User feedback information processing method and device in financial service, equipment and medium

Abstract

The invention relates to a user feedback information processing method and device in financial business, computer equipment, a storage medium and a computer program product, and relates to the field of computer technology and financial science and technology. According to the method and the device, the accuracy of extracting the feedback keywords in the user feedback information and the efficiency of analyzing and processing the user feedback information can be improved. The method comprises the steps of obtaining user feedback information of multiple preset channels, and judging whether the user feedback information is user feedback information related to a current financial institution or not; judging whether the preset channel is an internal system of the current financial institution; identifying feedback keywords in the user feedback information, and judging whether the feedback keywords belong to existing keywords in a user feedback information base or not; when the feedback keyword does not belong to the existing keyword in the user feedback information base, performing sentiment classification on the feedback keyword to obtain a target feedback keyword of a target sentiment type; and obtaining corresponding user feedback processing indication information according to the target feedback keyword.

Description

Technical Field

[0001] The present application relates to the fields of computer technology and financial technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for processing user feedback information in financial services. Background Art

[0002] With the development of computer technology and financial technology, the data processing capabilities of computers have been improved. Subsequently, the technology of using computers to process user feedback information has emerged in the field of financial technology. This technology is widely used in banks and other financial institutions, helping financial institutions to improve the processing speed of user feedback information in financial services.

[0003] In traditional technology, Chinese word segmentation methods are usually combined with different semantic analysis algorithms to process text information. However, the Chinese word segmentation methods in current technology are not accurate enough in identifying key information such as topics, objects and emotions, resulting in low accuracy of keywords extracted based on Chinese word segmentation, making the analysis and processing of user feedback information less effective. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for processing user feedback information in financial services in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for processing user feedback information in a financial service. The method comprises:

[0006] Obtain user feedback information from multiple preset channels, and determine whether the user feedback information is user feedback information related to the current financial institution;

[0007] When the user feedback information is user feedback information of the current financial institution, determining whether the preset channel corresponding to the user feedback information is an internal system of the current financial institution;

[0008] When the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identifying feedback keywords in the user feedback information, and determining whether the feedback keywords belong to existing keywords in the user feedback information database;

[0009] When the feedback keyword does not belong to the existing keywords in the user feedback information library, sentiment classification is performed on the feedback keyword to obtain a target feedback keyword whose sentiment type is a target sentiment type;

[0010] According to the target feedback keyword, corresponding user feedback processing instruction information is obtained.

[0011] In one embodiment, determining whether the user feedback information is user feedback information related to the current financial institution includes:

[0012] Acquire a text data corpus of the current financial institution; perform vector space division on the text data in the text data corpus according to preset topics, and assign corresponding classification labels to the division results; and identify user feedback information related to the current financial institution in the user feedback information based on the division results and classification labels.

[0013] In one embodiment, the identifying feedback keywords in the user feedback information includes:

[0014] The user feedback information is cleaned and the cleaned user feedback information is segmented using natural language processing technology; the parts of speech of words are marked in the user feedback information after the word segmentation processing; according to the parts of speech of the words, the user feedback information after the word segmentation processing is represented as a word frequency vector, and the words whose frequency vectors exceed a frequency threshold are used as the feedback keywords.

[0015] In one embodiment, the sentiment classification of the feedback keywords includes:

[0016] The feedback keywords are input into a trained naive Bayes classifier to obtain the sentiment scores corresponding to the feedback keywords; the feedback keywords whose sentiment scores are greater than the sentiment classification threshold are determined as feedback keywords of the first sentiment type; the feedback keywords whose sentiment scores are less than the sentiment classification threshold are determined as feedback keywords of the second sentiment type; the first sentiment type and the second sentiment type are two opposite sentiment types.

[0017] In one embodiment, the method further comprises:

[0018] When the preset channel corresponding to the user feedback information is not the internal system of the current financial institution, identify the external target feedback keyword whose emotion type in the user feedback information is the target emotion type; determine whether the user feedback information corresponding to the external target feedback keyword has generated abnormal user feedback information; when the user feedback information corresponding to the external target feedback keyword has not generated abnormal user feedback information, use a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information.

[0019] In one embodiment, the determining whether the user feedback information corresponding to the external target feedback keyword has generated abnormal user feedback information includes:

[0020] The external target feedback keyword is used as a search content, and a corresponding abnormal feedback keyword is determined in the search results; a matching degree between the external target feedback keyword and the abnormal feedback keyword is calculated; and according to the matching degree, it is determined whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information.

[0021] In one embodiment, the using of a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information includes:

[0022] A binary classification model is constructed using the support vector machine algorithm; external target feedback keywords and sample user feedback information in a preset format are input into the binary classification model for inner product operation to obtain similarity values ​​between the external target feedback keywords and the sample user feedback information; the similarity values ​​are input into the support vector machine algorithm to obtain prediction results.

