the sensory quality is precisely controlled.

[0012] The method for controlling cigarette flavor as described above, preferably, involves constructing a flavor standardization database, which includes:

[0013] Based on domestic and international literature on the sweet aroma, coolness, and fruity sweetness of cigarettes, chemical index data, sensory evaluation data, production area data, and experimental condition data were extracted to establish a standardized database containing material characteristics, experimental conditions, and sensory evaluation. The chemical indexes include at least one of total sugar, reducing sugar, nicotine, total nitrogen, potassium, chlorine, sugar-to-alkali ratio, nitrogen-to-alkali ratio, menthol, benzyl alcohol, and isoamyl acetate.

[0014] The method for controlling cigarette flavor as described above, preferably, includes the following tobacco text mining strategy: performing a global scan of text sets on sweet aroma, coolness, and fruity sweetness flavors, retaining words whose occurrence frequency is ≥ a preset frequency threshold, calculating the word frequency of words in a single document using the logarithmic TF method, introducing a smoothing term to calculate the inverse document frequency, calculating word weights based on word frequency and inverse document frequency, screening substances with word weights ≥ a preset weight threshold as candidates for core flavor substances, and determining the final core flavor substances after significance testing.

[0015] In the cigarette flavor control method described above, preferably, the preset threshold number of times is 5.

[0016] The method of calculating the word frequency in a single document using the logarithmic TF method includes: calculating the word frequency using the following formula:

[0017] (1)

[0018] Where TF represents word frequency, and i represents the number of times the same word appears in a single document.

[0019] The calculation of inverse document frequency by introducing a smoothing term includes: calculating the inverse document frequency using the following formula:

[0020] (2)

[0021] Where IDF represents inverse document frequency, m represents the number of documents containing the term, and n represents the total number of documents.

[0022] The calculation of lexical weights based on word frequency and inverse document frequency includes: calculating lexical weights using the following formula:

[0023] TFIDF=TF×IDF (3)

[0024] Here, TF-IDF represents word weights; the higher the TF-IDF value, the greater the representational value of the words to the text's topic.

[0025] The preset weight threshold is 1.6.

[0026] The method for controlling cigarette flavor as described above, preferably, involves identifying the core flavor-causing substances corresponding to each flavor based on the flavor standardization database and a tobacco text mining strategy, including:

[0027] Based on the aforementioned tobacco text mining strategy, a tobacco-specific terminology database and a stop word database were constructed. Combining the bag-of-words model and the TFIDF algorithm, the core flavor compounds for three flavor categories were screened. The core flavor compounds corresponding to the sweet aroma flavor include total sugar, sugar-to-alkali ratio, 5-methylfurfural, and maltol. The core flavor compounds corresponding to the cool flavor are menthol, menthol, WS-23, and solanone. The core flavor compounds corresponding to the fruity sweet flavor are benzyl alcohol, isoamyl acetate, and linalool.

[0028] The method for regulating cigarette flavor as described above, preferably, involves performing heterogeneity testing and effect size pooling on the identified core flavor compounds through meta-analysis, and using a random effects model to determine the quantitative correlation between each substance and sensory effects, in order to determine the optimal regulation range, including:

[0029] The heterogeneity test in the meta-analysis was performed using a combination of the Q test and the I² statistic. When I² ≤ 50%, a fixed-effects model was used, and when I² > 50%, a random-effects model was used. The robustness of the results was verified by funnel plot symmetry test and Egger linear regression.

[0030] The method for controlling cigarette flavor as described above, preferably, involves using a random effects model to determine the quantitative correlation between each substance and sensory effects in order to determine the optimal control range, including:

[0031] The dose-response relationship for sweet aroma is as follows: for every 1-point increase in sugar-to-alkali ratio, the balance score between sweetness and astringency on a 5-point scale increases by 1.23 points, with an optimal control range of 812; for every 1-point increase in total sugar content, the average sweetness intensity on a 10-point scale increases by 0.85 points, with an optimal control range of 18%-22%; for every 1-point increase in reducing sugar content, the sweetness freshness score increases by 0.72 points, with an optimal control range of 16%-20%; for every 0.1 μg / g increase in 5-methylfurfural, the sweet aroma intensity score increases by 0.46 points, with an optimal control range of 0.05-0.2 μg / g; for every 0.01% increase in maltol addition, the sweetness persistence score increases by 0.58 points, with an optimal control range of 0.02%~0.05%.

[0032] The dose-response relationship for the cooling flavor is as follows: for every 0.1 μg / mL increase in menthol content, the cooling intensity on a 5-point scale increases by an average of 0.16 points, with the optimal control range being 0.2-1.0 μg / mL; for every 1 μg / g increase in solanone content, the content of the active cooling ingredient after pyrolysis increases by 16.47 μg / g, with the optimal control range being 0.5-1.0 μg / g; for every 0.05 μg / g increase in 2-pentylfuran content, the cooling freshness score on a 10-point scale increases by 0.19 points, with the optimal control range being 0.05-0.2 μg / g; and for every 0.1 μg / g increase in the amount of butylated hydroxytoluene added, the cooling persistence is extended by an average of 0.75 minutes, with the optimal control range being 0.1-0.3 μg / g.

[0033] The dose-response relationship corresponding to the fruity sweet flavor is as follows: when benzyl alcohol, linalool and geraniol are compounded in a ratio of 2:3:1, the fruity sweet flavor realism score is 8.6 points on a 10-point scale, and the harmony score is 9.2 points on a 10-point scale; when the mass ratio of isoamyl acetate to benzyl alcohol is 1:1.07, the fruity sweet flavor characteristics are most significant.

[0034] The method for controlling cigarette flavor as described above, preferably, involves selecting raw materials from production areas based on single-factor ANOVA and cluster analysis, including:

[0035] Based on univariate ANOVA and cluster analysis, the main tobacco-producing areas were divided into three categories: high-sugar and low-alkali, high-alkali and medium-sugar, and low-sugar and low-alkali, so as to select tobacco raw materials from the corresponding producing areas according to the target flavor requirements.

[0036] Among them, the types corresponding to the main tobacco-producing areas of Heilongjiang, Yunnan and Fujian are high sugar and low alkali, which are suitable for the development of sweet cigarettes; the types corresponding to the main tobacco-producing areas of Henan, Hunan and Shandong are high alkali and medium sugar, which are suitable for the development of strong cigarettes; the types corresponding to the low sugar and low alkali of the main tobacco-producing areas are Inner Mongolia, Jilin, Guangdong, Guizhou, Shaanxi and Chongqing, which need to be used after the chemical composition content is improved through the optimization of planting technology.

