Multi-dimensional feature associated agricultural product repurchase willingness weak supervision analysis method and system

Abstract

The invention belongs to the technical field of natural language processing, and particularly relates to a multi-dimensional feature associated agricultural product repurchase willingness weak supervision analysis method and a multi-dimensional feature associated agricultural product repurchase willingness weak supervision analysis system. The method comprises the following steps: dividing an agricultural product evaluation data set into a comment text list and a score list, mining factors influencing a repurchase intention, and determining a fine-grained evaluation dimension; extracting triple information from the comment text, and processing to obtain a fine-grained aspect item list and a fine-grained opinion item list; then generating a pseudo-label data set, and correcting a pseudo-label training set by using a pseudo-label correction strategy embedded with confidence learning and multi-dimensional multi-label deep learning model preliminary training; the contribution evaluation model quantifies the contribution of each repurchasing willingness influence factor; and training under a weak supervision framework to mine multi-dimensional emotion in the comment text. According to the method, the problem of label missing in each feature dimension is solved, customer demands can be accurately grasped, and a targeted resource allocation strategy is further formulated.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural language processing, and specifically relates to a weakly supervised analysis method and system for agricultural product repurchase intention associated with multi-dimensional features. Background Art

[0002] With the rapid development of e-commerce platforms and the arrival of the big data era, online sales and purchases of agricultural products have become a widely accepted and increasingly popular trading method. While this sales-purchase method provides convenience for people, it also brings some problems that need to be considered and solved. Consumers' repurchase rate of agricultural products is often affected by multi-dimensional factors, such as product quality, merchant services, packaging appearance, etc. It is particularly important for merchants to grasp the various needs of customers and formulate corresponding resource allocation strategies to increase the repurchase rate of products. Consumers also often hope to quickly select agricultural products that meet their wishes from a wide range of stores and products. In order to build a communication bridge between merchants and customers, major e-commerce platforms have set up comment functions so that consumers can give feedback on their opinions on agricultural products and rate their shopping experience. The scores given by consumers intuitively reflect their repurchase willingness: the higher the score, the stronger the repurchase willingness. The huge volume of reviews provides a more comprehensive and powerful basis for merchants to analyze customer needs and consumer product selection. However, it is not easy for merchants to fully grasp the key factors that affect customers' willingness to repurchase. It is also difficult for customers to intuitively analyze how various products perform in the aspects they value. Traditional data analysis methods often cannot quantify the content reflected in the review text. This is an urgent problem to be solved for the development of agricultural product transactions on e-commerce platforms.

[0003] In this context, natural language processing tasks in machine learning and deep learning have been introduced into product review analysis and have achieved remarkable results in recent years. Deep learning models can learn the rules between semantic information and sentiment from a large amount of text data with sentiment labels, and apply them to text without sentiment labels to help people efficiently and intuitively analyze the sentiment information contained therein. In combination with relevant corpora and using sentiment analysis tools with functions such as part-of-speech tagging and sentiment intensity perception, factors in reviews that are associated with repurchase intention can be captured.

[0004] However, the results obtained do not directly reflect the impact of each factor on repurchase intention in the overall situation, and merchants cannot formulate reasonable resource allocation strategies based on this. Moreover, the agricultural product evaluation dataset obtained from the e-commerce platform only contains the overall sentiment score, and does not involve the scores of specific dimensions. There is a lack of fine-grained labels, which is an obstacle to quantifying the contribution of each influencing factor and building a multi-dimensional and multi-label deep learning model. Summary of the invention

[0005] Based on this, the present invention proposes a weakly supervised analysis method and system for agricultural product repurchase intention associated with multi-dimensional features, uses a topic model to model short texts to capture the factors that affect the repurchase intention and determine the fine-grained evaluation dimensions accordingly, generates pseudo labels on various coarse and fine granularity dimensions by extracting sentiment triplets from the comment text, constructs a multi-task deep learning model and corrects the pseudo labels in combination with confidence learning to obtain multi-dimensional pseudo labels with relatively high credibility. A global interpretation method based on game theory is applied to the machine model to construct a global contribution evaluation model to obtain a quantitative representation of the contribution of each factor to the repurchase intention. Then, a multi-dimensional and multi-label deep learning model is trained under a weakly supervised framework to realize the mining of fine-grained sentiment tendencies in comment texts.

[0006] The present invention is achieved through the following technical solutions.

[0007] A weakly supervised analysis method for agricultural product repurchase intention based on multi-dimensional feature associations, the steps are as follows: Step 1: Obtain review texts and ratings from e-commerce platforms to construct an agricultural product evaluation dataset; Step 2: Divide the agricultural product evaluation dataset into a review text list and a rating list, create a review noun matrix and a noun dictionary based on the review text list, and then convert the review text into a bag-of-words representation matrix based on the noun dictionary and the review noun matrix; then perform topic modeling to obtain the topic-noun distribution, thereby mining the factors that affect repurchase intention and determining the fine-grained evaluation dimensions; Step 3: Use the aspect-sentiment triple extraction model based on span-level interactive learning mechanism to extract triple information from the review text, including aspect items, opinion items and sentiment polarity; Step 4: Remove duplicates from the extracted aspect items and opinion items and classify them separately to obtain a fine-grained aspect item list and a fine-grained opinion item list; Step 5: Generate a pseudo-labeled dataset based on the extracted triple information, fine-grained aspect item list, and fine-grained opinion item list. Extract data and manually correct it as a clean test set, and use the rest as a pseudo-labeled training set. Step 6: Build a multi-dimensional multi-label deep learning model based on convolutional neural network and Transformer, and use the pseudo-label correction strategy embedded with confidence learning and preliminary training of the multi-dimensional multi-label deep learning model to correct the pseudo-label training set; Step 7: Obtain the data set required for the contribution evaluation model, use the histogram-based gradient boosting classifier, and further apply the feature contribution estimation method based on game theory to the histogram gradient boosting classifier, and use the contribution evaluation model to quantify the contribution of each repurchase intention influencing factor; Step 8: Pass in the corrected pseudo-label dataset and train a multi-dimensional multi-label deep learning model under a weakly supervised framework to mine the multi-dimensional emotions in the comment text.