[0023] In a second aspect, the present application also provides a user feedback information processing device in a financial service. The device comprises:

[0024] An information judgment module is used to obtain user feedback information from multiple preset channels and judge whether the user feedback information is user feedback information related to the current financial institution;

[0025] a channel determination module, configured to determine, when the user feedback information is user feedback information of the current financial institution, whether the preset channel corresponding to the user feedback information is an internal system of the current financial institution;

[0026] A word determination module, used for identifying feedback keywords in the user feedback information and determining whether the feedback keywords belong to existing keywords in the user feedback information database when the preset channel corresponding to the user feedback information is the internal system of the current financial institution;

[0027] A sentiment classification module, used for, when the feedback keyword does not belong to an existing keyword in the user feedback information library, to perform sentiment classification on the feedback keyword to obtain a target feedback keyword whose sentiment type is a target sentiment type;

[0028] The feedback instruction module is used to obtain corresponding user feedback processing instruction information according to the target feedback keyword.

[0029] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0030] Obtain user feedback information from multiple preset channels, and determine whether the user feedback information is user feedback information related to the current financial institution; when the user feedback information is user feedback information of the current financial institution, determine whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution; when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identify feedback keywords in the user feedback information, and determine whether the feedback keywords belong to existing keywords in a user feedback information library; when the feedback keywords do not belong to existing keywords in the user feedback information library, perform sentiment classification on the feedback keywords to obtain target feedback keywords whose sentiment type is a target sentiment type; and obtain corresponding user feedback processing instruction information according to the target feedback keywords.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0032] Obtain user feedback information from multiple preset channels, and determine whether the user feedback information is user feedback information related to the current financial institution; when the user feedback information is user feedback information of the current financial institution, determine whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution; when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identify feedback keywords in the user feedback information, and determine whether the feedback keywords belong to existing keywords in a user feedback information library; when the feedback keywords do not belong to existing keywords in the user feedback information library, perform sentiment classification on the feedback keywords to obtain target feedback keywords whose sentiment type is a target sentiment type; and obtain corresponding user feedback processing instruction information according to the target feedback keywords.

[0033] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0034] Obtain user feedback information from multiple preset channels, and determine whether the user feedback information is user feedback information related to the current financial institution; when the user feedback information is user feedback information of the current financial institution, determine whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution; when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identify feedback keywords in the user feedback information, and determine whether the feedback keywords belong to existing keywords in a user feedback information library; when the feedback keywords do not belong to existing keywords in the user feedback information library, perform sentiment classification on the feedback keywords to obtain target feedback keywords whose sentiment type is a target sentiment type; and obtain corresponding user feedback processing instruction information according to the target feedback keywords.

[0035] The user feedback information processing method, device, computer equipment, storage medium and computer program product in the above-mentioned financial business, by collecting user feedback information from multiple preset channels obtained by the server, distinguishes the target objects of the user feedback information into the current financial institution and the non-current financial institution; and when the user feedback information is the user feedback information of the current financial institution, further distinguishes the preset channels corresponding to the user feedback information into the internal system of the institution and the external system of the institution; then, the user feedback information is visualized, standardized and digitized using natural language processing technology to accurately identify the feedback keywords therein; secondly, the sentiment analysis technology is used to perform sentiment classification on the feedback keywords to identify the target feedback keywords therein; finally, the corresponding user feedback processing instruction information is generated according to the target feedback keywords through a rigorous algorithm. Compared with the Chinese word segmentation method in the existing technology, this solution performs word segmentation on user feedback information according to a basic dictionary, and then determines feedback keywords according to preset indicators of relevance to the text topic, and analyzes the user feedback information based on the feedback keywords, so that the user feedback information is visualized in the form of events; thereby improving the accuracy of feedback keyword extraction in user feedback information and the efficiency of analyzing and processing user feedback information, so that user feedback information can be followed up and processed in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 An application environment diagram of a method for processing user feedback information in a financial service in one embodiment;

[0037] Figure 2 A schematic diagram of a flow chart of a method for processing user feedback information in a financial service in one embodiment;

[0038] Figure 3 A schematic diagram of a flow chart of a step of sentiment classification of feedback keywords in one embodiment;

[0039] Figure 4 A schematic diagram of a flow chart of a step of predicting the generation of abnormal user feedback information in one embodiment;

[0040] Figure 5 It is a flowchart of a method for processing user feedback information in a financial service in a specific embodiment;

[0041] Figure 6 is a flow chart of a method for processing user feedback information in a financial service in another embodiment;

[0042] Figure 7 is a structural block diagram of a user feedback information processing device in a financial service in one embodiment;

[0043] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] The method for processing user feedback information in financial services provided in the embodiments of the present application can be applied to Figure 1 In the application environment shown, the terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be integrated on the server or placed on the cloud or other network servers.

[0046] Specifically, the method for processing user feedback information in financial services provided in the embodiments of the present application can be executed by a server.

[0047] Exemplarily, the server obtains user feedback information from multiple preset channels, and determines whether the user feedback information is user feedback information related to the current financial institution; when the user feedback information is user feedback information of the current financial institution, the server determines whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution; when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, the server identifies feedback keywords in the user feedback information, and determines whether the feedback keywords belong to existing keywords in the user feedback information library; when the feedback keywords do not belong to existing keywords in the user feedback information library, the server performs sentiment classification on the feedback keywords to obtain target feedback keywords with a target sentiment type; finally, the server obtains corresponding user feedback processing instruction information based on the target feedback keywords.