[0037] The selection of tobacco leaves from corresponding production areas based on target flavor requirements includes: tobacco leaves from high-sugar, low-alkali production areas with a total sugar content ≥23% and a nicotine content ≤1.8%; tobacco leaves from high-alkali, medium-sugar production areas with a nicotine content ≥2.4% and a total sugar content of 18%-21%; and tobacco leaves from low-sugar, low-alkali production areas, after optimization of planting techniques, with a total sugar content increased to over 20% and a nicotine content adjusted to 1.8%-2.2%.

[0038] The method for controlling cigarette flavor as described above, preferably, involves precise sensory quality control based on the identification results of core flavor compounds, the optimal control range of core flavor compounds, and the selection results of raw materials from the production area, including:

[0039] A random forest regression model was constructed, with core flavoring substances as feature variables and sensory indicators as target variables, to determine the core regulatory factors of each sensory indicator.

[0040] By adjusting the content and ratio of core regulatory factors, the sweet aroma, cool sensation, and fruity sweetness flavors can be precisely controlled.

[0041] In the cigarette flavor regulation method described above, preferably, the construction parameters of the random forest regression model are: 500 decision trees, node splitting criterion is minimizing the mean squared error, and core regulation factors are determined by calculating feature importance. Among them, the core regulation factor for aroma quality is nitrogen-base ratio, the core regulation factor for aroma quantity is total nitrogen, the core regulation factor for smoke concentration is total sugar, the core regulation factor for combustibility is potassium, the core regulation factor for off-flavors is nitrogen-base ratio, and the core regulation factor for irritation is total nitrogen.

[0042] The method involves precisely controlling the sweet aroma, cooling sensation, and fruity sweetness flavors by adjusting the content and ratio of core regulatory factors, including:

[0043] In the regulation of cooling flavor, the synergistic effect is strongest when the mass ratio of cooling agent to sweet substance is 1:5, which increases the duration of cooling by 40% and does not significantly reduce the sweetness. When fruit sweetness and cooling are synergistically regulated, the combination of phenylacetaldehyde and menthol can prolong the duration of cooling by 30%. The Maillard reaction product of banana extract combined with WS-23 can improve the coordination between sweetness and cooling by 2.3 points.

[0044] This invention provides a method for regulating cigarette flavor. By integrating multidisciplinary technologies to construct a complete regulation system, it achieves targeted, quantitative, and efficient flavor regulation. Through the synergistic application of multiple methods, it constructs a complete technical system from core substance identification and dose-effect relationship quantification to regional adaptation and quality regulation. It clarifies the dose-effect and intensity relationships of flavor-producing substances in cigarettes, providing a solid scientific basis for the precise regulation of sweet aroma, coolness, and fruity sweetness in cigarettes, and also providing important technical support for the diversified innovation of flavor in the tobacco industry. It can solve the technical problems of existing cigarette flavor regulation, such as lack of scientific quantitative basis, poor regional adaptability, and unclear synergistic effects of multiple substances. Attached Figure Description

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings, wherein:

[0046] Figure 1 A flowchart illustrating an embodiment of the cigarette flavor control method provided by the present invention;

[0047] Figure 2 Word cloud for the cooling sensation of cigarettes;

[0048] Figure 3 A schematic diagram of the TFIDF model for the cooling sensation of cigarettes;

[0049] Figure 4 A forest diagram of menthol;

[0050] Figure 5 A funnel diagram of menthone;

[0051] Figure 6 This is a schematic diagram of cluster analysis for total sugar and nicotine indicators. Detailed Implementation

[0052] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The descriptions of the exemplary embodiments are merely illustrative and are in no way intended to limit the present disclosure or its application or use. The present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided so that the present disclosure will be thorough and complete, and will fully express the scope of the disclosure to those skilled in the art. It should be noted that, unless specifically stated otherwise, the relative arrangement of components and steps, the composition of materials, numerical expressions, and values ​​set forth in these embodiments should be interpreted as exemplary only and not as limiting.

[0053] The terms “first,” “second,” and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different parts. Terms such as “including” or “contains” mean that the element preceding the term encompasses the element listed after it, and do not exclude the possibility of encompassing other elements as well. Terms such as “above” and “below” are used only to indicate relative positional relationships; when the absolute position of the described object changes, this relative positional relationship may also change accordingly.

[0054] In this disclosure, when a specific component is described as being located between a first component and a second component, an intermediary component may or may not be present between the specific component and the first or second component. When a specific component is described as connecting to other components, the specific component may be directly connected to the other components without having an intermediary component, or it may not be directly connected to the other components but may have an intermediary component.

[0055] All terms used in this disclosure (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in a general dictionary, such as a dictionary, should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.

[0056] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0057] like Figure 1 As shown, the cigarette flavor control method provided in this embodiment includes the following steps in actual implementation:

[0058] Step S1: Construct a flavor standardization database, which includes material characteristics, experimental conditions, and sensory evaluation results.

[0059] Specifically, based on domestic and international literature on the sweet aroma, coolness, and fruity sweetness of cigarettes, chemical index data, sensory evaluation data, production area data, and experimental condition data are extracted to establish a standardized database containing material characteristics, experimental conditions, and sensory evaluation. The chemical indexes include at least one of total sugar, reducing sugar, nicotine, total nitrogen, potassium, chlorine, sugar-to-alkali ratio, nitrogen-to-alkali ratio, menthol, benzyl alcohol, and isoamyl acetate.

[0060] In one embodiment of this invention, relevant literature on the sweet aroma, coolness, and fruity sweetness of cigarettes from 2004 to 2024, both domestically and internationally, is integrated for data extraction. The literature covers research findings from authoritative domestic databases and core international databases. The search employs a combination strategy of "core subject terms + flavor characteristic terms + material basis terms" to ensure comprehensive and targeted literature coverage. The retrieved literature undergoes a three-stage screening process of "duplicate removal - initial screening - secondary screening," ultimately including 218 valid articles.

[0061] Next, multi-dimensional data were extracted from valid literature, including chemical index data (total sugar, reducing sugar, nicotine, total nitrogen, potassium, chlorine, sugar-alkali ratio, nitrogen-alkali ratio, menthol, benzyl alcohol, isoamyl acetate, 5-methylfurfural, WS-23, etc.), sensory evaluation data (sweetness intensity, cooling intensity, fruit sweetness realism, aroma quality, aroma quantity, irritation, aftertaste, etc.), production area data (chemical component detection values ​​of tobacco leaves from 12 major tobacco-producing areas), and experimental condition data (flavoring method, reaction temperature, pH value, fermentation time, etc.). A standardized flavor database was established to eliminate the problems of data fragmentation and missing indicators.

