As consumer demand for tea beverage sensory experiences rises, flavor pairing theory becomes a key scientific basis for tea beverage innovation. This paper explores how the theory supports tea beverage blending by analyzing volatile compounds and their interactions, involving three key principles: shared flavor compounds, contrast, and synergy. Its implementation includes identifying aroma characteristics of raw materials, key aroma compounds and their senasory thresholds, and using data analysis and machine learning algorithms. Through application cases such as Ju-hong Tea, Osmanthus Black Tea, Polygonatum Pearl Milk Tea, and Blended Peach Oolong Tea, this study shows the theory’s practical results in tea beverage innovation. Flavor pairing theory provides strong momentum and abundant inspiration for innovation and upgrading in tea beverage products. It helps the tea beverage industry develop towards diversification and personalization, meets consumers’ diverse taste preferences, enhances product competitiveness, and promotes industry progress.

The safety supervision of food flavorings is of great significance for public health. The regulatory system in the United States, with its mature evaluation mechanism and transparent policies, has a significant global influence. The U.S. regulatory system originated from the Federal Food, Drug, and Cosmetic Act of 1938, and was further refined by the Food Additives Amendment of 1958, which introduced the definition of “generally recognized as safe” (GRAS) substances. The Expert Panel of the Flavor and Extract Manufacturers Association (FEMA) assesses flavor ingredients based on rigorous scientific criteria and publishes the results through the GRAS list. This system is not only widely applied in the United States but also has had a profound impact on the regulatory systems of international organizations and other countries. In contrast, the regulatory system of China for food flavorings has also been continuously improved to form a multi-department collaborative supervision model led by the State Administration for Market Regulation based on national standards and regulations, emphasizing the leading role of the government. Through comparative analysis, the differences in the regulatory systems of food flavorings between China and the United States are revealed, and suggestions for optimizing China’s regulatory system are put forward, with the aim of promoting the safe and orderly development of the food industry.

The objective of this study was to investigate the effects of natural fermentation and inoculated fermentations with Pichia pastoris and active dry wine yeast BV818 (Saccharomyces cerevisiae) on the flavor of yellow peach wine. Simultaneous inoculation, sequential inoculation of P. pastoris and BV818 with a time interval of one and three days, and inoculation of BV818 alone were used. Gas chromatography-mass spectrometry (GC-MS), gas chromatography-ion mobility spectrometry (GC-IMS) and quantitative descriptive sensory analysis were used to study the effects of different fermentation methods on volatile flavor substances in yellow peach wine. The results showed that for all groups, ethanol fermentation was completed within 15 days, and 31 and 22 compounds were detected by GC-MS and GC-IMS, respectively. In the sequential fermentation groups, the content of alcohols increased, the contents of acids and esters decreased, and the numbers of terpenes, alcohols and acids increased. In the sensory evaluation, the sequential fermentation groups had higher aroma richness and aroma balance than the other groups. The simultaneous fermentation group scored higher in sensory score for nutty and sweet aroma. The naturally fermented wine had the strongest sour aroma. In summary, mixed fermentation with P. pastoris and S. cerevisiae can improve the aroma of yellow peach wine, and the sensory score of the wine produced by sequential inoculation with a three-day interval was highest.

Food flavor plays an important role in people’s life. Traditional methods for flavor analysis and detection have limited ability to predict food flavor. In recent years, many researchers have used machine learning models to effectively process food flavor information and establish classification and prediction models, making flavor prediction more accurate and efficient. The principles of traditional and novel machine learning methods, such as support vector machine (SVM), random forest (RF), k-nearest neighbor (k-NN), and neural network, as well as recent progress on their combined application with flavor analysis instruments and molecular structure analysis for food flavor prediction are reviewed, aiming to provide new ideas for the application of machine learning models in food flavor analysis and prediction. It is found that machine learning models can be used to predict the impacts of different substance components on food flavor, identify the flavor characteristics of foods from different regions. The combination of multiple machine learning models can improve the accuracy and reliability of prediction, and promote in-depth research and development of food flavor.

Flavor plays a crucial role in the sensory perception of food and is a key determinant for consumer preference and choice. Therefore, flavor analysis methods are of paramount importance. Traditional methods for flavor analysis have limitations such as time-consuming and unable to handle large sample data. The emergence of machine learning is poised to address these problems. Machine learning possesses the capability to analyze and process vast amounts of data, identify complex patterns in high-dimensional variable spaces, autonomously learn useful information from known data, and automatically generate and optimize algorithms for prediction based on new data. The emergence of machine learning provides a new method for understanding the complex flavor characteristics of food. This article provides a comprehensive review of the advantages and disadvantages of traditional and novel machine learning methods as well as their various application scenarios in conjunction with analytical instruments such as electronic tongue, electronic nose, and gas chromatography-mass spectrometry (GC-MS). Additionally, it reviews the application of machine learning in food flavor analysis. Through research, it has been found that different scenarios of food flavor analysis require different machine learning methods. Machine learning holds significant potential for enhancing food quality, safety and consumer satisfaction. The combination of multiple machine learning models and analytical techniques will play a crucial role in food flavor analysis.

In this study, headspace solid phase microextraction in combination with comprehensive two-dimensional gas chromatography coupled to time-of-flight mass spectrometry (HS-SPME-GC × GC-TOFMS) was used to analyze the volatile compounds in whole (WM), low-fat milk (LFM) and non-fat milk (NFM). Altogether 49 volatile compounds were detected, among which methyl ketones with odd-numbered carbon chain lengths such as 2-nonanone and 2-undecanone constituted the main flavor compounds of WM. Using partial least squares discriminant analysis (PLS-DA), a model which could well differentiate among the 3 milks was developed and it was found to have good variance and cross-validation predictive ability. Nine differential key aroma compounds were identified using variable importance in the projection (VIP) > 1, P ≤ 0.05 and their contents ≥ 1% as criteria, which may be the main factors contributing to the differences in flavor profiles among the 3 milks. The heatmap from clustering analysis indicated that NFM had poor sensory performance due to the presence of off-flavor compounds (e.g., hexadecanal), whereas WM and LFM contained more aroma compounds, having a full and rich sensory aroma profile. The HS-SPME-GC × GC-TOFMS method can provide theoretical guidance for dairy flavor improvement and dairy flavoring formulation.