[0008] Further preferably, in step 2, a latent Dirichlet distribution model is used for topic modeling.

[0009] Further preferably, the process of extracting triplet information from the review text using the aspect sentiment triplet extraction model based on the span-level interactive learning mechanism in step 3 is as follows: Traverse each comment text, perform word segmentation encoding, use the pre-trained BERT model to capture the contextual representation of the comment text and select the last layer shape as the hidden state, loop through, and generate all possible spans in the comment text; Classify spans: obtain the probability distribution of all spans belonging to different span categories and predict whether the span is a valid aspect item or opinion item; The number of aspect items and opinion item spans that need to be retained for a given review text is determined based on the sequence length and pruning threshold, and then a dual-channel pruning strategy is adopted to select the top 10 aspect items and opinion item spans in the probability distribution of span categories. The span index that is most likely to be an aspect item and the span index that is most likely to be an opinion item; Construct a joint representation matrix for each pair of aspect terms and opinion terms in the review text; The sentiment polarity is judged on the joint representation matrix of aspect item-opinion item pairs, and the probability distribution of all aspect item-opinion item pairs belonging to different sentiment polarity categories is obtained.

[0010] Further preferably, in step four, firstly, the aspect items and opinion items are deduplicated respectively; then, the aspect items containing the same root word are grouped into one category; for the opinion items, after removing the adverbs expressing the function, the word form is restored and the items containing the same root word are grouped into one category; further, after manual adjustment and screening, a fine-grained aspect item list and a fine-grained opinion item list of each fine-grained evaluation dimension are generated.

[0011] Further preferably, the specific process of step five is as follows: First, set initial labels for each fine-grained evaluation dimension of each review; Then define three types of pseudo labels and sentiment polarity-pseudo label conversion function to convert sentiment polarity into corresponding pseudo labels; Next, the fine-grained initial labels corresponding to all comment texts are modified based on the pseudo-label judgment method; Finally, pseudo labels on coarse-grained dimensions are generated based on the triple information to obtain a pseudo label dataset; Some data are extracted from the pseudo-label dataset and manually corrected as a pure test set, and the remaining data is used as a pseudo-label training set.

[0012] Further preferably, the specific process of step six is ​​as follows: The pseudo-label training set is passed in, and the text classifier based on the Fasttext model is trained for each dimension separately, and then the confidence learning tool is used to find low-confidence samples; A multi-dimensional and multi-label deep learning model based on convolutional neural networks and Transformer is constructed. Data is extracted from the pseudo-label training set for manual annotation and passed into the constructed multi-dimensional and multi-label deep learning model for training. A preliminary model is obtained according to the accuracy of the training set and the preliminary model is used to predict the low-confidence labels of low-confidence samples to obtain the model labels. The model labels of the low-confidence samples are then compared with the true labels recognized by the text classifier based on the Fasttext model for the low-confidence samples. It is stipulated that: if the model label is consistent with the true label recognized by the text classifier based on the Fasttext model, the corrected label is the model label. If the two are inconsistent, they are corrected manually. The samples corresponding to the corrected labels and the high-confidence samples together constitute the corrected pseudo-label training set.

[0013] Further preferably, the specific process of step seven is as follows: Generate labels for all review texts in the corrected pseudo-label training set and the clean test set on the repurchase intention dimension based on the score list; then transform the pseudo-labels on all fine-grained evaluation dimensions in the corrected pseudo-label training set and the clean test set to meet the data requirements of the contribution evaluation model; splice the data sets that constitute the contribution evaluation model, and divide them into the training set and test set of the contribution evaluation model; Construct a contribution evaluation model, pass in the training set of the contribution evaluation model, use the pseudo labels on the fine-grained dimension as input features, use the labels on the repurchase intention dimension as target labels, traverse the hyperparameter grid and evaluate the performance of the gradient boosting classifier under each parameter combination through five-fold cross validation, and select the hyperparameter combination that makes the gradient boosting classifier perform best on multiple data subsets; The training set is then fed back into the contribution evaluation model to quantify the contribution of each factor influencing repurchase intention.

[0014] Further preferably, in step eight, the manually corrected pure test set and the corrected pseudo-label training set together constitute a corrected pseudo-label data set, which is used as the data set for the multi-dimensional and multi-label deep learning model.

