Determination of fat content in chicken hamburgers using NIR spectroscopy and the Successive Projections Algorithm for interval selection in PLS regression (iSPA-PLS)

Primary assessment of macronutrients in durian (CV Monthong) leaves using near infrared spectroscopy with wavelength selection

Farmers would be able to regulate fertilization and produce quality durian if they knew the nutrient concentration in durian leaves. A long period of time for traditional nutritional content determination is needed. Therefore, near-infrared spectroscopy is a good method for nondestructive and quick nutrient content evaluation. The leaf sample matrices (fresh leaves, dried ground leaves, and dried ground leaf pellets) were scanned by Fourier transform near-infrared (FT-NIR) with a wavelength of 12,500-3,600 cm-1. Regression models were developed using partial least squares (PLS) with full wavelength, short wavelength, and selected wavelength by successive projections algorithm (SPA). In this study, the model for N and K concentration was acceptable and the prediction was considered good but for P content not had succeeded. As a result, the PLS-SPA model using fresh leaf samples for evaluating N content in durian leaves exhibited performance of r2 = 0.852, SEP = 0.14%, RPD = 2.63 and bias = -0.020%. The PLS-SPA model using dried ground leaf samples for evaluating K content in durian leaves exhibited performance of r2 = 0.820, SEP = 0.13%, RPD = 2.36 and bias = 0.006%. This research found that it is possible to apply NIR waves to predict N and K concentrations in durian leaves. It is not necessary to predict directly from the wavelengths associated with -N or -K bonds. Instead, NIR can measure them indirectly from the bonding of proteins, which are products formed by N and K. In addition, selecting the wavelength that is related to the value to be measured can produce results that are not significantly different from using full or short wavelengths. These models can assist farmers in rapidly predicting N and K content in durian leaves for immediate fertilizer adjustment.

Detection of protein, starch, oil, and moisture content of corn kernels using one-dimensional convolutional autoencoder and near-infrared spectroscopy

Background

Analysis of the nutritional values and chemical composition of grain products plays an essential role in determining the quality of the products. Near-infrared spectroscopy has attracted the attention of researchers in recent years due to its advantages in the analysis process. However, preprocessing and regression models in near-infrared spectroscopy are usually determined by trial and error. Combining newly popular deep learning algorithms with near-infrared spectroscopy has brought a new perspective to this area.

Methods

This article presents a new method that combines a one-dimensional convolutional autoencoder with near-infrared spectroscopy to analyze the protein, moisture, oil, and starch content of corn kernels. First, a one-dimensional convolutional autoencoder model was created for three different spectra in the corn dataset. Thirty-two latent variables were obtained for each spectrum, which is a low-dimensional spectrum representation. Multiple linear regression models were built for each target using the latent variables of obtained autoencoder models.

Results

R², RMSE, and RMSPE were used to show the performance of the proposed model. The created one-dimensional convolutional autoencoder model achieved a high reconstruction rate with a mean RMSPE value of 1.90% and 2.27% for calibration and prediction sets, respectively. This way, a spectrum with 700 features was converted to only 32 features. The created MLR models which use these features as input were compared to partial least squares regression and principal component regression combined with various preprocessing methods. Experimental results indicate that the proposed method has superior performance, especially in MP5 and MP6 datasets.

A Review of Machine Learning for Near-Infrared Spectroscopy

The analysis of infrared spectroscopy of substances is a non-invasive measurement technique that can be used in analytics. Although the main objective of this study is to provide a review of machine learning (ML) algorithms that have been reported for analyzing near-infrared (NIR) spectroscopy from traditional machine learning methods to deep network architectures, we also provide different NIR measurement modes, instruments, signal preprocessing methods, etc. Firstly, four different measurement modes available in NIR are reviewed, different types of NIR instruments are compared, and a summary of NIR data analysis methods is provided. Secondly, the public NIR spectroscopy datasets are briefly discussed, with links provided. Thirdly, the widely used data preprocessing and feature selection algorithms that have been reported for NIR spectroscopy are presented. Then, the majority of the traditional machine learning methods and deep network architectures that are commonly employed are covered. Finally, we conclude that developing the integration of a variety of machine learning algorithms in an efficient and lightweight manner is a significant future research direction.