[0062] Step S2: Based on the flavor standardization database, and using a tobacco text mining strategy, identify the core flavor compounds corresponding to each flavor, where flavors include sweetness, coolness, and fruitiness.

[0063] Specifically, based on the aforementioned tobacco text mining strategy, a tobacco-specific terminology database and a stop word database are constructed to avoid the adaptability defects of general word segmentation tools. Combining the Bag-of-Words (BOW) model and the TFIDF algorithm, text information in the flavor standardization database is processed to screen the core flavor compounds for three flavor categories. Specifically, the core flavor compounds corresponding to sweet aroma include total sugar, sugar-to-alkali ratio, 5-methylfurfural, and maltol; the core flavor compounds corresponding to cool aroma are menthol, menthol, WS-23, and solanone; and the core flavor compounds corresponding to fruity sweetness are benzyl alcohol, isoamyl acetate, and linalool. Further, the generation pathway of the core flavor compounds for fruity sweetness includes Maillard reaction and microbial aroma production. The Maillard reaction conditions are a temperature of 120-150℃, a pH of 5.0-5.5, and a reaction time of 2-4 hours. Microbial aroma production uses Corynebacterium fermentation strains. The tobacco-specific terminology database includes: sugar-to-alkali ratio, OAV value, WS-23, and 5-methylfurfural, etc.

[0064] The tobacco text mining strategy includes: performing a global scan of text sets related to sweet aroma, coolness, and fruity sweetness, retaining words whose occurrence frequency is greater than or equal to a preset frequency threshold to construct a vocabulary covering core terms and related terms; using the logarithmic TF method to calculate the word frequency in a single document, which can reduce the excessive influence of high-frequency general words, and introducing a smoothing term to calculate the inverse document frequency (IDF), which can avoid the problem of the denominator being 0; calculating word weights based on word frequency and inverse document frequency, screening substances with word weights greater than or equal to a preset weight threshold as candidates for core flavor substances, and determining the final core flavor substances for sweet aroma, coolness, and fruity sweetness after a significance test (p<0.05).

[0065] Specifically, in some embodiments of the present invention, the preset number of times threshold is 5.

[0066] The method of calculating the word frequency in a single document using the logarithmic TF method includes: calculating the word frequency using the following formula:

[0067] (1)

[0068] Where TF represents word frequency, and i represents the number of times the same word appears in a single document.

[0069] The calculation of inverse document frequency by introducing a smoothing term includes: calculating the inverse document frequency using the following formula:

[0070] (2)

[0071] Where IDF represents inverse document frequency, m represents the number of documents containing the term, and n represents the total number of documents.

[0072] The calculation of lexical weights based on word frequency and inverse document frequency includes: calculating lexical weights using the following formula:

[0073] TFIDF=TF×IDF (3)

[0074] Here, TF-IDF represents word weights; the higher the TF-IDF value, the greater the representational value of the words to the text's topic.

[0075] The preset weight threshold is 1.6. It should be noted that the present invention does not specifically limit the preset number threshold and the preset weight threshold.

[0076] The identified core flavor compounds for the sweet and aromatic flavor are: total sugar, sugar-to-alkali ratio, 5-methylfurfural, and maltol; the core flavor compounds for the cool flavor are: menthone, menthol, WS-23, and solanone; and the core flavor compounds for the fruity sweet flavor are: benzyl alcohol, isoamyl acetate, and linalool.

[0077] In one embodiment of the present invention, tobacco-specific text mining successfully identified the core flavor compounds and research hotspots for three flavor categories, and the core results are summarized in Table 1. Data shows that in the sweet aroma flavor, the frequency and TFIDF weights of "total sugar," "sugar-to-alkali ratio," and "5-methylfurfural" are all among the top three, with "sugar-to-alkali ratio" having a TFIDF weight of 1.89, reflecting its core position in sweetness balance research. In the cooling flavor, "menthol" has the highest frequency (186 times), while "WS-23" has the highest TFIDF weight (1.91), reflecting the research hotspot attribute of novel cooling agents. In the fruity sweet flavor, "benzyl alcohol," "isoamyl acetate," and "linalool" are the core substances, with "isoamyl acetate" having a TFIDF weight of 1.88, highlighting its specificity as a marker of citrus fruit aroma. The frequency and TFIDF weights of all core substances passed the significance test (p<0.05), verifying the reliability of the results.

[0078] Table 1 Summary of the characteristics of the three types of core flavor compounds

[0079]

[0080] The following analysis will use the cooling sensation of cigarettes as an example. Figure 2 In the word cloud diagram showing cooling flavor, "menthol," "WS-23," "menthyl lactate," "TRPM8 receptor," and "cooling intensity" are highlighted in large font to form the visual core, directly pointing to the three-tiered logic of cooling research: substance-mechanism-sensory aspect. "Menthol" and "WS-23" represent core cooling agents, "TRPM8 receptor" relates to the cooling mechanism, and "cooling intensity" corresponds to sensory evaluation indicators. Terms like "menthol," "menthyl succinate," and "eucalyptol" are surrounded by medium font, reflecting the diversity of cooling substances. The appearance of terms like "added with flavoring beads" and "added with tobacco" suggests the influence of flavoring methods on the cooling effect, representing important related information in the cooling text. Overall, the word cloud diagram clearly shows that cooling research focuses on cooling agents while also considering mechanisms and applications.

[0081] In the BOW (Browser-Wide Word) frequency ranking of cooling sensations, "menthol" led by a significant margin with 186 mentions, followed by "WS-23" (98 times) and "menthol lactate" (76 times). "TRPM8 receptor" (68 times) and "cooling threshold" (59 times) also appeared, with both co-occurring with "cooling agent concentration" 38 times. Furthermore, natural cooling substances such as "eucalyptol" (45 times) and "menthol monosuccinate" (39 times) accounted for over 50% of the literature on "synthetic cooling agents-natural substance synergy," indicating that cooling sensation research is extending from single substances to complex systems.

[0082] Figure 3The TFIDF weighted bar chart further clarifies the domain specificity of cooling substances. "WS-23" ranks first with a weight value of 1.91 and an IDF value of 2.17 (specifically appearing in only 5 articles). Compared with "menthol" (IDF=0.23, commonly appearing in 78% of cooling literature), it highlights its distinguishing significance in the study of "novel non-mentholic cooling". Although "WS-23" has a lower frequency, it is a key indicator for identifying "novel cooling agents". The weight values ​​of "TRPM8 receptor" (1.85) and "menthol" (1.72) follow closely behind. "TRPM8 receptor" is specific in the molecular mechanism literature because it focuses on the cooling mechanism (IDF=1.89). Although "menthol" has the highest frequency, its weight is slightly lower than "WS-23" because of its broad coverage. From the perspective of sensory association, the correlation between high-weight substances and cooling indexes reached a significant level (r>0.85, p<0.001). Among them, "WS-23" had the highest correlation coefficient with the intensity of nasal cooling sensation (r=0.92), which is consistent with the conclusion in the literature that "the cooling intensity of WS-23 is 1.8 times that of menthol".

