Client evaluation information analysis method and device and electronic equipment

Abstract

The invention discloses a customer evaluation information analysis method and device and electronic equipment, and relates to the field of big data, the method comprises the following steps: obtaining N documents, and extracting M stem keywords from the N documents; establishing a target matrix according to the M stem keywords and the N documents; performing clustering according to the target matrix and the N documents to obtain a clustering result; and determining an analysis result of the N documents according to a clustering result. Through the method and the device, the problem that the analysis result of the customer comments is inaccurate due to inaccurate keyword extraction and unknown classification of customer feedback themes when the customer comments are analyzed in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of big data, and more specifically, to a method, device and electronic device for analyzing customer evaluation information. Background Art

[0002] After customers purchase relevant wealth management products on the mobile banking client, generally after using it for a period of time, we will guide them to write down their customer experience reviews after purchasing the bank's relevant wealth management products, in order to more accurately understand the market's preferences for the bank's newly launched wealth management products.

[0003] In the prior art, the comments related to the customer experience are initially screened and the three keywords with the highest text frequency are extracted. One disadvantage of the prior art is that the stems are not extracted when doing text mining on customer comments, resulting in some words with the same meaning being divided into different stems when the word frequency is accumulated; moreover, the prior art lacks effective means to screen out high-frequency words, and does not process customer comments according to the text mining process, resulting in the extracted keywords missing key information; in addition, the prior art lacks visual display, lacks the way to display 10 or more high-frequency words appearing in the document through bar charts, and lacks visual clustering of terms; finally, the prior art does not group customer comments, and does not use cluster analysis algorithms to classify tweets, and cannot more clearly display different topics corresponding to different classifications.

[0004] With regard to the problem that when analyzing customer reviews in related technologies, the analysis results of customer reviews are inaccurate due to inaccurate keyword extraction and unclear classification of customer feedback topics, no effective solution has been proposed so far. Summary of the invention

[0005] The main purpose of this application is to provide a method, device and electronic device for analyzing customer evaluation information to solve the problem of inaccurate analysis results of customer comments due to inaccurate keyword extraction and unclear classification of customer feedback topics when analyzing customer comments in related technologies.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for analyzing customer evaluation information is provided, the method comprising: obtaining N documents, extracting M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and M and N are both integers greater than 1; establishing a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; clustering is performed based on the target matrix and the N documents to obtain clustering results; determining the analysis results of the N documents based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions.

[0007] Furthermore, clustering is performed based on the target matrix and the N documents to obtain a clustering result, including: processing the M stem keywords based on the target matrix, and updating the M stem keywords based on the processed stem keywords; hierarchical clustering is performed on the M stem keywords to obtain L clusters, where L is a positive integer; random seeds are determined based on the L clusters, and a target document is randomly selected from the N documents using the random seeds; clustering is performed based on the distance between the stem keywords corresponding to the L clusters and the target document to obtain the clustering result.

[0008] Further, the M stem keywords are processed according to the target matrix, and the M stem keywords are updated according to the processed stem keywords, including: when the number of the M stem keywords is greater than a preset number, stem keywords with a word frequency higher than a preset value are screened from the M stem keywords, and the M stem keywords are updated based on the screened stem keywords; or when the number of the M stem keywords is greater than a preset number, the M stem keywords are filtered using word frequency inverse document frequency, and the M stem keywords are updated based on the filtered stem keywords.

[0009] Furthermore, after establishing a target matrix based on the M stem keywords and the N documents, the method also includes: determining the word frequency of each stem keyword among the M stem keywords based on the target matrix; constructing a target graph based on the word frequency of each stem keyword among the M stem keywords; determining high-frequency keywords based on the word frequency of each stem keyword in the target graph, and calculating the degree of association between the high-frequency keywords through a preset function, wherein the high-frequency keywords are stem keywords whose word frequencies are higher than a preset value; and displaying the degree of association between the target graph and the high-frequency keywords to the target object.

[0010] Furthermore, extracting M stem keywords from the N documents includes: for each document in the N documents, determining the terms contained in the document to obtain a term set for each document, wherein each term in the term set of each document includes at least one word; determining a target root word corresponding to each term in the term set of each document based on a dictionary; and determining the M stem keywords based on the target root word corresponding to each term in each document in the N documents.

[0011] Furthermore, N documents are obtained, including: using a first preset character to replace a second preset character in the customer evaluation text to obtain the N documents, wherein the second preset character is used to connect previous and next terms, and the first preset character is different from the second preset character; determining a target root word corresponding to each term in the term set of each document according to a dictionary, including: for each term in the term set of each document in the N documents, calculating the term frequency; when the term frequency is greater than a preset value, determining the term as the target root word; when the term frequency is less than or equal to the preset value, determining the term with the highest frequency among the synonyms of the term in the dictionary as the target root word.

[0012] Furthermore, a target matrix is ​​established based on the M stem keywords and the N documents, including: constructing a first matrix, wherein the first matrix includes M rows and N columns; for each document in the N documents, calculating the frequency of each stem keyword in the M stem keywords in the document, and obtaining the frequency of each stem keyword in the N documents; constructing each element of the first matrix based on the frequency of each stem keyword in the N documents, and obtaining the target matrix.

[0013] Furthermore, obtaining N documents includes: collecting N evaluation texts from multiple customers, and translating the N evaluation texts to obtain N first texts; converting each of the N first texts into a preset data structure to obtain N second texts; performing data preprocessing operations on the N second texts respectively to obtain the N documents, wherein the data preprocessing operations at least include: converting character formats, deleting preset characters, and deleting non-semantic characters.

[0014] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for analyzing customer evaluation information is provided, which includes: an acquisition unit, used to acquire N documents, and extract M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and M and N are both integers greater than 1; an establishment unit, used to establish a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; a clustering unit, used to cluster based on the target matrix and the N documents to obtain a clustering result; a first determination unit, used to determine the analysis result of the N documents based on the clustering result, wherein the analysis result includes at least one of the following: satisfaction level, customer suggestion.

[0015] Furthermore, the clustering unit includes: a first processing subunit, used to process the M stem keywords according to the target matrix, and update the M stem keywords according to the processed stem keywords; a first clustering subunit, used to hierarchically cluster the M stem keywords to obtain L clusters, where L is a positive integer; a selection subunit, used to determine a random seed based on the L clusters, and use the random seed to randomly select a target document from the N documents; a second clustering subunit, used to cluster based on the distance between the stem keywords corresponding to the L clusters and the target document to obtain the clustering result.