[0048] In such Figure 1In the application environment shown, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, and tablet computers. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0049] In one embodiment, Figure 2 As shown, a method for processing user feedback information in financial services is provided, and the method is applied to Figure 1 The server in the example is used to illustrate the following steps:

[0050] Step S201, obtaining user feedback information from multiple preset channels, and determining whether the user feedback information is user feedback information related to the current financial institution.

[0051] Specifically, the server obtains user feedback information from multiple preset channels, and determines whether the user feedback information is user feedback information related to the current financial institution.

[0052] How to use the user feedback information acquisition tool:

[0053] 1. Create a keyword list that includes bank names, product names, abbreviations, official logos, industry terms, and keywords related to banking business.

[0054] 2. Use Boolean logic operators: The user feedback information acquisition tool supports Boolean logic operators (AND, OR, NOT) to filter content more accurately. For example, you can set search conditions to require posts containing bank names and product names (AND operation) and exclude posts containing specific sensitive words (NOT operation).

[0055] 3. Identification through tags and topics: The user feedback information acquisition tool allows users to create custom tags or topics to further classify and filter content. Different tags can be created for different bank products, services or activities to better organize and analyze user feedback data.

[0056] 4. Advanced search settings: The user feedback information acquisition tool can provide advanced search settings, allowing users to precisely control the search conditions. By specifying parameters such as the search time range, social media platform, post type (text, image and video), user feedback information can be further accurately obtained.

[0057] Step S202: when the user feedback information is user feedback information of the current financial institution, it is determined whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution.

[0058] Specifically, when the user feedback information is user feedback information of the current financial institution, the server determines whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution.

[0059] For example, the server can determine whether the preset channel corresponding to the user feedback information is an in-bank system through the data source identifier, and identify the in-bank system as channel 1 and the out-of-bank system as channel 0.

[0060] Step S203, when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identifying the feedback keyword in the user feedback information, and determining whether the feedback keyword belongs to an existing keyword in the user feedback information database.

[0061] Specifically, when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, the feedback keyword in the user feedback information is identified, and the server determines whether the feedback keyword belongs to an existing keyword in the user feedback information database.

[0062] For example, the server identifies feedback keywords in the user feedback information, and these keywords may include words related to the internal business of the bank, such as product names, service names, and departments.

[0063] 1. Perform text cleaning on user feedback information to remove special characters, HTML (Hypertext Markup Language) tags, and unnecessary spaces.

[0064] 2. Use natural language processing technology to segment this article and mark the parts of speech of the words.

[0065] 3. Keyword extraction, bag-of-words model: Create a document-word matrix, represent the text as a word frequency vector, and then select words with higher frequency as keywords.

[0066] To further illustrate, the server determines whether the identified keyword is an existing keyword in the user feedback information database.

[0067] 1. Compare with the user feedback information database (the existing user feedback data comes from within the bank).

[0068] 2. Keyword matching: traverse the words in the user feedback information and match them with the internal keyword list; you can use the TF-IDF (Term Frequency-Inverse Document Frequency) matching method: TF-IDF is a method commonly used for text matching. First, calculate the TF-IDF value for each keyword in the user feedback information and the user feedback information library. Then, by comparing the TF-IDF value of the keyword in the user feedback information with the TF-IDF value of the corresponding word in the user feedback information library, the matching degree can be determined. A higher matching degree means that the keyword matches better. A high matching degree proves that the published user feedback information is a keyword in the existing user feedback information library. If the matching degree is low, it proves that there is no such feedback keyword in the existing user feedback information library, and it is necessary to further judge whether this keyword involves the user feedback information.

[0069] Step S204: when the feedback keyword does not belong to the existing keywords in the user feedback information library, sentiment classification is performed on the feedback keyword to obtain a target feedback keyword whose sentiment type is the target sentiment type.

[0070] Among them, sentiment analysis technology is a technology that automatically identifies and analyzes the emotional tendencies expressed in text, speech or images through computers.

[0071] Specifically, when the feedback keyword does not belong to the existing keywords in the user feedback information library, the server uses sentiment analysis technology to perform sentiment classification on the feedback keyword to obtain a target feedback keyword whose sentiment type is the target sentiment type.

[0072] Step S205: acquiring corresponding user feedback processing instruction information according to the target feedback keyword.

[0073] Specifically, the server obtains corresponding user feedback processing instruction information according to the target feedback keyword.

[0074] For example, the server classifies the user feedback information according to the user feedback information rating, and rates each negative feedback information according to the results of the user feedback information classification, sentiment analysis, and context analysis. The rating can be qualitatively divided into three parts: low, medium, and high to indicate the severity of the user feedback information.