[0015] The present invention also provides a weakly supervised analysis system of agricultural product repurchase intention associated with multi-dimensional features, the system comprising a memory, a processor and a computer program stored in the memory, and the computer program, when executed by the processor, implements the various steps of the above-mentioned weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features.

[0016] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the various steps of the above-mentioned weak supervision analysis method of agricultural product repurchase intention associated with multi-dimensional features.

[0017] The present invention combines four innovative strategies, namely, a fine-grained pseudo-label generation method based on a cross-level sentiment triple extraction model, a pseudo-label correction strategy embedded with confidence learning and preliminary training of a deep model, a contribution evaluation model with a global perspective, and a multi-dimensional multi-label deep learning model construction method under a weak supervision framework. It solves the problem of missing labels in each feature dimension, and helps merchants to grasp customer needs efficiently and accurately and further formulate targeted resource allocation strategies by mining and quantifying the contribution of each feature dimension in the global context. The constructed multi-dimensional multi-label deep learning model helps consumers to quickly select products by capturing the fine-grained sentiment tendencies in massive comment texts, and is also helpful for merchants to discover potential problems in the process of selling products, thereby promoting the development of e-commerce platforms in agricultural product transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of a weak supervision analysis method for agricultural product repurchase intention associated with multi-dimensional features provided by the present invention; Figure 2 This is the overall label distribution diagram of the crawled agricultural product evaluation dataset; Figure 3 It is the overall framework diagram of the method; Figure 4 It is a custom pseudo-label judgment method flowchart; Figure 5 It is a quantitative result chart of the contribution of repurchase intention factors. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0020] Reference Figure 1 , a weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features, the steps are as follows: Step 1: Obtain review texts and ratings from e-commerce platforms to construct an agricultural product evaluation dataset; Step 2: Divide the agricultural product evaluation dataset into a review text list and a rating list, create a review noun matrix and a noun dictionary based on the review text list, and then convert the review text into a bag-of-words representation matrix based on the noun dictionary and the review noun matrix; then perform topic modeling to obtain the topic-noun distribution, thereby mining the factors that affect repurchase intention and determining the fine-grained evaluation dimensions; Step 3: Use the aspect-sentiment triple extraction model based on span-level interactive learning mechanism to extract triple information from the review text, including aspect items, opinion items and sentiment polarity; Step 4: Remove duplicates from the extracted aspect items and opinion items and classify them separately to obtain a fine-grained aspect item list and a fine-grained opinion item list; Step 5: Generate a pseudo-labeled dataset based on the extracted triple information, fine-grained aspect item list, and fine-grained opinion item list. Extract data and manually correct it as a clean test set, and use the rest as a pseudo-labeled training set. Step 6: Build a multi-dimensional multi-label deep learning model based on convolutional neural network (CNN) and Transformer, and use the pseudo-label correction strategy embedded with confidence learning and preliminary training of the multi-dimensional multi-label deep learning model to correct the pseudo-label training set; Step 7: Obtain the data set required for the contribution evaluation model, use the histogram-based gradient boosting classifier HistGradientBoostingClassifier, and further apply the feature contribution estimation method based on game theory to the histogram gradient boosting classifier, and use the contribution evaluation model to quantify the contribution of each repurchase intention influencing factor; Step 8: Pass in the corrected pseudo-label dataset and train a multi-dimensional multi-label deep learning model under a weakly supervised framework to mine the multi-dimensional emotions in the comment text.

[0021] In step 1 of this embodiment, automated crawler technology is used to obtain evaluation data of three types of agricultural products from multiple stores on three major e-commerce platforms to construct an agricultural product evaluation dataset (agricultural products include tomatoes, apples, and watermelons). The evaluation data consists of two parts: comment text and ratings. The ratings of all evaluations are marked with five sentiment levels from 1 to 5. The rating distribution is as follows: Figure 2 The original data set obtained first needs to be preprocessed, including removing duplicate data, removing empty data, and using regular expressions to remove special symbols contained in the comment text to meet the text requirements for data analysis.

[0022] In step 2 of this embodiment, the agricultural product evaluation data set is divided into a comment text list and the rating list ,in Respectively represent Comment text, Respectively represent The score corresponding to the comment text is then processed using the word segmentation tool and part-of-speech tagging tool of the nltk library to create a comment noun matrix ,in Indicates Comments in text nouns and create a noun dictionary from them After filtering low-frequency and high-frequency words, all comment texts are converted into a bag-of-words representation matrix based on the noun dictionary and comment noun matrix. , the bag of words represents each element in the matrix Indicates Comment text The number of times a noun appears, , Indicates Nouns in the dictionary The corresponding mapping in Indicates the frequency of occurrence. Then, the Latent Dirichlet Allocation (LDA) model is used for topic modeling. In order to improve the robustness of the LDA model to sparse data, the hyperparameters of the LDA model are traversed and optimized to adapt to the characteristics of short texts and improve the accuracy of topic modeling. Under each hyperparameter combination, the optimal number of topics under the hyperparameter configuration is determined by finding the maximum value of the LDA model consistency under all topic number settings. The calculation process is shown in formula (1): (1); in Indicated in The number of topics, including of the topics Number of review texts with nouns; Indicates Other nouns in the subject The number of times a noun appears in the same comment text. is to adjust the parameters to avoid or When it is 0, the logarithmic function is undefined. Indicates The number of nouns contained in the topic, Indicates The total number of topics under the topic number setting is calculated, and the LDA model is trained based on this to obtain the topic-noun distribution under this hyperparameter configuration. Finally, the topic-noun distribution of the optimal LDA model is obtained according to the maximum consistency score under all parameter combinations, thereby excavating the factors affecting the intention to repurchase and determining the fine-grained evaluation dimensions accordingly.