Rapid determination of moisture content of multi-source solid waste using ATR-FTIR and multiple machine learning methods

Rapid determination of moisture content plays an important role in guiding the recycling, treatment and disposal of solid waste, as the moisture content of solid waste directly affects the leachate generation, microbial activities, pollutants leaching and energy consumption during thermal treatment. Traditional moisture content measurement methods are time-consuming, cumbersome and destructive to samples. Therefore, a rapid and nondestructive method for determining the moisture content of solid waste has become a key technology. In this work, an attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) and multiple machine learning methods was developed to predict the moisture content of multi-source solid waste (textile, paper, leather and wood waste). A combined model was proposed for moisture content regression prediction, and the applicability of 20 combinations of five spectral preprocessing methods and four regression algorithms were discussed to further improve the modeling accuracy. Furthermore, the prediction result based on the water-band spectra was compared with the prediction result based on the full-band spectra. The result showed that the combination model can efficiently predict the moisture content of multi-source solid waste, and the R² values of the validation and test datasets and the root mean square error for the moisture prediction reached 0.9604, 0.9660, and 3.80, respectively after the hyperparameter optimization. The excellent performance indicated that the proposed combined models can rapidly and accurately measure the moisture content of solid waste, which is significant for the existing waste characterization scheme, and for the further real-time monitoring and management of solid waste treatment and disposal process.

Qualitative and Quantitative Detection of Acacia Honey Adulteration with Glucose Syrup Using Near-Infrared Spectroscopy

Honey adulteration with cheap sweeteners such as corn syrup or invert syrup results in honey of lesser quality that can harm the objectives of both manufacturers and consumers. Therefore, there is a growing interest for the development of a fast and simple method for adulteration detection. In this work, near-infrared spectroscopy (NIR) was used for the detection of honey adulteration and changes in the physical and chemical properties of the prepared adulterations. Fifteen (15) acacia honey samples were adulterated with glucose syrup in a range from 10% to 90%. Raw and pre-processed NIR spectra of pure honey samples and prepared adulterations were subjected to Principal Component Analysis (PCA), Partial Least Squares (PLS) regression, and Artificial Neural Network (ANN) modeling. The results showed that PCA ensures distinct grouping of samples in pure honey samples, honey adulterations, and pure adulteration using NIR spectra after the Multiplicative Scatter Correction (MSC) method. Furthermore, PLS models developed for the prediction of the added adulterant amount, moisture content, and conductivity can be considered sufficient for screening based on RPD and RER values (1.7401 < RPD < 2.7601; 7.7128 < RER < 8.7157) (RPD of 2.7601; RER of 8.7157) and can be moderately used in practice. The R2validation of the developed ANN models was greater than 0.86 for all outputs examined. Based on the obtained results, it can be concluded that NIR coupled with ANN modeling can be considered an efficient tool for honey adulteration quantification.

Germplasm variability-assisted near infrared reflectance spectroscopy chemometrics to develop multi-trait robust prediction models in rice

Rice is a major staple food across the world in which wide variations in nutrient composition are reported. Rice improvement programs need germplasm accessions with extreme values for any nutritional trait. Near infrared reflectance spectroscopy (NIRS) uses electromagnetic radiations in the NIR region to rapidly measure the biochemical composition of food and agricultural products. NIRS prediction models provide a rapid assessment tool but their applicability is limited by the sample diversity, used for developing them. NIRS spectral variability was used to select a diverse sample set of 180 accessions, and reference data were generated using association of analytical chemists and standard methods. Different spectral pre-processing (up to fourth-order derivatization), scatter corrections (SNV-DT, MSC), and regression methods (partial least square, modified partial least square, and principle component regression) were employed for each trait. Best-fit models for total protein, starch, amylose, dietary fiber, and oil content were selected based on high RSQ, RPD with low SEP(C) in external validation. All the prediction models had ratio of prediction to deviation (RPD) > 2 amongst which the best models were obtained for dietary fiber and protein with R² = 0.945 and 0.917, SEP(C) = 0.069 and 0.329, and RPD = 3.62 and 3.46. A paired sample t-test at a 95% confidence interval was performed to ensure that the difference in predicted and laboratory values was non-significant.