[0075] In the above-mentioned user feedback information processing method in financial business, the target objects of user feedback information are distinguished into current financial institutions and non-current financial institutions by collecting user feedback information from multiple preset channels obtained by the server; and when the user feedback information is user feedback information of the current financial institution, the preset channels corresponding to the user feedback information are further distinguished into internal systems of the institution and external systems of the institution; then, the user feedback information is visualized, standardized and digitized by natural language processing technology to accurately identify the feedback keywords; secondly, the feedback keywords are sentimentally classified by sentiment analysis technology to identify the target feedback keywords; finally, the corresponding user feedback processing instruction information is generated according to the target feedback keywords through a rigorous algorithm. Compared with the Chinese word segmentation method in the current technology, this scheme performs word segmentation processing on the user feedback information according to the basic dictionary, and then determines the feedback keywords according to the preset indicators of the relevance to the text topic, and analyzes the user feedback information based on the feedback keywords, so that the user feedback information is visualized in the form of events; thereby improving the accuracy of feedback keyword extraction in the user feedback information and the efficiency of analyzing and processing the user feedback information, so that the user feedback information can be followed up and processed in a timely manner.

[0076] In one embodiment, in the above step S201, determining whether the user feedback information is user feedback information related to the current financial institution specifically includes the following steps:

[0077] Acquire the text data corpus of the current financial institution; divide the text data in the text data corpus into vector space according to preset topics, and assign corresponding classification labels to the division results; based on the division results and classification labels, identify user feedback information related to the current financial institution in the user feedback information.

[0078] Among them, a text data corpus refers to a collection of a large number of text documents, which is used for tasks such as text analysis, natural language processing, and machine learning.

[0079] Specifically, the server obtains the text data corpus of the current financial institution; divides the text data in the text data corpus into vector space according to preset topics, and assigns corresponding classification labels to the division results; then, the server identifies user feedback information related to the current financial institution in the user feedback information based on the division results and the classification labels.

[0080] For example, the user feedback information acquisition tool in the server acquires user feedback information from multiple preset channels, and determines whether the user feedback information is user feedback information related to the bank.

[0081] 1. Word Embeddings: Using the word embedding technology Word2Vec, the text is represented as a dense vector space, and then these vectors are used for classification. These methods can capture the semantic relationship between words and improve the accuracy of classification.

[0082] 2. Use the bank text data corpus, including text and corresponding classification labels, divide the corpus content into vector space and assign corresponding labels, 1 means related to the bank, and 0 means unrelated to the bank.

[0083] 3. Use Word2Vec to train a word embedding model to convert words in text data into dense vector representations.

[0084] for example:

[0085] from gensim.models import Word2Vec

[0086] # Word segmentation

[0087] tokenized_corpus = [text.split() for text in corpus]

[0088] # Train the Word2Vec model

[0089] model = Word2Vec(sentences=tokenized_corpus, vector_size=100, window=5, min_count=1, sg=0)

[0090] Sentences: A list of tokenized texts, a list of social media posts, news articles, or other texts, where each element is a text string.

[0091] vector_size: specifies the dimension of the generated word vector. The higher the dimension, the richer the semantic information that the word vector can capture.

[0092] window: Specifies the size of the context window, which is used to determine context terms. In user feedback acquisition tools, the window size is used to define the scope of finding relevant content in the text. A larger window may capture a wider range of context information.

[0093] min_count: Specifies the minimum word frequency threshold for filtering out low-frequency words. In the user feedback information acquisition tool, this parameter can be used to exclude rare words to reduce model noise and improve efficiency. Set it to 1 or higher to exclude words that appear very rarely.

[0094] sg: Select the training algorithm. 0 means using the CBOW (Continuous Bag of Words) algorithm, and 1 means using the Skip-gram algorithm. The Skip-gram algorithm is used in this solution to predict the words in the context of a word, which is convenient for generating high-quality word vectors and further identifying whether it is user feedback information related to the bank.

[0095] 4. Use the Naive Bayes classifier to determine whether the user feedback information that the user has posted is related to the bank. The Naive Bayes classifier determines whether the user feedback information that the user has posted is related to the bank by learning the probability distribution of text features. If the probability is greater than 0.5, the user feedback information is related to the bank; if it is less than 0.5, it is not related to the bank. For irrelevant data, it enters the data temporary storage stage and is no longer analyzed.

[0096] In this embodiment, the user feedback information is visualized, standardized and digitized by using a rigorous algorithm, thereby improving the scientificity and accuracy of classification and judgment.

[0097] In one embodiment, in the above step S203, identifying the feedback keywords in the user feedback information specifically includes the following steps:

[0098] The user feedback information is cleaned and the cleaned user feedback information is segmented using natural language processing technology; the parts of speech of the words in the user feedback information after the word segmentation are marked; according to the parts of speech of the words, the user feedback information after the word segmentation is represented as a word frequency vector, and the words whose word frequency vector exceeds the frequency threshold are used as feedback keywords.

[0099] Among them, natural language processing is an artificial intelligence technology that aims to realize the interaction between computers and natural language. It involves various technologies and methods, including text processing, speech recognition, speech synthesis, machine translation, sentiment analysis, information extraction and question-answering systems.

[0100] Specifically, the server performs text cleaning on the user feedback information, and uses natural language processing technology to perform word segmentation on the cleaned user feedback information; the parts of speech of the words in the user feedback information after the word segmentation are marked; the server represents the user feedback information after the word segmentation as a word frequency vector according to the parts of speech of the words, and uses the words whose word frequency vectors exceed the frequency threshold as feedback keywords.