[0016] Furthermore, the first processing sub-unit includes: a screening module, which is used to screen stem keywords with a word frequency higher than a preset value from the M stem keywords when the number of the M stem keywords is greater than a preset number, and update the M stem keywords based on the screened stem keywords; or a filtering module, which is used to filter the M stem keywords by using word frequency inverse document frequency when the number of the M stem keywords is greater than a preset number, and update the M stem keywords based on the filtered stem keywords.

[0017] Furthermore, the device also includes: a second determination unit, which is used to determine the word frequency of each of the M stem keywords according to the target matrix after establishing a target matrix according to the M stem keywords and the N documents; a construction unit, which is used to construct a target graph according to the word frequency of each of the M stem keywords; a calculation unit, which is used to determine high-frequency keywords according to the word frequency of each stem keyword in the target graph, and calculate the degree of association between the high-frequency keywords through a preset function, wherein the high-frequency keyword is a stem keyword whose word frequency is higher than a preset value; and a display unit, which is used to display the degree of association between the target graph and the high-frequency keyword to the target object.

[0018] Furthermore, the acquisition unit includes: a first determination subunit, used to determine, for each document in the N documents, the terms contained in the document, and obtain a term set for each document, wherein each term in the term set of each document includes at least one word; a second determination subunit, used to determine, based on a dictionary, a target root word corresponding to each term in the term set of each document; and a third determination subunit, used to determine the M stem keywords based on the target root word corresponding to each term in each document in the N documents.

[0019] Furthermore, the acquisition unit includes: a replacement subunit, used to replace a second preset character in the customer evaluation text with a first preset character to obtain the N documents, wherein the second preset character is used to connect the previous and next terms, and the first preset character is different from the second preset character; a second determination subunit includes: a calculation module, used to calculate the word frequency of each term in the term set of each document in the N documents; a first determination module, used to determine the term as the target root word when the word frequency is greater than a preset value; a second determination module, used to determine the term with the highest frequency among the synonyms of the term in the dictionary as the target root word when the word frequency is less than or equal to the preset value.

[0020] Furthermore, the establishment unit includes: a first construction subunit, used to construct a first matrix, wherein the first matrix includes M rows and N columns; a calculation subunit, used to calculate the frequency of each stem keyword among the M stem keywords in the document for each document in the N documents, and obtain the frequency of each stem keyword in the N documents; a second construction subunit, used to construct each element of the first matrix according to the frequency of each stem keyword in the N documents, and obtain the target matrix.

[0021] Furthermore, the acquisition unit includes: a collection subunit, used to collect N evaluation texts from multiple customers, and translate the N evaluation texts to obtain N first texts; a conversion subunit, used to convert each of the N first texts into a preset data structure to obtain N second texts; a second processing subunit, used to perform data preprocessing operations on the N second texts respectively to obtain the N documents, wherein the data preprocessing operations at least include: converting character formats, deleting preset characters, and deleting semanticless characters.

[0022] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer program product is provided, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for analyzing customer evaluation information, and which, when executed by a processor, implements the steps of the method for analyzing customer evaluation information described in each embodiment of the present application.

[0023] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes stored computer instructions, wherein when the computer instructions are executed by a processor, any one of the above-mentioned methods for analyzing customer evaluation information is implemented.

[0024] In order to achieve the above-mentioned purpose, according to one aspect of the present application, an electronic device is provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement any one of the above-mentioned methods for analyzing customer evaluation information.

[0025] In an embodiment of the present application, N documents are obtained, M stem keywords are extracted from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and M and N are both integers greater than 1; a target matrix is ​​established based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; clustering is performed based on the target matrix and the N documents to obtain clustering results; analysis results of the N documents are determined based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions, thereby solving the technical problem of inaccurate analysis results of customer comments due to inaccurate keyword extraction and unclear classification of customer feedback topics when analyzing customer comments.

[0026] By extracting M stem keywords from customer reviews and establishing a target matrix, the core content and theme of the review text can be determined, avoiding the changes in lexical form (such as different tenses of verbs) and inconsistent use of synonyms that may be caused by directly using the original vocabulary. The relationship between M stem keywords and N documents is clearly displayed, thereby improving the accuracy of customer review analysis. At the same time, the documents are divided into several groups through clustering algorithms, revealing the customer groups' satisfaction and suggestions for financial products, which helps companies understand the needs and pain points of different customer groups, thereby helping financial institutions to optimize products or services in a targeted manner and improve customer satisfaction and market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0028] Figure 1 This is a hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for analyzing customer evaluation information provided in Example 1 of the present application;

[0029] Figure 2 is a flowchart of an optional method for analyzing customer evaluation information provided in Example 1 of the present application;

[0030] Figure 3is a schematic diagram of an optional customer evaluation information mining process provided in Example 1 of the present application;

[0031] Figure 4 It is a software structure diagram of an optional method for analyzing customer evaluation information provided in Example 1 of the present application;

[0032] Figure 5 is a schematic diagram of a device for analyzing customer evaluation information provided in Embodiment 2 of the present application;

[0033] Figure 6 It is a schematic diagram of an electronic device for analyzing customer evaluation information provided in Embodiment 5 of the present application. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] It should be noted that the methods and devices determined by the processing methods, devices, storage media and electronic devices of the present application documents can be used in the field of financial technology to improve the accuracy of the analysis results of customer reviews in the process of analyzing customer reviews, and can also be used in any field except the field of financial technology. The application fields of the methods and devices of the processing methods, devices, storage media and electronic devices of the present application documents are not limited.

[0036] It should be noted that the collected information (including but not limited to user device information, user personal information, collected data, used data, generated data, processed data, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, collected information, used information, generated information, processed information, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with the relevant laws, regulations and standards of relevant countries and regions, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation entrances for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0037] Example 1

[0038] According to an embodiment of the present application, a method embodiment for analyzing customer evaluation information is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The method embodiment provided in Embodiment 1 of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for analyzing customer evaluation information is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (102a, 102..., 102n are used to illustrate in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (US) port (which may be included as one of the ports of the US bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0040] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method for analyzing customer evaluation information in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the method for analyzing customer evaluation information described above is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0044] Under the above operating environment, this application provides Figure 2 The analysis method of the customer evaluation information is shown. Figure 2 It is a flowchart of the method for analyzing customer evaluation information according to Example 1 of the present application.

[0045] The target system can be the subject of the analysis method of customer evaluation information provided in Example 1 of the present application. The target system combines text mining technology and clustering algorithms to analyze customer reviews of various financial products of the bank, and comprehensively mines and displays customer feedback on financial products through steps such as language and text conversion, word stem extraction, establishment of term-document matrix, frequent term and association analysis, term clustering and tweet clustering, so as to help business personnel more accurately understand the market's preference for financial products, identify product advantages and disadvantages, and thus contribute to product optimization and improvement of financial products.