Direct authentication and composition quantitation of red wines based on Tri-step infrared spectroscopy and multivariate data fusion

A robust data fusion strategy integrating Tri-step infrared spectroscopy (IR) with electronic nose (E-nose) was established for rapid qualitative authentication and quantitative evaluation of red wines using Cabernet Sauvignon as an example. The chemical fingerprints of four types of wines were thoroughly interpreted by Tri-step IR, and the defined spectral fingerprint region of alcohol and sugar was 1200-950 cm^-1. The wine types were authenticated by IR-based principal component analysis (PCA). Furthermore, ten quantitative models by partial least squares (PLS) were built to evaluate alcohol and total sugar contents. In particular, the model based on the fusion datasets of spectral fingerprint region and E-nose was superior to the others, in which RMSEP reduced by 47.95% (alcohol) and 79.90% (total sugar), r_p increased by 11.95% and 43.47%, and RPD >3.0. The developed methodology would be applicable for mass screening and rapid multi-chemical-component quantification of wines in a more comprehensive and efficient manner.

Agricultural Potentials of Molecular Spectroscopy and Advances for Food Authentication: An Overview

Meat, fish, coffee, tea, mushroom, and spices are foods that have been acknowledged for their nutritional benefits but are also reportedly targets of fraud and tampering due to their economic value. Conventional methods often take precedence for monitoring these foods, but rapid advanced instruments employing molecular spectroscopic techniques are gradually claiming dominance due to their numerous advantages such as low cost, little to no sample preparation, and, above all, their ability to fingerprint and detect a deviation from quality. This review aims to provide a detailed overview of common molecular spectroscopic techniques and their use for agricultural and food quality management. Using multiple databases including ScienceDirect, Scopus, Web of Science, and Google Scholar, 171 research publications including research articles, review papers, and book chapters were thoroughly reviewed and discussed to highlight new trends, accomplishments, challenges, and benefits of using molecular spectroscopic methods for studying food matrices. It was observed that Near infrared spectroscopy (NIRS), Infrared spectroscopy (IR), Hyperspectral imaging (his), and Nuclear magnetic resonance spectroscopy (NMR) stand out in particular for the identification of geographical origin, compositional analysis, authentication, and the detection of adulteration of meat, fish, coffee, tea, mushroom, and spices; however, the potential of UV/Vis, 1H-NMR, and Raman spectroscopy (RS) for similar purposes is not negligible. The methods rely heavily on preprocessing and chemometric methods, but their reliance on conventional reference data which can sometimes be unreliable, for quantitative analysis, is perhaps one of their dominant challenges. Nonetheless, the emergence of handheld versions of these techniques is an area that is continuously being explored for digitalized remote analysis.

Prediction of surface tension of solution in the presence of hydrophilic silica nanoparticle and anionic surfactant by ATR-FTIR spectroscopy and chemometric methods

In the current research, an analytical method was proposed for the quantitative determination of surface tension of anionic surfactant solutions in the presence of hydrophilic silica nanoparticles using attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy and chemometric methods. The surface tension behavior of anionic surfactant solutions considerably changes by the addition of silica nanoparticles with different particle size. The spectral data of solutions were used for prediction of surface tension using two calibration methods based on support vector machine regression (SVM-R) as a non-linear algorithm and partial least squares regression (PLS-R) as a linear algorithm. For preprocessing of data, baseline correction and standard normal variate (SNV) were also applied. Root mean square error of prediction (RMSEP) in SVM-R and PLS-R methods were 4.203 and 4.507, respectively. Considering the complexity of the samples, the SVM-R model was found to be reliable. The proposed method is fast and easy for measurement of the surface tension of surfactant solutions without any sample preparation step in chemical enhanced oil recovery (C-EOR).

Prediction of Soluble Solids Content in Green Plum by Using a Sparse Autoencoder

The soluble solids content (SSC) affects the flavor of green plums and is an important parameter during processing. In recent years, the hyperspectral technology has been widely used in the nondestructive testing of fruit ingredients. However, the prediction accuracy of most models can hardly be improved further. The rapid development of deep learning technology has established the foundation for the improvement of building models. A new hyperspectral imaging system aimed at measuring the green plum SSC is developed, and a sparse autoencoder (SAE)–partial least squares regression (PLSR) model is combined to further improve the accuracy of component prediction. The results of the experiment show that the SAE–PLSR model, which has a correlation coefficient of 0.938 and root mean square error of 0.654 for the prediction set, can achieve better performance for the SSC prediction of green plums than the three traditional methods. In this paper, integration approaches have combined three different pretreatment methods with PLSR to predict the SSC in green plums. The SAE–PLSR model has shown good prediction performance, indicating that the proposed SAE–PLSR model can effectively detect the SSC in green plums.