[0101] In this embodiment, by utilizing natural language processing technology, the computer can quickly and accurately understand user feedback information, thereby more efficiently processing text data and improving the accuracy and efficiency of feedback keyword recognition.

[0102] In one embodiment, if Figure 3 As shown, in the above step S204, sentiment classification of the feedback keywords is performed, which specifically includes the following steps:

[0103] Step S301: input the feedback keywords into the trained Naive Bayes classifier to obtain the sentiment score corresponding to each feedback keyword.

[0104] Step S302: Determine the feedback keywords with a sentiment score greater than the sentiment classification threshold as feedback keywords of the first sentiment type.

[0105] Step S303: Determine the feedback keywords whose sentiment scores are less than the sentiment classification threshold as feedback keywords of the second sentiment type.

[0106] Among them, the naive Bayes classifier is a probabilistic statistical classification algorithm based on Bayes' theorem and the feature conditional independence assumption. It is a simple but effective classification method and is widely used in natural language processing tasks such as text classification and sentiment analysis.

[0107] Specifically, the server inputs the feedback keywords into the trained naive Bayes classifier to obtain the sentiment score corresponding to each feedback keyword; the feedback keywords with sentiment scores greater than the sentiment classification threshold are determined as feedback keywords of the first sentiment type; the feedback keywords with sentiment scores less than the sentiment classification threshold are determined as feedback keywords of the second sentiment type.

[0108] In this embodiment, sentiment analysis technology is used to perform sentiment classification on feedback keywords, thereby improving the accuracy and reliability of sentiment classification, and can also greatly improve processing efficiency, saving time and labor costs.

[0109] In one embodiment, if Figure 4 As shown, the following steps are also included:

[0110] Step S401 : when the preset channel corresponding to the user feedback information is not the internal system of the current financial institution, identifying an external target feedback keyword whose emotion type in the user feedback information is a target emotion type.

[0111] Step S402: determine whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information.

[0112] Step S403 , when the user feedback information corresponding to the external target feedback keyword does not generate abnormal user feedback information, a support vector machine algorithm is used to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information.

[0113] Among them, support vector machine is a very commonly used machine learning algorithm, mainly used for classification and regression analysis. The core idea of ​​the algorithm is to find the maximum margin hyperplane and divide the training data set into different categories.

[0114] Specifically, when the preset channel corresponding to the user feedback information is not the internal system of the current financial institution, the server identifies the external target feedback keyword whose emotion type in the user feedback information is the target emotion type; determines whether the user feedback information corresponding to the external target feedback keyword has generated abnormal user feedback information; and when the user feedback information corresponding to the external target feedback keyword has not generated abnormal user feedback information, uses a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information.

[0115] In this embodiment, the support vector machine algorithm is used to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information, thereby improving the accuracy of the prediction result and increasing the practicality and universality of the solution.

[0116] In one embodiment, in the above step S402, determining whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information specifically includes the following steps:

[0117] The external target feedback keyword is used as the search content, and the corresponding abnormal feedback keyword is determined in the search results; the matching degree of the external target feedback keyword and the abnormal feedback keyword is calculated; and according to the matching degree, it is determined whether the user feedback information corresponding to the external target feedback keyword has generated the user abnormal feedback information.

[0118] Specifically, the server uses the external target feedback keyword as the search content, and determines the corresponding abnormal feedback keyword in the search results; calculates the matching degree of the external target feedback keyword and the abnormal feedback keyword; compares the matching degree with the corresponding threshold, and determines whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information based on the comparison result.

[0119] For example, the server uses the keywords of the user feedback information that the user has posted as the search content. Use the BM25 (Best Matching 25, an algorithm for text retrieval) algorithm to calculate the matching degree between the user feedback keywords and the Internet user feedback keywords. For each user feedback information that the user has posted, calculate the length of the document (number of words) and the frequency of each word. For each keyword in the Internet user feedback keywords, calculate its inverse document frequency. The inverse document frequency indicates the importance of a keyword in the entire user feedback data set. The calculation method is usually:

[0120] IDF(k) = log((N - n(k) + 0.5) / (n(k) + 0.5))

[0121] Where N represents the total number of documents in the user feedback information dataset, and n(k) represents the number of documents containing keyword k;

[0122] For each user feedback keyword that has been published by a user and each Internet user feedback keyword, calculate the BM25 score. The calculation method of the BM25 score is as follows:

[0123] BM25(Q, D) = ∑(IDF(w) * f(w, D) * (k1 + 1)) / (f(w, D) + k1 * (1 - b+ b * |D| / avgdl))

[0124] Among them, BM25(Q, D) represents the BM25 score between query Q and document D. w represents each word in the user feedback information, IDF(w) represents the inverse document frequency of keyword w, f(w, D) represents the word frequency of keyword w in document D, k1 and b are the adjustment parameters of BM25, avgdl represents the average length of the document set, and |D| represents the length of document D.

[0125] A high BM25 score means that the user feedback keywords published by the user have a high match with the current Internet user feedback keywords, that is, user abnormal feedback information has been generated. A low BM25 score means that the user feedback keywords published by the user have a low match with the current Internet user feedback keywords, that is, no user abnormal feedback information has been generated.