[0046] Step S101, obtaining N documents, and extracting M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and both M and N are integers greater than 1.

[0047] When performing text mining analysis of customer evaluations in this embodiment 1, it is necessary to perform preprocessing (such as language conversion, text cleaning, word segmentation, removal of stop words, etc.) operations on the original text data of customers' evaluations of financial products to obtain a processed customer evaluation text data set, namely the N documents mentioned above.

[0048] Then, from these N documents, stemming technology is used for further processing, the purpose is to unify different lexical forms with the same root, so as to avoid the dispersion of word frequency statistics due to different lexical forms. Exemplarily, stem keywords are extracted from each document of N documents based on word frequency to form a set of M stem keywords, where N and M are both integers greater than 1, representing the number of processed documents and the number of extracted stem keywords, respectively.

[0049] Through the above process, keywords in customer evaluation information can be extracted, providing a data basis for subsequent high-frequency vocabulary recognition, term association analysis, term clustering and document clustering, ensuring the accuracy and effectiveness of data mining, and helping to deeply understand the key information and implicit themes in customer evaluations.

[0050] Step S102: establishing a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to represent the association relationship between the M stem keywords and each document in the N documents.

[0051] In the text mining process of this embodiment 1, the M stem keywords that have been stemmed are combined with N preprocessed documents (i.e., customer evaluation texts) to create a term-document matrix, i.e., the target matrix mentioned above. The target matrix is ​​used to quantify and characterize the frequency or importance of the stem keywords in each document, as well as the correlation between these stem keywords.

[0052] In the target matrix, the rows represent M stem keywords and the columns represent N documents. Each element in the matrix represents the frequency of the corresponding stem keyword in a specific document or the value after some weighting process (such as TF-IDF). In this way, the stem keywords that appear frequently in each document can be determined, which helps analyze the main concerns and preferences of customers for financial products.

[0053] In addition, the target matrix is ​​used to analyze the association between stem keywords. This involves calculating the similarity or distance between terms, and determining how semantically related they are by analyzing the frequency of co-occurrence of terms in multiple documents. For example, if certain stem keywords often appear together in the same document, the association between them is considered high, which helps to reveal implicit themes and patterns in the text.

[0054] By constructing the target matrix, it not only reflects the distribution of stem keywords in the document collection, but also reveals the semantic associations between stem keywords, providing a data basis for subsequent clustering analysis and analysis, while helping business personnel quickly identify hot issues and product advantages in user feedback.

[0055] Step S103, clustering is performed based on the target matrix and the N documents to obtain a clustering result.

[0056] After constructing the target matrix in this embodiment 1, cluster analysis is performed on the N documents to discover the internal structure and pattern between the documents. For example, the 1200 customer reviews (i.e., the N documents mentioned above) after processing are grouped, and the similarity or distance between the documents is calculated by quantifying the frequency or importance of the stem keywords in each document, so that documents with similar features or themes are classified into the same category.

[0057] Clustering algorithms can clarify the potential semantic information in a document collection and intuitively display the subject of the document, thereby further analyzing the characteristics of each cluster and gaining a more comprehensive understanding of customer experience and needs.

[0058] Step S104, determining analysis results of the N documents based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestion.

[0059] In this embodiment 1, after completing the cluster analysis of N documents (i.e., customer evaluation texts), the customer satisfaction and suggestions reflected by the document set of each cluster are determined based on the cluster information obtained by clustering. For example, a cluster after clustering contains keywords such as "High interest rates", "High profit" and "IntegralFeedback", indicating that the documents in this cluster mainly reflect the customer's positive evaluation of the high returns and reward mechanism of the financial product, from which it can be inferred that the customer satisfaction of the cluster is high.

[0060] Interpreting N documents through cluster analysis can help business personnel quickly extract valuable information from massive customer reviews, which is helpful for optimizing financial products and improving customer service.

[0061] Optionally, in the method for analyzing customer evaluation information provided in Example 1 of the present application, clustering is performed based on the target matrix and N documents to obtain a clustering result, including: processing M stem keywords based on the target matrix, and updating the M stem keywords based on the processed stem keywords; hierarchically clustering the M stem keywords to obtain L clusters, where L is a positive integer; determining a random seed based on the L clusters, and randomly selecting a target document from the N documents using the random seed; and clustering based on the distance between the stem keywords corresponding to the L clusters and the target document to obtain a clustering result.

[0062] In this embodiment 1, after the target matrix, the M stem keywords are further processed, which may include frequency supplement weight calculation to ensure that the statistics of the stem keywords are more accurate and reflect the actual importance and frequency of occurrence of the words in the document. After processing, the M stem keyword list may be updated, for example, by removing low-frequency words or adding words optimized by frequency supplement.

[0063] Then, a hierarchical clustering algorithm is used to perform cluster analysis on the processed stem keywords, aiming to discover the internal connections and thematic structures between the keywords. Hierarchical clustering can generate a tree diagram that shows the distances between the M stem keywords, and finally divides the M stem keywords into L clusters, where L is a positive integer representing the number of clusters. In an optional embodiment, L can be set to 4, that is, the M stem keywords are clustered into 4 clusters. It should be noted that the value of L can be flexibly adjusted according to the actual production situation, and is not specifically limited in this embodiment 1.

[0064] Secondly, based on the L clusters obtained by hierarchical clustering, a random seed (an initial value for random number generation) is selected, and the random seed is used to randomly select target documents from the N documents for subsequent K-Means clustering. The purpose of this step is to provide an initial state for subsequent K-Means clustering to ensure the repeatability and stability of the clustering results. The selection of random seeds can be based on certain features of the L clusters, such as the combination of the most frequent keywords or the center of the cluster.

[0065] Finally, after determining the target document, the K-Means clustering algorithm is used to classify the stem keywords contained in the target document based on the word frequency information of the stem keywords in the target matrix. The K-Means algorithm divides the stem keywords contained in the target document into L clusters based on the distance between the stem keywords and the cluster center. The final clustering result reflects the grouping of stem keywords around different themes or features in the randomly selected target documents, which helps to deeply understand the core themes in mining and displaying customer reviews, and helps to optimize financial products and improve services.

[0066] By combining hierarchical clustering with K-Means clustering (also known as term clustering), the diversity and stability of clustering results are ensured, which can more comprehensively reflect the real opinions and suggestions of customers.