[0083] Step S3: Through meta-analysis, heterogeneity tests and effect size pooling are performed on the identified core flavoring substances. A random effects model (when I²>50%) is used to determine the quantitative correlation between each substance and sensory effects in order to determine the optimal control range.

[0084] The heterogeneity test in the meta-analysis uses a combination of Q test and I² statistic. When I² ≤ 50%, a fixed-effects model is used, and when I² > 50%, a random-effects model is used. The robustness of the results is verified by funnel plot symmetry test and Egger linear regression. Results with slight bias are corrected by trimming and supplementation to ensure the reliability of the dose-effect relationship.

[0085] Specifically, the Q-test: based on the chi-square distribution, it tests for heterogeneity among studies, with the null hypothesis being "all study effect sizes are homogeneous," and the calculation formula is as follows:

[0086] (4)

[0087] in, This represents the effect size of the i-th study. The mean of the effect size. Let $\mathbf{x}$ be the effect size variance and $k$ be the number of studies included. If $P < 0.1$, the null hypothesis is rejected, indicating significant heterogeneity among the studies.

[0088] I² statistic: quantifies the degree of heterogeneity; the calculation formula is:

[0089] (5)

[0090] in, Indicates degrees of freedom. The larger the value, the stronger the heterogeneity. The specific criteria for judgment are as follows: ≤25% indicates low heterogeneity, 25% < ≤50% is considered moderate heterogeneity. >50% indicates high heterogeneity.

[0091] Based on the heterogeneity test results, the model selection principle is clarified as follows: when I² ≤ 50% (low / medium heterogeneity), a fixed-effects model is used; when I² > 50% (high heterogeneity), a random-effects model is used to combine effect sizes. This model considers both "intra-study random error" and "inter-study true differences." The calculation formula for the random-effects model is as follows:

[0092] (6)

[0093] in, This indicates the variance of heterogeneity among studies, and the quantitative association between flavor-causing substances and sensory effects is clarified by pooling effect sizes.

[0094] Furthermore, when determining the optimal control range, the optimal control range for each core flavoring substance is clarified based on the combined effect size and confidence interval, combined with sensory evaluation thresholds.

[0095] Specifically, the method of using a random effects model to determine the quantitative correlation between each substance and sensory effects in order to determine the optimal control range includes:

[0096] The dose-response relationship for sweet aroma is as follows: for every 1-point increase in sugar-to-alkali ratio, the balance score between sweetness and astringency on a 5-point scale increases by 1.23 points, with an optimal control range of 812; for every 1-point increase in total sugar content, the average sweetness intensity on a 10-point scale increases by 0.85 points, with an optimal control range of 18%-22%; for every 1-point increase in reducing sugar content, the sweetness freshness score increases by 0.72 points, with an optimal control range of 16%-20%; for every 0.1 μg / g increase in 5-methylfurfural, the sweet aroma intensity score increases by 0.46 points, with an optimal control range of 0.05-0.2 μg / g; for every 0.01% increase in maltol addition, the sweetness persistence score increases by 0.58 points, with an optimal control range of 0.02%~0.05%.

[0097] The dose-response relationship for the cooling flavor is as follows: for every 0.1 μg / mL increase in menthol content, the cooling intensity on a 5-point scale increases by an average of 0.16 points, with the optimal control range being 0.2-1.0 μg / mL; for every 1 μg / g increase in solanone content, the content of the active cooling ingredient after pyrolysis increases by 16.47 μg / g, with the optimal control range being 0.5-1.0 μg / g; for every 0.05 μg / g increase in 2-pentylfuran content, the cooling freshness score on a 10-point scale increases by 0.19 points, with the optimal control range being 0.05-0.2 μg / g; and for every 0.1 μg / g increase in the amount of butylated hydroxytoluene (BHT), the cooling duration is extended by an average of 0.75 minutes, with the optimal control range being 0.1-0.3 μg / g.

[0098] The dose-response relationship corresponding to the fruity sweet flavor is as follows: when benzyl alcohol, linalool and geraniol are compounded in a ratio of 2:3:1, the fruity sweet flavor realism score is 8.6 points on a 10-point scale, and the harmony score is 9.2 points on a 10-point scale; when the mass ratio of isoamyl acetate to benzyl alcohol is 1:1.07, the fruity sweet flavor characteristics are most significant.

[0099] Specifically, in one embodiment of the present invention, the meta-analysis of the nine core flavor compounds all adopted a random-effects model (I²>50%), and the pooled effect sizes and confidence intervals are shown in Table 3. The results showed that the sugar-to-alkali ratio contributed the most to the sweetness coordination, with an MD value of 1.23 (95% CI [0.98, 1.48]), meaning that for every 1 increase in the sugar-to-alkali ratio, the balance score (out of 5) for sweetness and irritation increased by an average of 1.23 points. Mentholone is the core regulator of cooling intensity, with an MD of 0.16 (95% CI [0.14, 0.17]), and for every 0.1 μg / mL increase in its content, the cooling intensity (out of 5) increased by an average of 0.16 points; Solanone, as a cooling precursor, had an MD value as high as 16.47 (95% CI [14.20, 18.75]), indicating that for every 1 μg / g increase in its content, the content of the cooling active ingredient after pyrolysis can increase by 16.47 μg / g. The 95% CI for all effect sizes did not include 0, confirming that the results were statistically significant.

[0100] Heterogeneity test: All flavor-causing substances showed significant heterogeneity (I²>50%, P<0.05). The sources of heterogeneity were directly related to the stability of the substances and the differences in experimental conditions. Specific indicators and sources are shown in Table 2.

[0101] Table 2 Results of Heterogeneity Test for Cool / Sweet Core Substances

[0102]

[0103] Effect size pooling: A random effects model was used to pool effect sizes. The core results are shown in Table 3. A forest plot was added using menthol as an example. Figure 4 As shown. Figure 4 As shown, the weight distribution of the four studies ranged from 22.1% to 28.4%, and there was no overlap in 95% CI among all studies. The pooled effect size was marked with a diamond notation "MD=0.16, I²=75.6%, T²=0.0002", and the CI did not include 0, indicating that the effect size was statistically significant and could quantify the contribution of menthol to the cooling sensation.