[0126] In this embodiment, the corresponding Internet user feedback keywords are determined through information retrieval, and then the matching degree between the external user feedback keywords and the Internet user feedback keywords is calculated; thus, the judgment result of whether the abnormal user feedback information has been generated is obtained efficiently and accurately based on the matching degree parameter index, thereby increasing the reliability of the solution.

[0127] In one embodiment, in the above step S403, using a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information specifically includes the following steps:

[0128] A binary classification model is constructed using the support vector machine algorithm. External target feedback keywords and sample user feedback information in a preset format are input into the binary classification model for inner product operation to obtain similarity values ​​between the external target feedback keywords and sample user feedback information. The similarity values ​​are input into the support vector machine algorithm to obtain prediction results.

[0129] Specifically, the server uses the support vector machine algorithm to build a binary classification model; inputs the external target feedback keywords and sample user feedback information in a preset format into the binary classification model for an inner product operation to obtain a similarity value between the external target feedback keywords and the sample user feedback information; and inputs the similarity value into the support vector machine algorithm to obtain a prediction result.

[0130] For example, the server processes user feedback keywords, where each element represents a text document. The TF-IDF vectorizer is used to transform the text data and convert the text document into a TF-IDF feature vector. The feature vector with a large weight is usable data. The support vector machine algorithm is used to build a classification model. Here, the linear kernel function is chosen. The linear kernel function can be expressed as a simple inner product operation. For the feature vectors x and y, the linear kernel function is calculated as follows:

[0131] K(x,y)=x⋅y

[0132] Among them, K(x,y) represents the inner product in the original feature space, x and y are feature vectors respectively;

[0133] x represents the feature vector of a text document, which contains user feedback information. The feature vector is the result of converting text content into numerical features using the TF-IDF vectorization method.

[0134] y represents the feature vector of another text document. Usually, when predicting, this document is the content of user feedback information related to the bank. Similarly, this feature vector is also the result of converting text content into numerical features through the TF-IDF vectorization method.

[0135] In the support vector machine model, by calculating the inner product (x⋅y) of the two feature vectors, the similarity between them can be evaluated; if the inner product is larger, it means that the feature vectors of the two text documents are closer in the feature space, which may indicate that they are more semantically similar.

[0136] Use the trained support vector machine model to make predictions, which contains a list of prediction results. Each element represents the predicted label of the corresponding new text (1 means generating abnormal user feedback information, and 0 means not generating abnormal user feedback information).

[0137] In this embodiment, the support vector machine algorithm is used to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information, thereby improving the accuracy of the prediction result and increasing the scientific nature of the solution.

[0138] In one embodiment, Figure 5 As shown, a method for processing user feedback information in a financial service in a specific embodiment is provided, which specifically includes the following steps:

[0139] Step S501, obtaining user feedback information from multiple preset channels, and obtaining a text data corpus of the current financial institution; performing vector space division on the text data in the text data corpus according to preset topics, and assigning corresponding classification labels to the division results; based on the division results and the classification labels, identifying user feedback information related to the current financial institution in the user feedback information.

[0140] Step S502: when the user feedback information is user feedback information of the current financial institution, it is determined whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution.

[0141] Step S503, when the preset channel corresponding to the user feedback information is the internal system of the current financial institution, the user feedback information is cleaned, and the cleaned user feedback information is segmented using natural language processing technology; the parts of speech of the words in the user feedback information after the word segmentation are marked; according to the parts of speech of the words, the user feedback information after the word segmentation is represented as a word frequency vector, and the words whose word frequency vectors exceed the frequency threshold are used as feedback keywords.

[0142] Step S504, determining whether the feedback keyword belongs to an existing keyword in the user feedback information library. When the feedback keyword does not belong to an existing keyword in the user feedback information library, inputting the feedback keyword into a trained naive Bayes classifier to obtain a sentiment score corresponding to each feedback keyword.

[0143] Step S505, determine the feedback keywords with sentiment scores greater than the sentiment classification threshold as feedback keywords of the first sentiment type; determine the feedback keywords with sentiment scores less than the sentiment classification threshold as feedback keywords of the second sentiment type, and obtain the target feedback keywords with the target sentiment type.

[0144] Step S506: Obtain corresponding user feedback processing instruction information according to the target feedback keyword.

[0145] Step S507 , when the preset channel corresponding to the user feedback information is not the internal system of the current financial institution, identifying the external target feedback keyword whose emotion type in the user feedback information is the target emotion type.

[0146] Step S508, taking the external target feedback keyword as the search content, and determining the corresponding abnormal feedback keyword in the search results; calculating the matching degree between the external target feedback keyword and the abnormal feedback keyword; and determining whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information according to the matching degree.

[0147] Step S509, when the user feedback information corresponding to the external target feedback keyword does not generate user abnormal feedback information, a binary classification model is constructed using a support vector machine algorithm; the external target feedback keyword and the sample user feedback information in a preset format are input into the binary classification model for an inner product operation to obtain a similarity value between the external target feedback keyword and the sample user feedback information; the similarity value is input into the support vector machine algorithm to obtain a prediction result.