[0067] Optionally, in the customer evaluation information analysis method provided in Example 1 of the present application, M stem keywords are processed according to a target matrix, and the M stem keywords are updated based on the processed stem keywords, including: when the number of M stem keywords is greater than a preset number, stem keywords with a word frequency higher than a preset value are screened from the M stem keywords, and the M stem keywords are updated based on the screened stem keywords; or when the number of M stem keywords is greater than a preset number, the M stem keywords are filtered using word frequency inverse document frequency, and the M stem keywords are updated based on the filtered stem keywords.

[0068] In this embodiment 1, when the number of M stem keywords is too large and exceeds the preset number (for example, if 123 terms are initially screened, it exceeds the preset 100 terms, and the preset number of terms can be adjusted according to actual conditions), the processing speed of the model and the interpretability of the model are affected. Therefore, it is necessary to screen or filter the M stem keywords to reduce the number of terms.

[0069] In an optional embodiment, a screening method based on term frequency (TF) can be used, that is, the frequency of occurrence of words in a document. Exemplarily, if the number of M stem keywords is too large, a term frequency threshold can be set to retain only those stem keywords whose frequency of occurrence is higher than a preset value. For example, terms that appear less than 10 times will be discarded, which ensures that the focus of subsequent analysis is on more important keywords, avoiding the interference of low-frequency words on the analysis results. The screened stem keyword set will be used to update the M stem keywords for subsequent clustering analysis.

[0070] In another optional embodiment, a screening method based on term frequency-inverse document frequency (TF-IDF) can be used. TF-IDF is a weight calculation method that not only considers the frequency of terms in a document, but also the prevalence of terms in a document. The higher the frequency of a term in a document and the lower the frequency of its appearance in other documents, the higher its TF-IDF value, indicating that it is more unique and important to the document. If the number of M stem keywords is too large, TF-IDF can be used for filtering to remove those words that appear frequently in many documents but have a low contribution to the topic of a specific document, and retain the stem keywords that appear frequently in a single document and appear less frequently in other documents. The filtered stem keywords will be used to update the M stem keywords to optimize subsequent clustering analysis.

[0071] By filtering based on word frequency or TF-IDF, the number of stem keywords is reduced, so that we can focus on stem keywords that contribute more to the document topic, thereby improving the efficiency and accuracy of text analysis and ensuring that subsequent clustering analysis can more clearly reveal customers' real feedback and potential needs for financial products.

[0072] Optionally, in the method for analyzing customer evaluation information provided in Example 1 of the present application, after establishing a target matrix based on M stem keywords and N documents, the method further includes: determining the word frequency of each stem keyword among the M stem keywords based on the target matrix; constructing a target graph based on the word frequency of each stem keyword among the M stem keywords; determining high-frequency keywords based on the word frequency of each stem keyword in the target graph, and calculating the degree of association between high-frequency keywords through a preset function, wherein high-frequency keywords are stem keywords whose word frequency is higher than a preset value; and displaying the degree of association between the target graph and the high-frequency keywords to the target object.

[0073] In this embodiment 1, in order to help business personnel identify the topics and customer concerns in customer reviews, the frequency of occurrence of each stem keyword in M ​​stem keywords in N documents can be calculated based on the target matrix. The word frequency reflects the importance of a single word in a document. A higher word frequency means that the stem keyword appears more frequently and may be a hot spot or focus of customer concern.

[0074] Then, a target graph, usually a bar chart, can be constructed based on the calculated word frequency data. The bar chart can intuitively display the frequency of each stem keyword, allowing business personnel to intuitively and clearly understand customer concerns. Based on the constructed bar chart, a word frequency threshold (i.e., the preset value mentioned above) can be set to identify stem keywords with a frequency higher than this threshold as high-frequency keywords. These high-frequency keywords usually contain the most core feedback information in customer reviews. Afterwards, the preset function (for example, the findAccess function in the R language) is used to calculate the degree of association between high-frequency keywords. The calculation of the degree of association can be based on the co-occurrence frequency of words in the document, reflecting whether high-frequency keywords often appear together in the text context and the strength of the connection between them.

[0075] Finally, the constructed target graph and the calculated correlation between high-frequency keywords are displayed to the target object (e.g., business personnel, business managers, or data analysis teams) (i.e., visual clustering of terms). The graphical display not only helps to intuitively understand the importance of high-frequency keywords, but the display of correlation also reveals the potential connection between keywords, which helps to deeply analyze the specific evaluation and feelings of customers on financial products, and provide data support for product optimization and improving customer satisfaction.

[0076] Through this series of steps, valuable information can be effectively extracted from a large number of customer reviews, and presented in the form of easy-to-understand graphics and correlation data, thereby providing business personnel with a clear perspective to understand customers' main feedback on financial products and identify product advantages and areas for improvement.

[0077] Optionally, in the method for analyzing customer evaluation information provided in Example 1 of the present application, M stem keywords are extracted from N documents, including: for each document in the N documents, determining the terms contained in the document to obtain a term set for each document, wherein each term in the term set of each document includes at least one word; determining a target root word corresponding to each term in the term set of each document based on a dictionary; and determining M stem keywords based on the target root word corresponding to each term in each document in the N documents.

[0078] In this embodiment 1, each of the N documents may be preprocessed to identify all terms contained therein, and construct a term set for each document. A term here refers to any meaningful word that appears in a document, which may be a single word or a phrase (i.e., multiple words). For example, in the customer review "Product A has high returns and good safety", "Product A", "high returns", "safety", and "very good" may all be considered as terms.

[0079] Then, after determining the set of terms in the document, determine the target root corresponding to each term based on a dictionary (e.g., an English dictionary or a specialized root dictionary). The root refers to the basic form of a word, the core part after removing the affixes, which helps to unify different forms of words with the same basic meaning during statistics and analysis. For example, the terms "safety" and "secure" may be mapped to the same target root in the dictionary, such as "safe".

[0080] Finally, by analyzing the target root word corresponding to each term in each document in N documents, the root words that appear are summarized to determine M stem keywords. Stem keywords refer to more representative and general key words extracted from the term set after root word mapping. These keywords can reflect the theme of the document and are of great significance for subsequent text analysis, such as cluster analysis and sentiment analysis. For example, if terms related to the root word "income" frequently appear in multiple documents, "income" can be extracted as a stem keyword and become the focus of subsequent analysis.

[0081] Through this series of steps, the text mining process can effectively extract M representative stem keywords from the original N documents, providing a basis for subsequent in-depth analysis. The process of root mapping and keyword extraction helps to reduce vocabulary redundancy and improve the efficiency and accuracy of analysis, thereby more clearly showing the themes and patterns in customer reviews.