[0104] Table 3. Combined Results of Cooling Flavor Core Cooling Component Effect Quantities

[0105]

[0106] Next, publication bias is tested: specifically, a funnel plot symmetry test + Egger linear regression is used to assess publication bias. Taking menthol as an example, the funnel plot is as follows: Figure 5 As shown. Figure 5 Among the samples, the funnel plots showed an approximately symmetrical distribution, with only one small sample study (weight 17.9%) slightly deviating from the center line, presumably due to random error caused by differences in detection methods (titerization vs. GC-MS); Egger's linear regression test showed P=0.18 (P>0.05), indicating no significant publication bias. The funnel plots for the other four substances showed no significant asymmetry (Butylhydroxytoluene Egger test P=0.22, 2-pentylfuran P=0.25), with only 5-methylfurfural showing a slight deviation due to one high-dose study (addition amount >0.2 μg / g). After correction using the "cut-and-paste" method, the pooled effect size did not change significantly (MD=0.12, 95% CI: 0.09–0.15), indicating robust results.

[0107] Step S4: Select raw materials for production areas based on single-factor ANOVA and cluster analysis.

[0108] Specifically, based on single-factor ANOVA and cluster analysis, the main tobacco-producing areas are divided into three categories: high-sugar and low-alkali, high-alkali and medium-sugar, and low-sugar and low-alkali, so as to select tobacco raw materials from the corresponding producing areas according to the target flavor requirements.

[0109] Specifically, in some embodiments of the present invention, the core chemical indicators such as total sugar, nicotine, and sugar-to-nicotine ratio of 12 major tobacco-producing areas are used as analytical variables. The one-way ANOVA formula shown in formula (7) is used to test the differences in indicators between producing areas, with a significance level of α=0.05. If the differences between groups are significant (P<0.05), pairwise comparisons are further performed using the Tukey HSD post-hoc test to locate the specific differences in producing areas and clarify the differences in chemical characteristics of tobacco leaves from different producing areas.

[0110] The formula for the one-way ANOVA model is:

[0111] (7)

[0112] in, Represents the sum of squares between groups. The sum of squares within a group is represented by k; k represents the number of regions. This represents the total sample size. It follows a set of degrees of freedom (k-1, Nk). distributed.

[0113] Next, based on the core difference indicators (total sugar and nicotine) verified by ANOVA, the sample similarity of 12 production areas was calculated using Euclidean distance (Equation 8). System clustering was performed using the Ward method (minimizing the sum of squared deviations within groups) to provide a basis for selecting raw material production areas for different flavored cigarettes.

[0114] (8)

[0115] in, Indicates production area With production area The Euclidean distance; Indicates production area Total sugar content, Indicates production area Nicotine content; Indicates production area Total sugar content, Indicates production area Nicotine content.

[0116] The final analysis results are as follows: the tobacco-producing areas of Heilongjiang, Yunnan, and Fujian correspond to the high-sugar, low-alkali type, which is suitable for the development of sweet-flavored cigarettes; the tobacco-producing areas of Henan, Hunan, and Shandong correspond to the high-alkali, medium-sugar type, which is suitable for the development of strong-flavored cigarettes; the tobacco-producing areas of Inner Mongolia, Jilin, Guangdong, Guizhou, Shaanxi, and Chongqing correspond to the low-sugar, low-alkali type, which needs to be used after the chemical content is improved through planting technology optimization.

[0117] The selection of tobacco leaves from corresponding production areas based on target flavor requirements includes: tobacco leaves from high-sugar, low-alkali production areas with a total sugar content ≥23% and a nicotine content ≤1.8%; tobacco leaves from high-alkali, medium-sugar production areas with a nicotine content ≥2.4% and a total sugar content of 18%-21%; and tobacco leaves from low-sugar, low-alkali production areas, after optimization of planting techniques (such as adjusting fertilization plans and fermentation processes), with a total sugar content increased to over 20% and a nicotine content adjusted to 1.8%-2.2%.

[0118] In one embodiment of the present invention, the differences in eight major chemical components (total sugar, nicotine, total nitrogen, chlorine, protein, reducing sugar, pH value, and potassium) in tobacco leaves from 12 regions (Fujian, Guangdong, Guizhou, Henan, Heilongjiang, Hunan, Jilin, Inner Mongolia, Shandong, Shaanxi, Yunnan, and Chongqing) were systematically compared. The results are as follows: the regional differences in total sugar, nicotine, total nitrogen, chlorine, reducing sugar, pH value, and potassium were extremely significant, while the regional differences in protein were not significant.

[0119] Taking total sugar as an example (Table 4), the results of one-way ANOVA on the total sugar content of tobacco leaves showed that the regional effect was extremely significant (F=7.328, p<0.001). The between-group variance (60.96) was significantly higher than the within-group variance (8.32), indicating that there are statistically significant differences in the total sugar content of tobacco leaves in different regions.

[0120] Table 4. Results of one-way ANOVA of total sugar content

[0121]

[0122] Given the significant ANOVA results, pairwise comparisons (66 groups in total) were then performed using the Tukey HSD method at a pedigree error rate of α=0.05. Only the regions with significant differences are listed in Table 5.

[0123] Table 5. Comparison of regions with significant differences in total sugar content

[0124]

[0125] Furthermore, the Tukey-HSD multiple comparison revealed specific differences among several regions. The largest mean difference was observed between Heilongjiang and Henan (8.00%, p<0.001), while significant differences were also found between Yunnan and Henan (6.87%, p<0.001), and Henan and Fujian (-5.14%, p=0.0069). The results indicate that the total sugar content of Yunnan tobacco leaves is generally higher than that of other regions (such as Guangdong, Guizhou, and Henan), while the sugar content of Henan tobacco leaves is relatively lower. These differences may stem from the combined effects of factors such as climate, soil, or cultivation management, providing important evidence for regional studies of tobacco quality.

[0126] Finally, based on the conclusions of the above analysis of variance and related knowledge, we selected total sugar and nicotine as two indicators for cluster analysis, such as... Figure 6 As shown, the 12 regions are divided into 3 categories:

[0127] Category 1: Heilongjiang, Yunnan, and Fujian (high sugar, high reducing sugar), suitable for developing sweet-flavored cigarettes.

[0128] Category 2: Henan, Hunan, and Shandong (high nicotine, medium total sugar), suitable for developing cigarettes with strong stimulating properties.

[0129] Category 3: Other regions (low sugar, low nicotine) can improve the content of chemical components by improving planting techniques.