Application of Vis-NIR spectroscopy for determination the content of organic matter in saline-alkali soils

Visible and near-infrared diffuse reflectance spectroscopy (Vis-NIR) has been recognized as a fast method to evaluate the content of soil organic matter (SOM) in various types of soil. The accuracy of Vis-NIR is comparable to conventional laboratory methods for estimating SOM. However, very few studies have applied Vis-NIR to estimate SOM in saline-alkali soil. This study aimed to investigate the efficiency of spectral data for evaluating SOM in saline-alkali soil. Soil samples (n = 291) were collected from the five major saline-alkali soil regions in Shaanxi. SOM was measured using standard methods and the samples were scanned using ASD Fieldspec4 at wavelength of 350-2500 nm to obtain spectral data. Twenty-six pre-processing methods were tested and partial least squares regression (PLSR) was used to estimate SOM. The best preprocessing was R + SG + SNV + FD. The calibration results were Pc = 15, Rc² = 0.92, RMSEC = 1.11, SEC = 1.12, Slope = 0.92, Offset = 0.45; the validation results were Rv² = 0.97, RPD = 5.21, RMSEP = 0.38, SEP = 0.38, Slope = 0.97, Offset = 0.17. Therefore, this main objective of the study was to propose an effective approach based on Vis-NIR spectroscopy and Chemometrics for predicting saline-alkali SOM contents in the center of Shaanxi, China.

Rapid spectral analysis of agro-products using an optimal strategy: dynamic backward interval PLS-competitive adaptive reweighted sampling

A novel strategy of variable selection approach named dynamic backward interval partial least squares-competitive adaptive reweighted sampling (DBiPLS-CARS) was proposed in this study. Near-infrared data sets of three different agro-products, namely corn, crop processing lamina, and plant leaf samples, were collected to investigate the performance of the proposed method. Weak relevant variables were first removed by DBiPLS and a refined selection of the remaining variables was then conducted by CARS. The Monte Carlo uninformative variable elimination (MCUVE) was used as a classical beforehand uninformative variable elimination method for comparison. Results showed that DBiPLS can select informative variables more continuously than MCUVE. Some synergistic variables which may be omitted by MCUVE can be retained by DBiPLS. By contrast, MCUVE can hardly avoid the disturbance of certain weak relevant variables as a result of its calculation based on the full spectrum regression. Therefore, DBiPLS exhibited the advantage of removing the weak relevant variables before CARS, and simultaneously improved the prediction performance of CARS.

Comparison of near-infrared (NIR) and mid-infrared (MIR) spectroscopy for the determination of nutritional and antinutritional parameters in common beans

Common beans (Phaseolus vulgaris L.), represent the most consumed legume worldwide and constitute an important source of protein, being also known to contain antinutritional compounds, which compromise nutrients' bioavailability. However, the standard methodologies to assess these constituents are time-consuming and complex. Therefore, the present study evaluated the suitability of near-infrared (NIR) and mid-infrared (MIR) spectroscopies for the development of simple and reliable methods to assess protein, lipids, tannins and phytic acid contents, besides specific amino acids, in whole bean flours. Partial least squares (PLS) regression was used to develop analytical models, and external validation was performed. NIR displayed better performance for the evaluation of protein, lipids, tannins and phytic acid contents, and MIR, for the assessment of specific amino acids. In both techniques, the use of the 1st derivative was the best data treatment. Overall, both techniques represent reliable methods to evaluate the proximate and antinutritional composition of bean flours.