[0023] The process of extracting triple information from the review text using the aspect sentiment triple extraction model based on the span-level interactive learning mechanism described in step 3 of this embodiment is as follows: Traverse each comment text, perform word segmentation encoding and other processing to make it meet the input requirements of the BERT model, use the pre-trained BERT model to capture the contextual representation of the comment text and select the last layer with the shape of The hidden state of represents the length of the encoded sequence, Represents the hidden layer dimension of the BERT model, from 1 to Loop through and generate all possible spans in the comment text, consisting of a shape The span represents the tensor and a list of span index pairs Record span information, where N represents the total number of spans generated, is the dimension of the width embedding vector. Each span is represented by the concatenation of the tokens representing its start and end positions and a width embedding vector representing the span width. The span index is the element in the list. Respectively represent The position of the span in the original sequence of the comment text, , indicating the Span index information is a list consisting of two elements, represents the starting position index of the i-th span in the original sequence of the comment text, Indicates the end position index of the i-th span in the original sequence of the comment text.

[0024] Next, we classify spans: spans represent tensors A feedforward neural network consisting of multiple linear layers and activation functions is passed in. The input feature dimension of the feedforward network is , after three layers of linear transformation, it will eventually be mapped to three feature dimensions, representing "invalid span", "aspect span" and "opinion span", thereby obtaining the output transformed span representation tensor , the specific process is shown in formula (2): (2); in Respectively represent The weight matrix of the linear layer, Respectively represent The bias vector of the linear layer, ReLU and softmax are the activation functions, and then normalized to obtain the probability distribution of all spans belonging to different span categories , predict whether the span is a valid aspect item or opinion item, as shown in formula (3): (3); in represents the probability of the i-th span in the c-th span category, , They represent the original scores of the i-th span in the c-th span category and the k-th span category after linear transformation, Indicates the number of span categories.

[0025] According to the sequence length and pruning threshold Determine the number of aspect and opinion spans to retain for a given review , and then adopt a dual-channel pruning strategy to select the span category probability distribution Center front The span index that is most likely to be a facet item and the span index that is most likely to be an opinion item ,in Respectively represent the first The span represents the tensor The position index in , Respectively represent the opinions of the The span represents the tensor The position index in .

[0026] Next, a joint representation matrix is ​​constructed for each pair of aspect terms and opinion terms in the review text. , , the joint representation matrix Each element in Indicates that in the retained span, the Aspects and The joint representation of opinion items; Indicates The terms in span represent tensors The corresponding embedding representation in is, Indicates Opinion items in The corresponding embedding representation in is, It includes Aspects and The embedding representation of the relative position information between opinion items is generated as shown in formula (4): (4); in and Respectively represent The starting and ending positions of each aspect item in the original sequence of the comment text, and Respectively represent The starting and ending positions of the opinion items in the original sequence of the comment text are taken and then the minimum value is calculated. The operation maps it into a trainable embedding vector and returns a list of original position indices of candidate aspect item-opinion item pairs. , used to record the positions of aspect items and opinion items in all candidate aspect item-opinion item pairs in the original sequence of the comment text, Included in Aspects and The index of the starting and ending positions of the opinion items in the original sequence of the comment text, .

[0027] Then the joint representation matrix of aspect item-opinion item pair The input feature dimension of the feedforward neural network is similar to the previous one. ,in is the distance embedding vector dimension, after three layers of linear transformation and activation function After activation, it is mapped to four feature dimensions, representing "invalid", "positive", "negative", and "neutral" sentiment polarities respectively. After normalization, the probability distribution of all aspect item-opinion item pairs belonging to different sentiment polarity categories is finally obtained. : ; in Each element in Respectively represent The aspect item-opinion item pair belongs to The probability value of the sentiment polarity category, , M is the number of aspect item-opinion item pairs.

[0028] According to the original position index list of candidate aspect item-opinion item pairs and the probability distribution of sentiment polarity categories, the text corresponding to each valid aspect item-opinion item pair in the original sentence and the corresponding sentiment polarity are obtained. Thus, the (aspect item, opinion item, sentiment polarity) triple information extraction of all agricultural product review texts is completed.

[0029] In step 4 of this embodiment, first, the aspect items and opinion items are deduplicated. Then, for aspect items, the aspect items containing the word roots "apple", "watermelon", and "tomato" are divided into three categories, and then the remaining aspect items containing the same word roots are grouped into one category; for opinion items, adverbs expressing negation, modification, etc. are removed, and then word form restoration is performed and the word roots containing the same word roots are grouped into one category. After further manual adjustment and screening, the first A list of fine-grained aspect items of the fine-grained evaluation dimension and a list of fine-grained opinion items . Label the list consisting of all fine-grained aspect items as List 1, and the list consisting of all fine-grained opinion items as List 2.

[0030] The process of generating a pseudo-label data set, a clean test set, and a pseudo-label training set according to the extracted triple information, the fine-grained aspect item list, and the fine-grained opinion item list in step 5 of this embodiment is as follows: First, an initialization label is set for each fine-grained evaluation dimension of each comment, with the default value being 2.