[0130] Step S5: Based on the identification results of the core flavoring substances, the optimal control range of the core flavoring substances, and the selection results of raw materials from the production area, conduct precise control of sensory quality.

[0131] In one embodiment of the cigarette flavor control method of the present invention, step S5 may specifically include:

[0132] Step S51: Construct a random forest regression model, using core flavoring substances as feature variables and sensory indicators as target variables, to determine the core regulatory factors of each sensory indicator.

[0133] In this study, a random forest regression model is constructed using core flavor compounds as characteristic variables (X) and sensory indicators as target variables (Y) to achieve precise control of sensory quality. The parameters for constructing the random forest regression model are: 500 decision trees, node splitting criterion of minimizing mean squared error (MSE), and the contribution of each chemical substance to the control of sensory indicators is quantified by calculating Mean Decrease MSE to determine core regulatory factors and provide targets for precise flavor control. The formula for calculating Mean Decrease MSE is as follows:

[0134] (8)

[0135] Where T represents the number of decision trees, This represents the mean square error of the original data. This represents the mean square error of the permutation data.

[0136] Furthermore, the core regulatory pathways were determined: based on the importance scores of the features, the core regulatory factors for each sensory indicator were identified. Specifically, the core regulatory factor for aroma quality was the nitrogen-to-base ratio (importance score 0.377), the core regulatory factor for aroma quantity was total nitrogen (importance score 0.334), the core regulatory factor for smoke concentration was total sugar (importance score 0.451), the core regulatory factor for combustibility was potassium (importance score 0.416), the core regulatory factor for off-odors was the nitrogen-to-base ratio (importance score 0.361), and the core regulatory factor for irritation was total nitrogen (importance score 0.283).

[0137] Step S52: By adjusting the content and ratio of core regulatory factors, the sweet aroma, cool sensation, and fruity sweetness flavors are precisely controlled respectively.

[0138] In step S52, multi-substance synergistic regulation is performed: based on the quantitative relationship of the core regulatory factors and combined with the synergistic effect rules of the three flavor categories, the flavor-inducing substances are precisely proportioned. Specifically, in the regulation of cooling flavor, menthol is added at 0.2-1.0 μg / mL, combined with 0.5-1.0 μg / g of solanone. When the mass ratio of cooling agent to sweet substance is 1:5, the synergistic effect is the strongest, the duration of cooling is increased by 40%, and the sweetness does not decrease significantly. In the synergistic regulation of fruit sweetness and cooling, the combination of phenylacetaldehyde and menthol can prolong the duration of cooling by 30%. The Maillard reaction product of banana extract combined with WS-23 can improve the coordination between sweetness and cooling by 2.3 points.

[0139] In sweet aroma control: total sugar is controlled at 18%-22%, sugar-to-alkali ratio is maintained at 8-12, 0.05-0.2μg / g of 5-methylfurfural and 0.02%-0.05% of maltol are added, and Maillard reaction (120-150℃, pH 5.0-5.5, 2-4h) is carried out to enhance the sweet aroma intensity.

[0140] In the sweetness regulation: benzyl alcohol and isoamyl acetate are compounded in a ratio of 2:1.86, and linalool is added to form a complex system of "benzyl alcohol-isoamyl acetate-linalool = 2:3:1", which, combined with the fermentation of aroma-producing microbial strains, enhances the realism of the sweetness.

[0141] In one embodiment of the present invention, to further explore the intrinsic relationship between the chemical content of cigarettes and sensory evaluation indicators, the contents of nine chemical substances (total sugar, reducing sugar, potassium, chlorine, nicotine, total nitrogen, sugar-to-alkali ratio, nitrogen-to-alkali ratio, and potassium-to-chlorine ratio) and ten sensory evaluation indicators (aroma quality, aroma quantity, fineness, concentration, off-flavors, strength, irritation, combustibility, dryness, and aftertaste) were ultimately determined. The data records for the contents of the following nine chemical substances are abundant and reliable, and can accurately reflect the actual situation of the relevant chemical substances in cigarettes; while the ten sensory evaluation indicators cover the main aspects of cigarette sensory evaluation and are highly representative. The random forest model identified the core regulatory substances for the ten sensory indicators, and the core results are summarized in Table 6. The results showed that the nitrogen-base ratio, with an importance score of 0.377, was the core regulatory factor for aroma quality. Total nitrogen (0.334) dominated aroma quantity, total sugar (0.451) had the most significant impact on smoke concentration, and potassium (0.416) was a key factor affecting combustibility. Among the negative sensory indicators, excessively high total nitrogen (0.283) exacerbated irritation, an imbalanced nitrogen-base ratio (0.361) easily produced off-flavors, and excessively low reducing sugar (0.277) led to a dry feeling. Chemical substances with importance scores between 0.1 and 0.2 constituted a secondary regulatory layer, synergistically influencing sensory quality with the core substances, while substances with scores below 0.1 mainly played a role through indirect regulation or risk control.

[0142] Table 6 Core Regulatory Substances for Sensory Indicators

[0143]

[0144] Furthermore, the method for controlling cigarette flavor of the present invention further includes:

[0145] Step S6: Verify the flavor control effect.

[0146] Specifically, a sensory evaluation method is used, with a smoking evaluation group of 7 to 10 people to score the regulated cigarettes on indicators such as sweetness, coolness, fruity sweetness, aroma, aroma quantity, irritation, and aftertaste. If the score of each indicator is ≥7.5 points (out of 10), the regulation is considered qualified.

[0147] In summary, this invention focuses on the analysis and quantitative regulation of the flavor-causing mechanisms of three categories of cigarette flavors: sweet aroma, coolness, and fruity sweetness. It integrates 218 valid domestic and international publications from 2004 to 2024, and through a progressive study using text analysis, meta-analysis, one-way ANOVA, and random forest models, clarifies the flavor-causing patterns and key regulatory points. The core conclusions are as follows:

[0148] I. Identify the core flavor-producing substances

[0149] By using tobacco-specific text mining technology, the core flavor compounds and research hotspots of three flavor categories were identified. The core research on sweet aroma focuses on the "natural sugars-Maillard products" system. The frequency of mention of "total sugars," "reducing sugars," and "sugar-base ratio" all exceeded 110 times. The TFIDF weights of Maillard reaction products such as "5-methylfurfural" and "maltol" exceeded 1.6, which are key substances for enhancing sweet aroma. The research on cooling aroma revolves around "traditional / novel cooling agents + mechanisms of action." "Menthol" was mentioned 186 times (coverage rate 82%), which is the traditional basic substance for cooling. Although "WS-23" was mentioned less frequently (98 times), its TFIDF weight reached 1.91, which is the core marker for identifying novel cooling agents. At the same time, the high frequency of "TRPM8 receptor" (68 times) reflects the key direction of cooling mechanism research. The research on fruity sweetness focuses on "esters / alcohols + generation pathways." The frequency of mention of "benzyl alcohol" and "isoamyl acetate" both exceeded 90 times. The TFIDF weights of "Maillard reaction" and "microbial aroma production" exceeded 1.7, which clarified the main generation pathways of fruity sweetness substances.