ATR-FTIR spectroscopy and chemometric techniques for determination of polymer solution viscosity in the presence of SiO2 nanoparticle and salinity

An analytical method was proposed for quantitative determination of rheological properties of polyacrylamide (PAM) solution in the presence of SiO₂ nanoparticle and NaCl. The viscosity of PAM-SiO₂ nanohybrid solution was predicted using attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy in the wavenumber range of 800-3000 cm^-1 and chemometrics methods. Support vector machine regression (SVM-R) as a non-linear multivariate calibration procedure and partial least squares regression (PLS-R) as a linear procedure were applied for calibration. Preprocessing methods such as baseline correction and standard normal variate (SNV) were also utilized. Root mean square error of prediction (RMSEP) in SNV-SVM and SNV-PLS methods were 3.231 and 6.302, respectively. Considering the complexity of the samples, the SVM-R model was found to be reliable. The proposed method is rapid and simple without any sample preparation step for measurement of the viscosity of polymer solutions in chemical enhanced oil recovery (CEOR).

Selection of characteristic wavelengths using SPA for laser induced fluorescence spectroscopy of mine water inrush

In the process of prevention and control of water inrush disaster, it is of great significance to identify the type of water inrush source for coal mine safety production accurately and quickly. The application of laser induced fluorescence (LIF) technology to identify the water inrush in coal mine broke the shortage of the traditional hydrochemical method, which could realize the accurate and rapid identification of water inrush types. Firstly, in order to avoid the influence of random variations of spectral data, four kinds of common pretreatment methods were analyzed and compared, and the moving average smoothing method was chosen to preprocess the original fluorescence spectral data. Then, for the purpose of selecting the appropriate sample division method to improve the predictive performance of the model, four common sample division methods were compared, and the sample set partitioning based on joint x-y distance (SPXY) method was used to divide the samples into training set and test set. Further, the 10 characteristic wavelengths were selected by successive projections algorithm (SPA) to reduce the amount of data. Finally, the selected data was taken as input, the sigmoid function was selected as the activation function of extreme learning machine (ELM), and the number of hidden layer neurons was set to 34, which realized the construction of water source identification model. The prediction accuracy of ELM model for the training set and test set were 99.0% and 94.0%, respectively. In addition, the water samples collected at different time were mixed in the same way to form the independent verification set, and the prediction accuracy of the ELM water source identification model for independent verification set was 91.5%. The results shown that it was feasible to select the characteristic wavelengths of fluorescence spectra by using the SPA. The data of 10 characteristic wavelengths could fully represent the effective information of whole band spectrum. And it also provided a theoretical basis for the development of a special online identification instrument for mine water inrush.

Application of near-infrared spectroscopy for screening the potato flour content in Chinese steamed bread

Near-infrared (NIR) spectroscopy combined with chemometrics was used as a technique to predict the potato flour content in Chinese steamed bread (CSB). The inner core of CSB was chosen as the measuring position for acquiring the NIR spectra. Spectra between 4000 and 10,000 cm^-1 were analysed using a partial least-squares regression. The coefficient of determination (R _CV ²) and the root mean square error of cross-validation in the calibration set were found to be 0.7940-0.8955 and 4.22-5.93, depending on the pre-treatment of the spectra. The external validation set gave an R² and a ratio to performance deviation of 0.8865 and 3.07. Reasonable recovery (93.1-102.5%) and good intra-assay (3.3-8.3%) and inter-assay (7.6-17.2%) precision illustrated the feasibility of this method. The result of this study reveals that NIR spectroscopy could be used as rapid tool to determine the potato flour content in CSB (> 20%).

Quantitative Analysis of Cadmium in Tobacco Roots Using Laser-Induced Breakdown Spectroscopy With Variable Index and Chemometrics

The study investigated some new developed variable indices and chemometrics for the fast detection of cadmium (Cd) in tobacco root samples by laser-induced breakdown spectroscopy. The variables selection methods of interval partial least squares (iPLS), backward interval partial least squares (BiPLS), and successive projections algorithm (SPA) were used to locate the optimal Cd emission line for univariate analysis and to select the maximal relevant variables for multivariate analysis. iPLS and BiPLS located 10 Cd emission lines to establish univariate analysis models. Univariate analysis model based on Cd I (508.58 nm) performed best with the coefficient of determination of prediction (R_p ²) of 0.9426 and root mean square error of prediction (RMSEP) of 1.060 mg g^-1. We developed two new variable indices to remove negative effects for Cd content prediction, including Index1 = (I _508.58 + I _361.05)/2 × I _466.23 and Index2 = I _508.58/I _466.23 based on Cd emission lines at 508.58, 361.05, and 466.23 nm. Univariate model based on Index2 obtained better result (R_p ² of 0.9502 and RMSEP of 0.988 mg g^-1) than univariate analysis based on the best Cd emission line at 508.58 nm. PLS and support vector machines (SVM) were adopted and compared for multivariate analysis. The results of multivariate analysis outperformed univariate analysis and the best quantitative model was achieved by the iPLS-SVM model (R_c ² of 0.9820, RMSECV of 0.214 mg g^-1, R_p ² of 0.9759, and RMSEP of 0.712 mg g^-1) using the maximal relevant variables in the range of 474-526 nm. The results indicated that LIBS coupled with new developed variable index and chemometrics could provide a feasible, effective, and economical approach for fast detecting Cd in tobacco roots.