[0082] Optionally, in the method for analyzing customer evaluation information provided in Example 1 of the present application, N documents are obtained, including: using a first preset character to replace a second preset character in the customer evaluation text to obtain N documents, wherein the second preset character is used to connect the previous and next terms, and the first preset character is different from the second preset character; determining the target root word corresponding to each term in the term set of each document according to the dictionary, including: for each term in the term set of each document in the N documents, calculating the term frequency; when the term frequency is greater than a preset value, determining the term as the target root word; when the term frequency is less than or equal to the preset value, determining the term with the highest frequency among the synonyms of the term in the dictionary as the target root word.

[0083] In the text mining process of this embodiment 1, the original text needs to be preprocessed first to eliminate noise and standardize vocabulary in the text. In N customer evaluation documents, a preset character (such as a space) is used to replace another preset character (such as a comma or a hyphen), which is usually used to connect terms or phrases. For example, if there is "Long-cycle" in the original text, it will become "Long cycle" after the replacement operation. This step helps to ensure that subsequent vocabulary processing and stem extraction can be carried out more accurately.

[0084] Then, for each term in the term set of each document, calculate its frequency of occurrence in the document, that is, the term frequency (TF). The term frequency reflects the importance of the term in the document and is the basis for subsequent keyword extraction and stem determination. Calculating the term frequency usually involves counting the number of times each term appears and then dividing it by the total number of terms in the document.

[0085] After calculating the word frequency, the target root is determined based on the word frequency and the preset value. If the word frequency of the word is greater than the preset value, it means that the word appears more frequently in the document, and the word is directly determined as a target root. However, if the word frequency is less than or equal to the preset value, it means that the importance of the word is low, and it is necessary to find synonyms of the word in the dictionary, and determine the word with the highest frequency among the synonyms in the dictionary as the target root. This process helps to reduce the complexity of text analysis and focus on important words in the text without losing information.

[0086] Through this series of text processing and word frequency analysis steps, we can effectively extract word roots with high word frequency from the original customer evaluation text as the basis for subsequent text analysis and mining. This method not only considers the frequency of occurrence of words in the document, but also introduces a synonym processing mechanism to ensure that the selection of word roots can reflect the most frequently occurring semantic information, providing more accurate and powerful support for subsequent advanced analysis such as clustering analysis and topic extraction.

[0087] Optionally, in the customer evaluation information analysis method provided in Example 1 of the present application, a target matrix is ​​established based on M stem keywords and N documents, including: constructing a first matrix, wherein the first matrix includes M rows and N columns; for each document in the N documents, calculating the frequency of each stem keyword in the M stem keywords in the document, and obtaining the frequency of each stem keyword in the N documents; constructing each element of the first matrix based on the frequency of each stem keyword in the N documents, and obtaining the target matrix.

[0088] In this embodiment 1, a first matrix with M rows and N columns and empty elements is created, where M represents the number of stem keywords retained after stem extraction and frequency screening, and N represents the number of documents. Each row of the first matrix corresponds to a stem keyword, and each column corresponds to a document.

[0089] For each of the N documents, the frequency of occurrence of the M stem keywords needs to be counted. Frequency refers to the number of times a stem keyword appears in a document. This step traverses each document and matches the words that appear in it. If a word in the M stem keyword list is matched, the count of the word in the corresponding document is increased. For example, if the frequency of the stem keyword "income" in all 1,200 comments (i.e., N documents) is counted, then the number of times the word "income" appears in each comment needs to be found.

[0090] The elements of the first matrix are filled according to the frequency information of each stem keyword in N documents to form a target matrix. Each element in the target matrix represents the frequency relationship between the stem keyword and the document. For example, the element in the i-th row and j-th column in the matrix represents the frequency of occurrence of stem keyword i in document j. In an optional embodiment, some elements in the first matrix whose frequency is lower than a preset threshold (for example, the preset threshold is 10) can be deleted from the first matrix to reduce the impact of noise vocabulary on the analysis results.

[0091] Through this series of steps, the target matrix can effectively quantify the relationship between words and documents, providing a basis for subsequent text mining tasks such as clustering analysis and association rule mining. Each row in the target matrix can be regarded as a vector representation of a term, and each column can be regarded as a vector representation of a document. This information is the key to advanced text analysis.

[0092] Optionally, in the method for analyzing customer evaluation information provided in Example 1 of the present application, obtaining N documents includes: collecting N evaluation texts from multiple customers, and translating the N evaluation texts to obtain N first texts; converting each of the N first texts into a preset data structure to obtain N second texts; performing data preprocessing operations on the N second texts respectively to obtain N documents, wherein the data preprocessing operations include at least: converting character format, deleting preset characters, and deleting non-semantic characters.

[0093] In this embodiment 1, N evaluation texts are collected from multiple customers. These N evaluation texts may be in Chinese or other languages. Therefore, these non-English evaluation texts need to be translated into English to obtain N first texts to ensure that the text content can be understood by the subsequent English text mining algorithm.

[0094] Each of the N first texts is converted into a preset data structure, such as a data frame or a corpus, to obtain N second texts. A data frame is a tabular data structure commonly used in data analysis algorithms such as R language, which facilitates data manipulation and analysis; a corpus is a collection of text data, which is a common data representation form in text mining and natural language processing. This step organizes text data into a structure that is easier to count and analyze.

[0095] Data preprocessing operations are performed on the N second texts to clean and standardize the text data to obtain N documents that are ultimately used for text mining analysis. Data preprocessing includes at least the following operations: converting character formats, that is, converting all characters in the text to a unified format, such as converting all letters to lowercase, so as to avoid vocabulary statistics errors caused by different uppercase and lowercase letters; deleting preset characters, where preset characters usually refer to punctuation marks, numbers or specific special characters. These characters often do not contain semantic information in text mining, but will interfere with the accurate statistics and subsequent analysis of vocabulary. For example, deleting punctuation marks, numbers and stop characters, etc., as well as deleting, are all to reduce interference; deleting non-semantic characters, that is, removing characters that do not carry semantic information, such as HTTP-starting strings, extra spaces, line breaks or other non-text characters, to ensure that each word item is a meaningful word or phrase.

[0096] By collecting, translating and preprocessing customer evaluation texts, we can convert the original texts into structured, clean English text data, providing high-quality input for subsequent text mining steps (such as stem extraction, target matrix construction, etc.), thereby ensuring the accuracy and effectiveness of the analysis results.