[0031] Then define three types of pseudo labels (1 for positive, 0 for negative, and 2 for neutral or not mentioned) and a sentiment polarity-pseudo label conversion function to convert the sentiment polarities 'POS', 'NEG' and 'NEU' into corresponding pseudo labels.

[0032] Then, according to Figure 4 The custom pseudo-label judgment method shown modifies the fine-grained initial labels corresponding to all comment texts: traverse all comment texts, and for each comment text, traverse its corresponding triple list to process the triple information one by one. If the aspect item in a triple is in List 1, further determine which fine-grained evaluation dimension the aspect item represents. If the pseudo-label on this dimension is neither equal to 2 nor equal to the pseudo-label corresponding to the sentiment polarity, then set the pseudo-label of this dimension to 2, otherwise set it to the pseudo-label corresponding to the sentiment polarity. If the aspect item is not in List 1, check the opinion item. If the opinion item is in List 2, further determine which fine-grained evaluation dimension the opinion item represents. Similar to the aspect item, modify the pseudo-label of this dimension. If the aspect item of the triple is not in List 1 and the opinion is not in List 2, process the next tuple.

[0033] Finally, pseudo labels on the coarse-grained dimension are generated based on the triple information: if the triple list corresponding to a certain comment text is all triples representing positive relationships, the pseudo label of the text on the coarse-grained dimension is 1; if it is all triples representing negative relationships, the pseudo label on the coarse-grained dimension is 0; if there are triples representing both negative and positive relationships, the pseudo label on the coarse-grained dimension is 2, which reflects the customer's game psychology for repurchasing products. In this way, a pseudo-label dataset is obtained.

[0034] 525 data were extracted from the pseudo-labeled dataset and manually corrected as a pure test set. .in represents the pure test set comment text, They represent the pure test set comment text in clean labels on fine-grained dimensions, Represents the clean labels of the clean test set review text in the coarse-grained dimension. The remaining data is used as the pseudo-label training set.

[0035] The process of step six of this embodiment is as follows: First, find the coarse-grained and fine-grained pseudo labels that need to be corrected. Pass in the pseudo-label training set, use the text classifier FirstColFtClassifier based on the Fasttext model to train each dimension separately, and then use the confidence learning tool cleanlab to find low-confidence samples. For each dimension, use the comment text as the feature part of the training data, and use the pseudo-label on this dimension as the label part of the training data. After word mapping, obtain the vector representation of all comment texts. ,in Respectively represent The vector representation of the comment text is obtained, and then the label category probability distribution of all comment texts in this fine-grained dimension is obtained through formula (5): (5); in Indicates Comment texts are predicted to belong to The probability value of the class label, and Respectively , The weight matrix of the class label elements, Indicates The first comment in the text vector elements, and Respectively represent The bias term for the class label, is the number of label categories. Then, the confidence learning tool cleanlab constructs the joint distribution matrix of predicted labels and pseudo labels based on the obtained label category probability distribution , the joint distribution matrix Each element in Indicates that the given pseudo label category is Under the class label, the true label category is identified as The probability value of the class label is calculated as shown in formula (6).

[0036] (6); in, represents the true label category, represents the pseudo label category, Indicates that the true label category is Class labels, Indicates that the pseudo label category is Class label, Indicates The true label category of samples, Indicates The pseudo-label category given by the samples, represents the indicator function, is the total number of comment text samples, and the label quality score (between 0 and 1) of each sample is calculated based on it. The closer the quality score is to 0, the more likely the label is a noise label and the lower the credibility. The confidence learning tool cleanlab will select a threshold based on the quality score distribution to distinguish between low-confidence labels and high-confidence labels. The samples corresponding to low-confidence labels and high-confidence labels are low-confidence samples and high-confidence samples, respectively.

[0037] Next, the low-confidence labels are corrected and screened. A multi-dimensional multi-label deep learning model based on convolutional neural network (CNN) and Transformer is constructed. The multi-dimensional multi-label deep learning model contains an embedding layer, a position encoding layer, a convolution layer, a Transformer encoder, and a fully connected layer, which aims to map the input sequence to multiple category predictions on multiple evaluation dimensions. 105 data sets are extracted from the pseudo-label training set for manual annotation and passed into the constructed multi-dimensional multi-label deep learning model for training. The preliminary model is obtained based on the accuracy of the training set and the preliminary model is used to predict the low-confidence labels of the low-confidence samples to obtain the model labels. The model labels of the low-confidence samples are then compared with the true labels identified by FirstColFtClassifier for the low-confidence samples. It is stipulated that if the model label is consistent with the true label identified by FirstColFtClassifier, the corrected label is the model label. If the two are inconsistent, they are corrected manually. In addition, there are inevitably vague labels in the evaluation dataset, which may cause confusion in the model training process and affect the performance and generalization ability of the multi-dimensional multi-label deep learning model. Therefore, it is necessary to remove some data with unclear label categories to improve the quality of the dataset and the robustness of the model, so as to obtain the final corrected labels. The samples corresponding to the corrected labels and the high-confidence samples together constitute the corrected pseudo-label training set. ,in represents the corrected pseudo-label training set comment text, They represent the corrected pseudo-label training set comment text in Pseudo labels on fine-grained dimensions, Represents the pseudo-labels of the review texts in the coarse-grained dimension of the pseudo-label training set after correction.