Comparing causal techniques for rainfall variability analysis using causality algorithms in Iran

Causal analysis (CA) is a strong quantitative approach whose mechanisms have climatic predictions. In this study, we studied the patterns of causality (PC) on the effect of rainfall (ER) using climatic series collected from 170 stations for the period 1975-2014 in Iran. Next, we predicted the causal relationships of climatic variables using causal models, including first-generation techniques (FGT), second-generation techniques (SGT), third-generation techniques (TGT), and causal hybrid techniques (CHT). Then, we estimated the causal models using partial squares algorithms (PSA), mechanical equations modeling algorithms (MEMA) such as exploratory and confirmatory methods, and spatial variability methods such as geostatistics and spatial statistical methods. Finally, we evaluated the quality of the methods using the goodness of fit indices, including absolute fit indices (AFI), comparative fit indices (CFI), and parsimonious fit indices (PFI). The results showed that CHT algorithm more suitably predicted the climatic spatiotemporal effect variability (SEV) by extracting direct, indirect, and total effects of climatic variables. Based on the CHT algorithm, the highest and lowest effect values were observed in total effects of winter rainfall (0.98) and summer rainfall variables (0.1), respectively. The SEV ranged from 0.8 to 0.98 for the winter rainfall total effects of CHT in Iran. Using CHT, most of the predicted SEV, particularly the rainfall series, displayed SEV varying from 80% to 98% of the winter rainfall total effects to the annual rainfall in Iran. Similarly, based on the CHT, the highest and lowest SEV values were in western, eastern, and southern regions and in central regions, respectively. In addition, the SEV varied within the range of 0.6-0.74 (varying from 60% to 74% for the autumn rainfall total effects of the annual rainfall in Iran) for the autumn rainfall total effects in Iran. Finally, the SEV of this type of analytical pattern as well as designated subject of CA applications in the atmospheric science and environmental science are discussed.

Simultaneous measurement of contents of liquirtin and glycyrrhizic acid in liquorice based on near infrared spectroscopy

OBJECTIVE

To establish calibration models for simultaneous determination of contents of liquirtin and glycyrrhizic acid, and to investigate the variable selection methods.

METHODS

The contents of liquirtin and glycyrrhizic acid determined by HPLC were as the reference values, which were associated with samples spectra by using near infrared spectrum (NIR) analysis technology. Calibration models were developed using partial least squares (PLS) regression algorithm, and evaluated by the independent dataset test with calculating the metrics of coefficients of determination of calibration and prediction (R²_c, R²_p), the root mean square errors of calibration and prediction (RMSEC, RMSEP), the mean absolute errors of calibration and prediction (MAEC, MAEP), and the residual prediction deviation (RPD). Five variable selection methods including variable importance in projection (VIP), competitive adaptive reweighted sampling (CARS), Monte Carlo uninformative variable elimination (MCUVE), particle swarm optimization (PSO) and genetic algorithm (GA), were investigated.

RESULTS

Compared to the original full spectra, both quantification models for liquirtin and glycyrrhizic acid performed better with a clear ranking of GA>PSO>CARS>MCUVE≅VIP>Full. Especially for GA-PLS models, RMSEC and RMSEP were <0.05%, R²_c and R²_p were >0.94, and RPD were both >4, indicating that both the models had good robustness and excellent prediction accuracy.

CONCLUSION

The present calibration models can be utilized to simultaneously determine the contents of liquirtin and glycyrrhizic acid in liquorice samples, and thus are of great help for rapid quality evaluation and control of liquorice.