[0097] Optionally, in this embodiment 1, Figure 3 It is a schematic diagram of the optional mining process of customer evaluation information provided in Example 1 of the present application. First of all, this solution mainly considers and analyzes from four dimensions. The first dimension is to pay attention to extracting stems to avoid dividing some words with the same meaning into different stems when accumulating word frequencies; the second dimension is to use a bar chart to clearly display the ten most frequently appearing words, and at the same time find out the words with the highest correlation with the specified words, so as to have a more comprehensive understanding of customers' evaluation of financial products; the third dimension is to display high-frequency words through bar charts, word visualization clustering and other dimensions, so as to intuitively understand customers' evaluation of financial products; the fourth dimension is to clearly display different themes corresponding to different categories, and extract keywords from multiple dimensions such as product yield, product price, and target audience, which is conducive to helping business personnel to fully understand bank financial products.

[0098] Then, this solution uses text mining to process customer review data, including: converting language and text, extracting stems, establishing a term-document matrix (i.e., the target matrix mentioned above), determining frequent terms and associations, and performing term clustering and tweet clustering.

[0099] Secondly, this solution specifically includes: converting language and text, converting user comments from Chinese to English; extracting stems, introducing a frequency-complement weight calculation mechanism to extract stem keywords; establishing a term-document matrix to provide a data source for subsequent text mining; marking the occurrence of multiple high-frequency word combinations in comments, and displaying multiple terms with high term correlation; dividing user comments into 4 clusters through term clustering, dividing stem keywords into 4 clusters through tweet clustering, and determining the most frequently occurring terms in each cluster.

[0100] Finally, the customers’ evaluation of financial products is determined from the four clusters in the clustering results. For example, it is determined that a certain financial product has defects such as a long repayment period and difficulty in making reservations, but the product also has advantages such as a higher safety factor and higher returns.

[0101] Optionally, in this embodiment 1, Figure 4 It is a software structure diagram of the optional customer evaluation information analysis method provided in Example 1 of the present application, showing the logical connection between each module. Among them, the entire software structure includes a text mining algorithm design module, a K-MEANS clustering algorithm improvement module and a financial product evaluation improvement module. By standardizing text data, improving keyword extraction, optimizing matrix size and deep clustering analysis, the analysis efficiency and depth of the bank's comments on the user experience of financial products are effectively improved. Specifically, the improved stem extraction combined with frequency weight calculation makes the selection of keywords more accurate and can more truly reflect the main purpose of the comment. The optimization of the term-document matrix reduces noise by eliminating low-frequency words and using TF-IDF filtering, focusing on high-frequency words that are crucial to understanding customer experience. The frequent terms and association analysis, as well as the clustering of terms and tweets, further refine the theme of customer feedback, helping banks identify the specific needs and concerns of different user groups, which is the key to improving customer satisfaction and product service quality.

[0102] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0103] In summary, the method for analyzing customer evaluation information provided in the embodiment of the present application obtains N documents, extracts M stem keywords from the N documents, wherein each of the N documents is a document processed from the customer evaluation text, and M and N are both integers greater than 1; a target matrix is ​​established based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; clustering is performed based on the target matrix and the N documents to obtain clustering results; and analysis results of the N documents are determined based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions, which solves the problem of inaccurate analysis results of customer comments due to inaccurate keyword extraction and unclear classification of customer feedback topics when analyzing customer comments in related technologies.

[0104] By extracting M stem keywords from customer reviews and establishing a target matrix, the core content and theme of the review text can be determined, avoiding the changes in lexical form (such as different tenses of verbs) and inconsistent use of synonyms that may be caused by directly using the original vocabulary. The relationship between M stem keywords and N documents is clearly displayed, thereby improving the accuracy of customer review analysis. At the same time, the documents are divided into several groups through clustering algorithms, revealing the customer groups' satisfaction and suggestions for financial products, which helps companies understand the needs and pain points of different customer groups, thereby helping financial institutions to optimize products or services in a targeted manner and improve customer satisfaction and market competitiveness.

[0105] Example 2

[0106] The embodiment of the present application also provides a device for analyzing customer evaluation information. It should be noted that the device for analyzing customer evaluation information in the embodiment of the present application can be used to execute the method for analyzing customer evaluation information provided in the embodiment of the present application. The device for analyzing customer evaluation information provided in the embodiment of the present application is introduced below.

[0107] According to an embodiment of the present application, a device for implementing the above-mentioned method for analyzing customer evaluation information is also provided. Figure 5 As shown, the device comprises:

[0108] Specifically, the acquisition unit 501 is used to acquire N documents and extract M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and both M and N are integers greater than 1.

[0109] The establishing unit 502 is used to establish a target matrix according to the M stem keywords and the N documents, wherein the target matrix is ​​used to represent the association relationship between the M stem keywords and each document in the N documents.

[0110] The clustering unit 503 is used to perform clustering according to the target matrix and the N documents to obtain a clustering result.

[0111] The first determining unit 504 is used to determine the analysis results of the N documents according to the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestion.

[0112] The analysis device for customer evaluation information provided in the embodiment of the present application obtains N documents through an acquisition unit 501, extracts M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and M and N are both integers greater than 1; an establishment unit 502 establishes a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; a clustering unit 503 performs clustering based on the target matrix and the N documents to obtain a clustering result; a first determination unit 504 determines the analysis results of the N documents based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions, and solves the problem in the related art that when analyzing customer comments, the analysis results of customer comments are inaccurate due to inaccurate keyword extraction and unclear classification of customer feedback topics.

[0113] By extracting M stem keywords from customer reviews and establishing a target matrix, the core content and theme of the review text can be determined, avoiding the changes in lexical form (such as different tenses of verbs) and inconsistent use of synonyms that may be caused by directly using the original vocabulary. The relationship between M stem keywords and N documents is clearly displayed, thereby improving the accuracy of customer review analysis. At the same time, the documents are divided into several groups through clustering algorithms, revealing the customer groups' satisfaction and suggestions for financial products, which helps companies understand the needs and pain points of different customer groups, thereby helping financial institutions to optimize products or services in a targeted manner and improve customer satisfaction and market competitiveness.

[0114] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned clustering unit 503 includes: a first processing sub-unit, used to process M stem keywords according to the target matrix, and update the M stem keywords according to the processed stem keywords; a first clustering sub-unit, used to hierarchically cluster the M stem keywords to obtain L clusters, where L is a positive integer; a selection sub-unit, used to determine a random seed based on the L clusters, and randomly select a target document from N documents using the random seed; a second clustering sub-unit, used to cluster based on the distance between the stem keywords corresponding to the L clusters and the target document to obtain a clustering result.