[0148] In the user feedback information processing method in the above-mentioned financial business, the user feedback information is segmented according to the basic dictionary, and then the feedback keywords are determined according to the preset indicators of relevance to the text topic, and the user feedback information is analyzed based on the feedback keywords, so that the user feedback information is visualized in the form of events; thereby, the accuracy of extracting feedback keywords in the user feedback information and the efficiency of analyzing and processing the user feedback information are improved, so that the user feedback information can be followed up and processed in a timely manner.

[0149] In order to more clearly illustrate the method for processing user feedback information in a financial service provided by the embodiment of the present application, the method for processing user feedback information in a financial service is specifically described below with a specific embodiment. Figure 6 As shown, the present application also provides a method for processing user feedback information in a financial service, which specifically includes the following steps:

[0150] Step 1: Obtain user feedback information from multiple preset channels;

[0151] Step 2: Determine whether the user feedback information is related to the bank;

[0152] Step 3: Determine whether the preset channel corresponding to the user feedback information is the bank's internal system;

[0153] Step 4: Identify keywords in user feedback information;

[0154] Step 5: Determine whether the identified keyword belongs to an existing keyword in the user feedback information database;

[0155] Step 6: Determine whether the user feedback information generates abnormal feedback information;

[0156] Step 7: Add keywords to the user feedback information database;

[0157] Step 8: Process according to the classification and rating of keywords;

[0158] Step 9: Set up solutions for different situations;

[0159] Step 10: Identify external target feedback keywords in user feedback information;

[0160] Step 11: determining whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information;

[0161] Step 12: Predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information;

[0162] Step 13: Initiate user abnormal feedback warning and formulate corresponding public relations measures.

[0163] The beneficial effects brought by the above embodiments are as follows:

[0164] This solution performs word segmentation on user feedback information according to a basic dictionary, then determines feedback keywords according to preset indicators of relevance to the text topic, and analyzes the user feedback information based on the feedback keywords, so that the user feedback information is visualized in the form of events; thereby improving the accuracy of feedback keyword extraction in user feedback information and the efficiency of analyzing and processing user feedback information, so that user feedback information can be followed up and processed in a timely manner.

[0165] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0166] Based on the same inventive concept, the embodiment of the present application also provides a user feedback information processing device in a financial service for implementing the user feedback information processing method in the financial service involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the user feedback information processing device in one or more financial services provided below can refer to the limitations of the user feedback information processing method in the financial service above, and will not be repeated here.

[0167] In one embodiment, Figure 7 As shown, a user feedback information processing device in a financial service is provided, comprising:

[0168] The information determination module 701 is used to obtain user feedback information from multiple preset channels and determine whether the user feedback information is user feedback information related to the current financial institution;

[0169] The channel determination module 702 is used to determine whether the preset channel corresponding to the user feedback information is the internal system of the current financial institution when the user feedback information is the user feedback information of the current financial institution;

[0170] The word determination module 703 is used to identify the feedback keywords in the user feedback information and determine whether the feedback keywords belong to existing keywords in the user feedback information database when the preset channel corresponding to the user feedback information is the internal system of the current financial institution;

[0171] The sentiment classification module 704 is used to perform sentiment classification on the feedback keyword when the feedback keyword does not belong to the existing keywords in the user feedback information library, and obtain a target feedback keyword whose sentiment type is the target sentiment type;

[0172] The feedback instruction module 705 is used to obtain corresponding user feedback processing instruction information according to the target feedback keyword.

[0173] In one embodiment, the information judgment module 701 is also used to obtain a text data corpus of the current financial institution; perform vector space division on the text data in the text data corpus according to preset topics, and assign corresponding classification labels to the division results; and identify user feedback information related to the current financial institution in the user feedback information based on the division results and the classification labels.

[0174] In one embodiment, the word judgment module 703 is also used to perform text cleaning on the user feedback information, and use natural language processing technology to perform word segmentation on the cleaned user feedback information; mark the part of speech of the words in the user feedback information after the word segmentation processing; according to the part of speech of the words, the user feedback information after the word segmentation processing is represented as a word frequency vector, and the words whose word frequency vector exceeds the frequency threshold are used as feedback keywords.

[0175] In one embodiment, the sentiment classification module 704 is also used to input feedback keywords into a trained naive Bayes classifier to obtain a sentiment score corresponding to each feedback keyword; feedback keywords with a sentiment score greater than a sentiment classification threshold are determined as feedback keywords of a first sentiment type; feedback keywords with a sentiment score less than the sentiment classification threshold are determined as feedback keywords of a second sentiment type; the first sentiment type and the second sentiment type are two opposite sentiment types.

[0176] In one embodiment, the user feedback information processing device in the financial business also includes an abnormality prediction module, which is used to identify the external target feedback keyword whose emotion type in the user feedback information is the target emotion type when the preset channel corresponding to the user feedback information is not the internal system of the current financial institution; determine whether the user feedback information corresponding to the external target feedback keyword has generated abnormal user feedback information; and use a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information when the user feedback information corresponding to the external target feedback keyword has not generated abnormal user feedback information.