[0038] The process of step seven of this embodiment is as follows: First, obtain the data set required for the contribution evaluation model. Generate the labels of all review texts in the repurchase intention dimension in the corrected pseudo-label training set and the clean test set according to the score list: Figure 3 As shown in the figure, if the score is 1 or 2, the label on the repurchase intention dimension is set to 0, if the score is 3 or 4, it is set to 1, and if the score is 5, it is set to 2. Then, the pseudo labels on all fine-grained evaluation dimensions in the corrected pseudo-label training set and the clean test set are transformed to meet the data requirements of the contribution evaluation model: pseudo labels indicating positive are transformed to 2, and pseudo labels indicating neutral or unmentioned are transformed to 1. The datasets that form the contribution evaluation model are concatenated .in Respectively represent the pseudo-label training set and the pure test set after correction in The pseudo-labels after transformation on a fine-grained dimension, Represents the label on the repurchase intention dimension, thereby obtaining the training set and test set of the contribution evaluation model.

[0039] Then, the contribution evaluation model is constructed, and the training set of the contribution evaluation model is passed in for parameter optimization, and the performance of the contribution evaluation model is verified on the test set of the contribution evaluation model. The histogram-based gradient boosting classifier (HistGradientBoostingClassifier) ​​is used, and multiple possibilities are set for its learning rate, L2 regularization term strength, maximum number of iterations, etc. The training set of the contribution evaluation model is passed in, the pseudo labels on the fine-grained dimension are used as input features, and the labels on the repurchase intention dimension are used as target labels. The hyperparameter grid is traversed and the performance of the gradient boosting classifier under each parameter combination is evaluated through five-fold cross validation, and the hyperparameter combination that makes the gradient boosting classifier perform best on multiple data subsets is selected. In order to further verify the reliability of the gradient boosting classifier, the trained gradient boosting classifier is placed on the test set of the contribution evaluation model to predict the repurchase intention. After verification, its accuracy rate reaches 85.42%, proving that it has met the requirements for further construction of the contribution evaluation model.

[0040] Finally, the training set is passed to the contribution evaluation model again to quantify the contribution of each repurchase intention influencing factor (each fine-grained evaluation dimension). The whole process will be cross-validated in five folds to avoid misjudgment due to overfitting or underfitting of a single model and calculate the contribution value of each factor. In this process, the data set will be divided into 5 subsets, 4 subsets will be used in turn for training under the best hyperparameter combination, and the remaining 1 subset will be used for validation. In each fold validation, the accuracy of the repurchase intention prediction on the current validation subset will be calculated. , and then estimate the contribution of various factors affecting repurchase intention on the validation subset by masking the features and observing the changes in model performance ,in Respectively expressed in Fold the first The contribution value of the factors affecting repurchase intention. The calculation process is shown in formula (7): (7); in, Indicated in Fold the first The contribution value of the factors influencing repurchase intention, Indicates that all the covered The feature subset of the factors affecting repurchase intention is classified, and the cross entropy loss function is used to obtain the performance quantitative representation of a feature subset. And add the first Quantitative representation of the performance of the factors affecting repurchase intention after characterization , by observing the difference, we can get The marginal contribution of class features to the performance of the contribution evaluation model, It is Characteristics of factors influencing repurchase intention The corresponding weight value, and Represents all feature numbers and The number of features in .

[0041] After the five-fold cross validation is completed, the validation set accuracy matrix is ​​constructed and the contribution matrix ,in Respectively represent The accuracy of the validation set is Represents the factors affecting repurchase intention in the first The contribution matrix on the validation subset is then normalized to obtain the matrix ,in, Respectively represent The relative importance value of the validation subset. Finally, the contribution value is weighted averaged and normalized to obtain the global feature contribution vector ,in Respectively represent The contribution of the factors influencing repurchase intention to consumers' repurchase intention is quantified. Each element in the vector The calculation process is shown in formula (8): (8); in, represents the relative importance value of the i-th validation subset, and Respectively expressed in On the validation subset The contribution value of the factors affecting repurchase intention.

[0042] According to Figure 5 The quantitative analysis results shown in the figure provide resource allocation suggestions for merchants based on the contribution value: Figure 5 As shown, the three factors with contribution values ​​greater than the average are "taste", "freshness" and "appearance". Among them, the contribution value of "taste" is the largest, much higher than other topics. It can be seen that the taste of agricultural products is the most important factor affecting customers' repurchase willingness. Agricultural product suppliers should focus their resources on improving the taste of agricultural products to increase customer repurchase rate. The second and third are freshness and product appearance factors, respectively. Their contribution values ​​are slightly lower than the taste factor. These are also the two aspects that customers care about more. It is recommended that agricultural product sellers increase resource investment in transportation and improved preservation technology to ensure the freshness of agricultural products. At the same time, certain resources should be allocated to optimizing the appearance of products. Secondly, the fourth is the "texture" factor. It can be seen that the public is generally more tolerant of whether the product has a powdery and glutinous taste or a crisp taste, and whether the watermelon has seeds or no seeds. Although this factor does not play a decisive role in the repurchase intention, merchants should also maintain their advantages in this aspect and further optimize it under the premise of ensuring the "taste", "freshness" and "appearance" of the product. From Figure 5 It can be seen that there are few comments on the prices of agricultural products. Merchants may raise prices appropriately and set reasonable prices by observing changes in repurchase rates. Although merchant service attitude does not have a decisive influence on customers' repurchase intention, if resources permit, they can also consider improving customers' shopping experience and the convenience of delivery services. Otherwise, more resources can be invested in more important aspects.