[0150] II. Quantification of Dose-Effect Relationship and Intensity Relationship

[0151] A random effects model was used to process highly heterogeneous data (I² > 65% for all substances), clarifying the quantitative correlation between the nine core substances and sensory effects. The specific dose-effect relationships and intensity relationships are shown in Table 7.

[0152] Table 7. Quantity-Effect Relationship and Intensity Relationship of Core Flavor Substances in Cigarettes

[0153]

[0154] The relative contribution of sensory intensity is based on the core substance in each flavor category as 1.0, and the remaining substances are converted according to the effect size ratio.

[0155] III. Differences in Production Areas and Clear Selection of Raw Materials

[0156] Analysis of eight core chemical indicators across 12 major tobacco-producing regions revealed highly significant regional differences in the remaining seven indicators (p<0.001), except for protein. In terms of total sugar content, Heilongjiang (28.0%) and Yunnan (26.87%) had significantly higher levels than other regions, while Henan had the lowest (20.0%), with a mean difference of 8.00% between Heilongjiang and Henan (p<0.001). Conversely, nicotine content was highest in Henan (2.81%) and Shandong (2.49%), and lowest in Heilongjiang (1.46%). Clustering based on the two core flavor indicators of total sugar and nicotine, the 12 regions can be divided into three categories: high-sugar, low-alkali (Heilongjiang, Yunnan, Fujian), suitable for developing sweet cigarettes; high-alkali, medium-sugar (Henan, Hunan, Shandong), suitable for strong-flavored cigarettes; and low-sugar, low-alkali (Inner Mongolia, Jilin, etc.), requiring optimization of planting techniques to increase component content. This provides a clear direction for selecting cigarette raw materials.

[0157] IV. Clear Sensory Quality Control Path

[0158] Random forest models reveal the weights of each chemical substance's influence on sensory quality, providing a clear path for precise control. Among positive sensory indicators, the nitrogen-to-alkali ratio (0.377) is the core influencing factor of aroma quality, total nitrogen (0.334) dominates aroma quantity, total sugar (0.451) has the greatest impact on smoke concentration, and total sugar (0.331) also significantly determines aftertaste quality. Among negative sensory indicators, an imbalance in the nitrogen-to-alkali ratio (0.361) is most likely to produce off-flavors, excessively high total nitrogen (0.283) will exacerbate irritation, and excessively low reducing sugar (0.277) will lead to a dry feeling in the mouth. Among neutral sensory indicators, total sugar (0.306) affects the fullness of the flavor, and potassium (0.416) is a key regulator of combustibility. When the potassium-to-chlorine ratio is greater than 4, it can effectively reduce the combustion failure rate.

[0159] The cigarette flavor control method provided in this invention integrates multidisciplinary technologies to construct a complete control system, achieving directional, quantitative, and efficient flavor control. Through the synergistic application of multiple methods, a complete technical system is constructed, from core substance identification and dose-effect relationship quantification to regional adaptation and quality control. The dose-effect and intensity relationships of flavor-producing substances in cigarettes are clarified, providing a solid scientific basis for the precise control of sweet aroma, coolness, and fruity sweetness in cigarettes, and also providing important technical support for the diversified innovation of flavors in the tobacco industry. It can solve the technical problems of existing cigarette flavor control methods, such as lack of scientific quantitative basis, poor regional adaptability, and unclear synergistic effects of multiple substances.

[0160] The embodiments of this disclosure have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0161] While specific embodiments of this disclosure have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for controlling cigarette flavor, characterized in that, include: A standardized flavor database is constructed, which includes material characteristics, experimental conditions, and sensory evaluation results. Based on the flavor standardization database, and using a tobacco text mining strategy, the core flavor compounds corresponding to each flavor are identified, including sweetness, coolness, and fruitiness. Meta-analysis was used to test the heterogeneity and combine the effect sizes of the identified core flavoring substances. A random effects model was used to determine the quantitative correlation between each substance and sensory effects in order to determine the optimal control range. Based on univariate ANOVA and cluster analysis, the raw material selection of the production area is carried out; Based on the identification results of the core flavor compounds, the optimal control range of the core flavor compounds, and the selection results of raw materials from the production area, the sensory quality is precisely controlled.

2. The method of modulating the flavor of a cigarette according to claim 1, wherein, The construction of the flavor standardization database includes: Based on domestic and international literature on the sweet aroma, coolness, and fruity sweetness of cigarettes, chemical index data, sensory evaluation data, production area data, and experimental condition data were extracted to establish a standardized database containing material characteristics, experimental conditions, and sensory evaluation. The chemical indexes include at least one of total sugar, reducing sugar, nicotine, total nitrogen, potassium, chlorine, sugar-to-alkali ratio, nitrogen-to-alkali ratio, menthol, benzyl alcohol, and isoamyl acetate.

3. The method of modulating the flavor of a cigarette according to claim 1, wherein, The tobacco text mining strategy includes: performing a global scan of text sets on sweet aroma, coolness, and fruity sweetness, retaining words whose occurrence frequency is greater than or equal to a preset frequency threshold, calculating the word frequency of words in a single document using the logarithmic TF method, introducing a smoothing term to calculate the inverse document frequency, calculating word weights based on word frequency and inverse document frequency, screening substances with word weights greater than or equal to a preset weight threshold as candidates for core flavor substances, and determining the final core flavor substances after a significance test.

4. The method of modulating the flavor of a cigarette according to claim 3, wherein The preset threshold number of times is 5. The method of calculating the word frequency in a single document using the logarithmic TF method includes: calculating the word frequency using the following formula: （1） Where TF represents word frequency, and i represents the number of times the same word appears in a single document. The calculation of inverse document frequency by introducing a smoothing term includes: calculating the inverse document frequency using the following formula: （2） Where IDF represents inverse document frequency, m represents the number of documents containing the term, and n represents the total number of documents. The calculation of lexical weights based on word frequency and inverse document frequency includes: calculating lexical weights using the following formula: TFIDF=TF×IDF (3) Here, TF-IDF represents word weights; the higher the TF-IDF value, the greater the representational value of the words to the text's topic. The preset weight threshold is 1.

6.