[0177] In one embodiment, the anomaly prediction module is further used to use the external target feedback keyword as the search content and determine the corresponding anomaly feedback keyword in the search results; calculate the matching degree between the external target feedback keyword and the anomaly feedback keyword; and determine whether the user feedback information corresponding to the external target feedback keyword has generated user anomaly feedback information based on the matching degree.

[0178] In one embodiment, the anomaly prediction module is also used to construct a binary classification model using a support vector machine algorithm; input external target feedback keywords and sample user feedback information in a preset format into the binary classification model for an inner product operation to obtain a similarity value between the external target feedback keywords and the sample user feedback information; and input the similarity value into the support vector machine algorithm to obtain a prediction result.

[0179] Each module in the user feedback information processing device in the above financial business can be implemented in whole or in part by software, hardware and their combination. Each module can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0180] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as feedback keywords. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for processing user feedback information in a financial business is implemented.

[0181] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0182] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0183] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0184] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0185] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0186] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0187] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0188] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing user feedback information in financial services, characterized in that: The method comprises: Obtain user feedback information from multiple preset channels, and determine whether the user feedback information is user feedback information related to the current financial institution; When the user feedback information is user feedback information of the current financial institution, determining whether the preset channel corresponding to the user feedback information is an internal system of the current financial institution; When the preset channel corresponding to the user feedback information is the internal system of the current financial institution, identifying feedback keywords in the user feedback information, and determining whether the feedback keywords belong to existing keywords in the user feedback information database; When the feedback keyword does not belong to the existing keywords in the user feedback information library, sentiment classification is performed on the feedback keyword to obtain a target feedback keyword whose sentiment type is a target sentiment type; According to the target feedback keyword, corresponding user feedback processing instruction information is obtained.

2. The method according to claim 1, characterized in that The determining whether the user feedback information is user feedback information related to the current financial institution includes: Acquire a text data corpus of the current financial institution; Performing vector space division on the text data in the text data corpus according to preset topics, and assigning corresponding classification labels to the division results; Based on the division results and the classification labels, the user feedback information related to the current financial institution is identified in the user feedback information.

3. The method according to claim 1, characterized in that The identifying feedback keywords in the user feedback information includes: Performing text cleaning on the user feedback information, and performing word segmentation processing on the cleaned user feedback information using natural language processing technology; Mark the part of speech of words in the user feedback information after word segmentation processing; According to the part of speech of the word, the user feedback information after the word segmentation processing is represented as a word frequency vector, and the words whose word frequency vector exceeds a frequency threshold are used as the feedback keywords.

4. The method according to claim 1, characterized in that: The sentiment classification of the feedback keywords includes: Inputting the feedback keywords into the trained Naive Bayes classifier to obtain the sentiment score corresponding to each feedback keyword; Determine the feedback keywords whose sentiment scores are greater than the sentiment classification threshold as feedback keywords of the first sentiment type; The feedback keyword whose emotion score is less than the emotion classification threshold is determined as a feedback keyword of a second emotion type; the first emotion type and the second emotion type are two emotion types of opposite types.

5. The method according to claim 4, characterized in that The method further comprises: When the preset channel corresponding to the user feedback information is not the internal system of the current financial institution, identifying that the emotion type present in the user feedback information is an external target feedback keyword of the target emotion type; Determining whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information; When the user feedback information corresponding to the external target feedback keyword does not generate abnormal user feedback information, a support vector machine algorithm is used to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information.

6. The method according to claim 5, characterized in that The determining whether the user feedback information corresponding to the external target feedback keyword has generated abnormal user feedback information includes: Using the external target feedback keyword as search content, and determining the corresponding abnormal feedback keyword in the search results; Calculating the matching degree between the external target feedback keyword and the abnormal feedback keyword; According to the matching degree, it is determined whether the user feedback information corresponding to the external target feedback keyword has generated user abnormal feedback information.

7. The method according to claim 6, characterized in that The using of a support vector machine algorithm to predict whether the user feedback information corresponding to the external target feedback keyword will generate abnormal user feedback information includes: Using the support vector machine algorithm, a binary classification model is constructed; Inputting the external target feedback keywords and sample user feedback information in a preset format into the binary classification model to perform an inner product operation to obtain a similarity value between the external target feedback keywords and the sample user feedback information; The similarity value is input into the support vector machine algorithm to obtain a prediction result.

8. A user feedback information processing device in financial services, characterized in that: The device comprises: An information judgment module is used to obtain user feedback information from multiple preset channels and judge whether the user feedback information is user feedback information related to the current financial institution; a channel determination module, configured to determine, when the user feedback information is user feedback information of the current financial institution, whether the preset channel corresponding to the user feedback information is an internal system of the current financial institution; A word determination module, used for identifying feedback keywords in the user feedback information and determining whether the feedback keywords belong to existing keywords in the user feedback information database when the preset channel corresponding to the user feedback information is the internal system of the current financial institution; A sentiment classification module, used for, when the feedback keyword does not belong to an existing keyword in the user feedback information library, performing sentiment classification on the feedback keyword to obtain a target feedback keyword whose sentiment type is a target sentiment type; The feedback instruction module is used to obtain corresponding user feedback processing instruction information according to the target feedback keyword.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images