[0115] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned first processing sub-unit includes: a screening module, which is used to screen stem keywords with a word frequency higher than a preset value among the M stem keywords when the number of M stem keywords is greater than a preset number, and update the M stem keywords based on the screened stem keywords; or a filtering module, which is used to filter the M stem keywords by using word frequency inverse document frequency when the number of M stem keywords is greater than a preset number, and update the M stem keywords based on the filtered stem keywords.

[0116] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned device also includes: a second determination unit, which is used to determine the word frequency of each stem keyword among the M stem keywords according to the target matrix after establishing a target matrix based on M stem keywords and N documents; a construction unit, which is used to construct a target graph based on the word frequency of each stem keyword among the M stem keywords; a calculation unit, which is used to determine high-frequency keywords based on the word frequency of each stem keyword in the target graph, and calculate the degree of association between high-frequency keywords through a preset function, wherein the high-frequency keyword is a stem keyword whose word frequency is higher than a preset value; and a display unit, which is used to display the degree of association between the target graph and the high-frequency keywords to the target object.

[0117] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned acquisition unit 501 includes: a first determination subunit, used to determine, for each document in N documents, the terms contained in the document, and obtain a term set for each document, wherein each term in the term set of each document includes at least one word; a second determination subunit, used to determine, based on a dictionary, a target root word corresponding to each term in the term set of each document; and a third determination subunit, used to determine M stem keywords based on the target root word corresponding to each term in each document in the N documents.

[0118] Optionally, in the analysis device for customer evaluation information provided in Example 2 of the present application, the above-mentioned acquisition unit 501 includes: a replacement subunit, used to replace a second preset character in the customer evaluation text with a first preset character to obtain N documents, wherein the second preset character is used to connect the previous and next terms, and the first preset character is different from the second preset character; the second determination subunit includes: a calculation module, used to calculate the word frequency of each term in the term set of each document in the N documents; a first determination module, used to determine the term as a target root when the word frequency is greater than a preset value; and a second determination module, used to determine the term with the highest frequency among the synonyms of the term in the dictionary as the target root when the word frequency is less than or equal to a preset value.

[0119] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned establishment unit 502 includes: a first construction subunit, used to construct a first matrix, wherein the first matrix includes M rows and N columns; a calculation subunit, used to calculate the frequency of each stem keyword among the M stem keywords in the document for each document in the N documents, and obtain the frequency of each stem keyword in the N documents; a second construction subunit, used to construct each element of the first matrix according to the frequency of each stem keyword in the N documents, and obtain the target matrix.

[0120] Optionally, in the customer evaluation information analysis device provided in Example 2 of the present application, the above-mentioned acquisition unit 501 includes: a collection subunit, used to collect N evaluation texts of multiple customers, and translate the N evaluation texts to obtain N first texts; a conversion subunit, used to convert each first text of the N first texts into a preset data structure to obtain N second texts; a second processing subunit, used to perform data preprocessing operations on the N second texts respectively to obtain N documents, wherein the data preprocessing operations at least include: converting character format, deleting preset characters, and deleting semanticless characters.

[0121] It should be noted that the acquisition unit 501, the establishment unit 502, the clustering unit 503 and the first determination unit 504 correspond to steps S201 to S204 in Example 1, and the two modules and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102 analysis of customer evaluation information, ..., 102n), and the above-mentioned modules can also be run in the computer terminal 10 provided in Example 1 as part of the device.

[0122] Example 3

[0123] An embodiment of the present application may provide an electronic device, Figure 6 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 6 As shown, the electronic device may include: one or more ( Figure 6 (only one is shown) processor 602, memory 604, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0124] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0125] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: obtain N documents, extract M stem keywords from the N documents, wherein each of the N documents is a document processed from the customer evaluation text, and M and N are both integers greater than 1; establish a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each of the N documents; cluster the target matrix and the N documents to obtain clustering results; determine the analysis results of the N documents based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions.

[0126] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: clustering according to the target matrix and N documents to obtain clustering results, including: processing M stem keywords according to the target matrix, and updating the M stem keywords according to the processed stem keywords; hierarchical clustering of the M stem keywords to obtain L clusters, where L is a positive integer; determining random seeds based on the L clusters, and randomly selecting target documents from the N documents using the random seeds; clustering based on the distances between the stem keywords corresponding to the L clusters and the target documents to obtain clustering results.

[0127] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: process M stem keywords according to the target matrix, and update the M stem keywords according to the processed stem keywords, including: when the number of M stem keywords is greater than a preset number, filter the stem keywords with a word frequency higher than a preset value from the M stem keywords, and update the M stem keywords based on the filtered stem keywords; or when the number of M stem keywords is greater than a preset number, filter the M stem keywords using word frequency inverse document frequency, and update the M stem keywords based on the filtered stem keywords.

[0128] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: after establishing a target matrix based on M stem keywords and N documents, the above method also includes: determining the word frequency of each stem keyword in the M stem keywords based on the target matrix; constructing a target graph based on the word frequency of each stem keyword in the M stem keywords; determining high-frequency keywords based on the word frequency of each stem keyword in the target graph, and calculating the degree of association between high-frequency keywords through a preset function, wherein the high-frequency keyword is a stem keyword whose word frequency is higher than a preset value; and displaying the degree of association between the target graph and the high-frequency keyword to the target object.

[0129] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: extract M stem keywords from N documents, including: for each document in the N documents, determine the terms contained in the document to obtain a term set for each document, wherein each term in the term set of each document includes at least one word; determine the target root word corresponding to each term in the term set of each document based on the dictionary; determine the M stem keywords based on the target root word corresponding to each term in each document in the N documents.

[0130] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain N documents, including: use the first preset character to replace the second preset character in the customer evaluation text to obtain N documents, wherein the second preset character is used to connect the previous and next terms, and the first preset character is different from the second preset character; determine the target root word corresponding to each term in the term set of each document according to the dictionary, including: for each term in the term set of each document in the N documents, calculate the term frequency; when the term frequency is greater than a preset value, determine the term as the target root word; when the term frequency is less than or equal to the preset value, determine the term with the highest frequency among the synonyms of the term in the dictionary as the target root word.

[0131] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: establish a target matrix based on M stem keywords and N documents, including: construct a first matrix, wherein the first matrix includes M rows and N columns; for each document in the N documents, calculate the frequency of each stem keyword in the M stem keywords in the document to obtain the frequency of each stem keyword in the N documents; construct each element of the first matrix based on the frequency of each stem keyword in the N documents to obtain the target matrix.