5. The method of modulating the flavor of a cigarette according to claim 3, wherein The step of identifying the core flavor compounds corresponding to each flavor based on the flavor standardization database and a tobacco text mining strategy includes: Based on the aforementioned tobacco text mining strategy, a tobacco-specific terminology database and a stop word database were constructed. Combining the bag-of-words model and the TFIDF algorithm, the core flavor compounds for three flavor categories were screened. The core flavor compounds corresponding to the sweet aroma flavor include total sugar, sugar-to-alkali ratio, 5-methylfurfural, and maltol. The core flavor compounds corresponding to the cool flavor are menthol, menthol, WS-23, and solanone. The core flavor compounds corresponding to the fruity sweet flavor are benzyl alcohol, isoamyl acetate, and linalool.

6. The method of modulating the flavor of a cigarette according to claim 1, wherein, The process involves meta-analysis to examine the heterogeneity and pool the effect sizes of the identified core flavor-causing substances. A random-effects model is then used to determine the quantitative correlation between each substance and sensory effects, in order to identify the optimal control range. This includes: The heterogeneity test in the meta-analysis was performed using a combination of the Q test and the I² statistic. When I² ≤ 50%, a fixed-effects model was used, and when I² > 50%, a random-effects model was used. The robustness of the results was verified by funnel plot symmetry test and Egger linear regression.

7. The method of modulating the flavor of a cigarette according to claim 1, wherein, The method of using a random effects model to determine the quantitative correlation between each substance and sensory effects, in order to determine the optimal control range, includes: The dose-response relationship for sweet aroma is as follows: for every 1-point increase in sugar-to-alkali ratio, the balance score between sweetness and astringency on a 5-point scale increases by 1.23 points, with an optimal control range of 812; for every 1-point increase in total sugar content, the average sweetness intensity on a 10-point scale increases by 0.85 points, with an optimal control range of 18%-22%; for every 1-point increase in reducing sugar content, the sweetness freshness score increases by 0.72 points, with an optimal control range of 16%-20%; for every 0.1 μg / g increase in 5-methylfurfural, the sweet aroma intensity score increases by 0.46 points, with an optimal control range of 0.05-0.2 μg / g; for every 0.01% increase in maltol addition, the sweetness persistence score increases by 0.58 points, with an optimal control range of 0.02%~0.05%. The dose-response relationship for the cooling flavor is as follows: for every 0.1 μg / mL increase in menthol content, the cooling intensity on a 5-point scale increases by an average of 0.16 points, with the optimal control range being 0.2-1.0 μg / mL; for every 1 μg / g increase in solanone content, the content of the active cooling ingredient after pyrolysis increases by 16.47 μg / g, with the optimal control range being 0.5-1.0 μg / g; for every 0.05 μg / g increase in 2-pentylfuran content, the cooling freshness score on a 10-point scale increases by 0.19 points, with the optimal control range being 0.05-0.2 μg / g; and for every 0.1 μg / g increase in the amount of butylated hydroxytoluene added, the cooling persistence is extended by an average of 0.75 minutes, with the optimal control range being 0.1-0.3 μg / g. The dose-response relationship corresponding to the fruity sweet flavor is as follows: when benzyl alcohol, linalool and geraniol are compounded in a ratio of 2:3:1, the fruity sweet flavor realism score is 8.6 points on a 10-point scale, and the harmony score is 9.2 points on a 10-point scale; when the mass ratio of isoamyl acetate to benzyl alcohol is 1:1.07, the fruity sweet flavor characteristics are most significant.

8. The method for controlling cigarette flavor according to claim 1, characterized in that, The method of selecting raw materials from production areas based on univariate ANOVA and cluster analysis includes: Based on univariate ANOVA and cluster analysis, the main tobacco-producing areas were divided into three categories: high-sugar and low-alkali, high-alkali and medium-sugar, and low-sugar and low-alkali, so as to select tobacco raw materials from the corresponding producing areas according to the target flavor requirements. Among them, the types corresponding to the main tobacco-producing areas of Heilongjiang, Yunnan and Fujian are high sugar and low alkali, which are suitable for the development of sweet cigarettes; the types corresponding to the main tobacco-producing areas of Henan, Hunan and Shandong are high alkali and medium sugar, which are suitable for the development of strong cigarettes; the types corresponding to the low sugar and low alkali of the main tobacco-producing areas are Inner Mongolia, Jilin, Guangdong, Guizhou, Shaanxi and Chongqing, which need to be used after the chemical composition content is improved through the optimization of planting technology. The selection of tobacco leaves from corresponding production areas based on target flavor requirements includes: tobacco leaves from high-sugar, low-alkali production areas with a total sugar content ≥23% and a nicotine content ≤1.8%; tobacco leaves from high-alkali, medium-sugar production areas with a nicotine content ≥2.4% and a total sugar content of 18%-21%; and tobacco leaves from low-sugar, low-alkali production areas, after optimization of planting techniques, with a total sugar content increased to over 20% and a nicotine content adjusted to 1.8%-2.2%.

9. The method of modulating the flavor of a cigarette according to claim 1, wherein, The precise control of sensory quality based on the identification results of core flavor compounds, the optimal control range of core flavor compounds, and the selection results of raw materials from the production area includes: A random forest regression model was constructed, with core flavoring substances as feature variables and sensory indicators as target variables, to determine the core regulatory factors of each sensory indicator. By adjusting the content and ratio of core regulatory factors, the sweet aroma, cool sensation, and fruity sweetness flavors can be precisely controlled.

10. The method of modulating the flavor of a cigarette according to claim 9, wherein, The construction parameters of the random forest regression model are as follows: the number of decision trees is 500, the node splitting criterion is minimizing the mean square error, and the core regulatory factors are determined by calculating the importance of features. Among them, the core regulatory factor for aroma quality is nitrogen-base ratio, the core regulatory factor for aroma quantity is total nitrogen, the core regulatory factor for smoke concentration is total sugar, the core regulatory factor for combustibility is potassium, the core regulatory factor for off-gas is nitrogen-base ratio, and the core regulatory factor for irritation is total nitrogen. The method involves precisely controlling the sweet aroma, cooling sensation, and fruity sweetness flavors by adjusting the content and ratio of core regulatory factors, including: In the regulation of cooling flavor, the synergistic effect is strongest when the mass ratio of cooling agent to sweet substance is 1:5, which increases the duration of cooling by 40% and does not significantly reduce the sweetness. When fruit sweetness and cooling are synergistically regulated, the combination of phenylacetaldehyde and menthol can prolong the duration of cooling by 30%. The Maillard reaction product of banana extract combined with WS-23 can improve the coordination between sweetness and cooling by 2.3 points.

Images