[0043] In step eight of this embodiment, the manually corrected pure test set and the corrected pseudo-label training set together constitute a corrected pseudo-label dataset, which is used as the dataset of the multi-dimensional multi-label deep learning model; After each comment text is segmented, encoded, and filled, it is passed into the multi-dimensional multi-label deep learning model together with the mask marking its filling position; First, the embedding layer maps each word index in the text into a high-dimensional vector. The position encoding layer then adds position information to each embedding vector, thereby explicitly introducing position information in the sequence and helping the multi-dimensional multi-label deep learning model capture the sequential relationship in the sequence. Next, the convolution layer of the model will capture the local features in the input sequence through one-dimensional convolution operations, and introduce nonlinearity in combination with the ReLU activation function so that the neural network can learn more complex and nonlinear features. The convolutional features are then batch normalized to make the model easier to converge, which speeds up the training process. In order to prevent information loss after the convolution operation, the multi-dimensional multi-label deep learning model will save the input information of the convolution layer and add it to the output of the convolution layer to form a residual connection, which is used as the input of the tranformer encoder together with the mask marking the filling position; The Transformer encoder is composed of multiple Transformer Encoder Layers, which process each time step in the text sequence in parallel: focus on each position in the sequence through a multi-head self-attention mechanism to capture the global dependencies in the sequence, and perform nonlinear transformations on the features of each position in the sequence through a feedforward neural network (FFN). Finally, a hidden state containing contextual information will be generated. Among them, the provided src_key_padding_mask mask will cause the self-attention mechanism to ignore the padding part in the sequence, avoiding the padding part from introducing noise into the calculation of the attention weight; The hidden state of the last time step is extracted and passed to the shared connection layer, aiming to map the extracted features to multi-class labels of all fine-grained evaluation dimensions. Finally, the prediction result is presented as a three-dimensional tensor through a tensor operation. Each element in the tensor represents the probability that a given text belongs to the three categories of labels 0, 1, and 2 on a given evaluation dimension. Finally, the label corresponding to the maximum probability value is taken as the prediction result of the model.

[0044] During the training process, early stopping and dropout regularization techniques are used to prevent the model from overfitting. In order to make the multi-dimensional multi-label deep learning model more finely adjusted in the later stage of training to find the optimal solution, the learning rate decay method is adopted. Due to the imbalance of the labels of the dataset used for the training of the multi-dimensional multi-label deep learning model, the cross entropy loss function is adopted and the loss weights of the three types of labels are set to 0, 1, and 2. During training, the mean loss of all dimensions is used as the loss backpropagation of a batch. Accuracy, precision, recall, and F1 value are used as evaluation indicators of the multi-dimensional multi-label deep learning model. The prediction effects of the overall and each fine-grained dimension are measured respectively, and the best multi-dimensional multi-label deep learning model is determined based on the comprehensive results of these four evaluation indicators on the test set as a whole. The accuracy measurement process is shown in formula (9): (9); in, Indicates the number of influencing factors, represents the number of test set samples, and Respectively represent The sample in The prediction results and true labels on the class influencing factors, Represents the indicator function.

[0045] The obtained optimal multi-dimensional multi-label deep learning model is further evaluated and compared with the four models. The results are shown in Table 1. The present invention has better performance.

[0046] Table 1

[0047] Another embodiment of the present invention provides a weakly supervised analysis system of agricultural product repurchase intention associated with multi-dimensional features, the system comprising a memory, a processor, and a computer program stored in the memory, and when the computer program is executed by the processor, it implements the various steps of the weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features as described in the above embodiment.

[0048] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the various steps of the weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features as described in the above-mentioned embodiment.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A weakly supervised analysis method for agricultural product repurchase intention based on multi-dimensional feature association, characterized in that: Here are the steps: Step 1: Obtain review texts and ratings from e-commerce platforms to construct an agricultural product evaluation dataset; Step 2: Divide the agricultural product evaluation dataset into a review text list and a rating list, create a review noun matrix and a noun dictionary based on the review text list, and then convert the review text into a bag-of-words representation matrix based on the noun dictionary and the review noun matrix; then perform topic modeling to obtain the topic-noun distribution, thereby mining the factors that affect repurchase intention and determining the fine-grained evaluation dimensions; Step 3: Use the aspect-sentiment triple extraction model based on span-level interactive learning mechanism to extract triple information from the review text, including aspect items, opinion items and sentiment polarity; Step 4: Remove duplicates from the extracted aspect items and opinion items and classify them separately to obtain a fine-grained aspect item list and a fine-grained opinion item list; Step 5: Generate a pseudo-labeled dataset based on the extracted triple information, fine-grained aspect item list, and fine-grained opinion item list. Extract data and manually correct it as a clean test set, and use the rest as a pseudo-labeled training set. Step 6: Build a multi-dimensional multi-label deep learning model based on convolutional neural network and Transformer, and use the pseudo-label correction strategy embedded with confidence learning and preliminary training of the multi-dimensional multi-label deep learning model to correct the pseudo-label training set; Step 7: Obtain the data set required for the contribution evaluation model, use the histogram-based gradient boosting classifier, and further apply the feature contribution estimation method based on game theory to the histogram gradient boosting classifier, and use the contribution evaluation model to quantify the contribution of each repurchase intention influencing factor; Step 8: Pass in the corrected pseudo-label dataset and train a multi-dimensional multi-label deep learning model under a weakly supervised framework to mine the multi-dimensional emotions in the comment text.

2. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: In step 2, the latent Dirichlet allocation model is used for topic modeling.

3. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: The process of extracting triple information from the review text using the aspect sentiment triple extraction model based on the span-level interactive learning mechanism described in step 3 is as follows: Traverse each comment text, perform word segmentation encoding, use the pre-trained BERT model to capture the contextual representation of the comment text and select the last layer shape as the hidden state, loop through, and generate all possible spans in the comment text; Classify spans: obtain the probability distribution of all spans belonging to different span categories and predict whether the span is a valid aspect item or opinion item; The number of aspect items and opinion item spans that need to be retained for a given review text is determined based on the sequence length and pruning threshold, and then a dual-channel pruning strategy is adopted to select the top 10 aspect items and opinion item spans in the probability distribution of span categories. The span index that is most likely to be an aspect item and the span index that is most likely to be an opinion item; Construct a joint representation matrix for each pair of aspect terms and opinion terms in the review text; The sentiment polarity is judged on the joint representation matrix of aspect item-opinion item pairs, and the probability distribution of all aspect item-opinion item pairs belonging to different sentiment polarity categories is obtained.

4. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: In step 4, first, the aspect items and opinion items are deduplicated separately; then, the aspect items with the same root word are grouped into one category; for the opinion items, after removing the adverbs indicating the function, the word form is restored and the items with the same root word are grouped into one category; further, after manual adjustment and screening, the fine-grained aspect item list and fine-grained opinion item list of each fine-grained evaluation dimension are generated.

5. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: The specific process of step five is as follows: First, set initial labels for each fine-grained evaluation dimension of each review; Then define three types of pseudo labels and sentiment polarity-pseudo label conversion functions to convert sentiment polarity into corresponding pseudo labels; Next, the fine-grained initial labels corresponding to all comment texts are modified based on the pseudo-label judgment method; Finally, pseudo labels on coarse-grained dimensions are generated based on the triple information to obtain a pseudo label dataset; Some data are extracted from the pseudo-label dataset and manually corrected as a pure test set, and the remaining data is used as a pseudo-label training set.

6. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: The specific process of step six is ​​as follows: The pseudo-label training set is passed in, and the text classifier based on the Fasttext model is trained for each dimension separately, and then the confidence learning tool is used to find low-confidence samples; A multi-dimensional and multi-label deep learning model based on convolutional neural networks and Transformer is constructed. Data is extracted from the pseudo-label training set for manual annotation and passed into the constructed multi-dimensional and multi-label deep learning model for training. A preliminary model is obtained according to the accuracy of the training set and the preliminary model is used to predict the low-confidence labels of low-confidence samples to obtain the model labels. The model labels of the low-confidence samples are then compared with the true labels recognized by the text classifier based on the Fasttext model for the low-confidence samples. It is stipulated that: if the model label is consistent with the true label recognized by the text classifier based on the Fasttext model, the corrected label is the model label. If the two are inconsistent, they are corrected manually. The samples corresponding to the corrected labels and the high-confidence samples together constitute the corrected pseudo-label training set.

7. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: The specific process of step seven is as follows: Generate labels for all review texts in the corrected pseudo-label training set and the clean test set on the repurchase intention dimension based on the score list; then transform the pseudo-labels on all fine-grained evaluation dimensions in the corrected pseudo-label training set and the clean test set to meet the data requirements of the contribution evaluation model; The data sets constituting the contribution evaluation model are concatenated and divided into a training set and a test set of the contribution evaluation model; Construct a contribution evaluation model, pass in the training set of the contribution evaluation model, use the pseudo labels on the fine-grained dimension as input features, use the labels on the repurchase intention dimension as target labels, traverse the hyperparameter grid and evaluate the performance of the gradient boosting classifier under each parameter combination through five-fold cross validation, and select the hyperparameter combination that makes the gradient boosting classifier perform best on multiple data subsets; The training set is then fed back into the contribution evaluation model to quantify the contribution of each factor influencing repurchase intention.

8. The weakly supervised analysis method of agricultural product repurchase intention based on multi-dimensional feature association according to claim 1 is characterized in that: In step eight, the manually corrected pure test set and the corrected pseudo-label training set together constitute the corrected pseudo-label dataset, which is used as the dataset for the multi-dimensional and multi-label deep learning model.

9. A weakly supervised analysis system for agricultural product repurchase intention based on multi-dimensional feature association, characterized in that: It includes a memory, a processor and a computer program stored in the memory, and when the computer program is executed by the processor, it implements the various steps of the weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, each step of the weakly supervised analysis method of agricultural product repurchase intention associated with multi-dimensional features as described in any one of claims 1 to 8 is implemented.

Images