[0132] The processor can call the information and application stored in the memory through the transmission device to execute the following steps: obtain N documents, including: collect N evaluation texts from multiple customers, and translate the N evaluation texts to obtain N first texts; convert each first text in the N first texts into a preset data structure to obtain N second texts; perform data preprocessing operations on the N second texts respectively to obtain N documents, wherein the data preprocessing operations at least include: converting character formats, deleting preset characters, and deleting semanticless characters.

[0133] By adopting the embodiment of the present application, a method for analyzing customer evaluation information is provided. By obtaining N documents, M stem keywords are extracted from the N documents, wherein each document in the N documents is a document processed from the customer evaluation text, and M and N are both integers greater than 1; a target matrix is ​​established based on the M stem keywords and the N documents, wherein the target matrix is ​​used to characterize the association relationship between the M stem keywords and each document in the N documents; clustering is performed based on the target matrix and the N documents to obtain clustering results; the analysis results of the N documents are determined based on the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestions, thereby solving the technical problem that the analysis results of customer comments are inaccurate due to inaccurate keyword extraction and unclear classification of customer feedback topics when analyzing customer comments.

[0134] By extracting M stem keywords from customer reviews and establishing a target matrix, the core content and theme of the review text can be determined, avoiding the changes in lexical form (such as different tenses of verbs) and inconsistent use of synonyms that may be caused by directly using the original vocabulary. The relationship between M stem keywords and N documents is clearly displayed, thereby improving the accuracy of customer review analysis. At the same time, the documents are divided into several groups through clustering algorithms, revealing the customer groups' satisfaction and suggestions for financial products, which helps companies understand the needs and pain points of different customer groups, thereby helping financial institutions to optimize products or services in a targeted manner and improve customer satisfaction and market competitiveness.

[0135] It can be understood by those skilled in the art that Figure 6 The structure shown is for illustration only, and the electronic device may also be a terminal device such as a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, etc. Figure 6 The structure of the electronic device is not limited. Figure 6 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 6 Different configurations are shown.

[0136] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0137] Example 4

[0138] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the method for analyzing customer evaluation information provided in the above embodiment 1.

[0139] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0140] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the program steps of the method for analyzing customer evaluation information.

[0141] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0142] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0144] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0146] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly or all or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc., which can store program code.

[0147] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for analyzing customer evaluation information, characterized in that: include: Obtain N documents, and extract M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and both M and N are integers greater than 1; Establishing a target matrix based on the M stem keywords and the N documents, wherein the target matrix is ​​used to represent the association relationship between the M stem keywords and each document in the N documents; Clustering is performed according to the target matrix and the N documents to obtain a clustering result; The analysis results of the N documents are determined according to the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestion.

2. The method according to claim 1, characterized in that: Clustering is performed based on the target matrix and the N documents to obtain a clustering result, including: Processing the M stem keywords according to the target matrix, and updating the M stem keywords according to the processed stem keywords; Performing hierarchical clustering on the M stem keywords to obtain L clusters, where L is a positive integer; Determine a random seed according to the L clusters, and randomly select a target document from the N documents using the random seed; Clustering is performed based on the distances between the L clusters and the stem keywords corresponding to the target document to obtain the clustering result.

3. The method according to claim 2, characterized in that Processing the M stem keywords according to the target matrix, and updating the M stem keywords according to the processed stem keywords, including: When the number of the M stem keywords is greater than a preset number, stem keywords with a frequency higher than a preset value are screened from the M stem keywords, and the M stem keywords are updated based on the screened stem keywords; or When the number of the M stem keywords is greater than a preset number, the M stem keywords are filtered using word frequency inverse document frequency, and the M stem keywords are updated based on the filtered stem keywords.

4. The method according to claim 1, characterized in that: After establishing a target matrix according to the M stem keywords and the N documents, the method further includes: Determining the word frequency of each stem keyword in the M stem keywords according to the target matrix; Constructing a target graph according to the word frequency of each stem keyword in the M stem keywords; Determine a high-frequency keyword according to the frequency of each stem keyword in the target graph, and calculate the degree of association between the high-frequency keywords through a preset function, wherein the high-frequency keyword is a stem keyword with a frequency higher than a preset value; The degree of association between the target graphic and the high-frequency keyword is displayed to the target object.

5. The method according to claim 1, characterized in that Extracting M stem keywords from the N documents, including: For each of the N documents, determine the terms contained in the document to obtain a term set for each document, wherein each term in the term set for each document includes at least one word; Determine the target root word corresponding to each term in the term set of each document according to the dictionary; The M stem keywords are determined according to the target root word corresponding to each word item in each document in the N documents.

6. The method according to claim 5, characterized in that Acquiring N documents, including: replacing a second preset character in the customer evaluation text with a first preset character to obtain the N documents, wherein the second preset character is used to connect preceding and following terms, and the first preset character is different from the second preset character; Determining the target root word corresponding to each term in the term set of each document according to the dictionary includes: calculating the term frequency of each term in the term set of each document in the N documents; when the term frequency is greater than a preset value, determining the term as the target root word; when the term frequency is less than or equal to the preset value, determining the term with the highest term frequency among the synonyms of the term in the dictionary as the target root word.

7. The method according to claim 1, characterized in that A target matrix is ​​established based on the M stem keywords and the N documents, including: Constructing a first matrix, wherein the first matrix includes M rows and N columns; For each document in the N documents, calculate the frequency of each stem keyword in the M stem keywords in the document to obtain the frequency of each stem keyword in the N documents; Each element of the first matrix is ​​constructed according to the frequency of each stem keyword in the N documents to obtain the target matrix.

8. The method according to claim 1, characterized in that Get N documents, including: Collecting N evaluation texts from multiple customers, and translating the N evaluation texts to obtain N first texts; Convert each of the N first texts into a preset data structure to obtain N second texts; Performing data preprocessing operations on the N second texts respectively to obtain the N documents, wherein the data preprocessing operations at least include: converting character formats, deleting preset characters, and deleting non-semantic characters.

9. A device for analyzing customer evaluation information, characterized in that: include: An acquisition unit, configured to acquire N documents and extract M stem keywords from the N documents, wherein each of the N documents is a document processed from a customer evaluation text, and both M and N are integers greater than 1; An establishing unit, configured to establish a target matrix according to the M stem keywords and the N documents, wherein the target matrix is ​​used to represent an association relationship between the M stem keywords and each document in the N documents; A clustering unit, used for performing clustering according to the target matrix and the N documents to obtain a clustering result; The first determining unit is used to determine the analysis results of the N documents according to the clustering results, wherein the analysis results include at least one of the following: satisfaction level, customer suggestion.

10. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 8 when running.

Images