Generation method and device of target report, program product and storage medium

Abstract

The embodiment of the invention provides a target report generation method and device, a program product and a storage medium, and the method comprises the steps: extracting a word vector in each first text in a first text set, and generating a first matrix, a second matrix and a third matrix by using the word vector of each first text; determining a target theme category of each first text by utilizing the first matrix, the second matrix and the third matrix; extracting a semantic relationship among a plurality of keywords included in each first text, and integrating the semantic relationship in each first text and the corresponding target topic category to obtain key information of the first text set; and generating a target report based on the second text set and the key information. The problem that the report cannot be efficiently generated in related technologies is solved, and the effect of efficiently generating the report is achieved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular, to a method and device for generating a target report, a program product, and a storage medium. Background Art

[0002] In the traditional bank report generation process, specialized analysts regularly collect data from various departments. Analysts analyze and judge the data based on their own experience and expertise to draw conclusions such as market trends and risk assessments, and then manually input, calculate and summarize data in spreadsheets or specific report software. The entire report generation process is highly dependent on manual labor, and the process of manual analysis and judgment is easily limited by personal level, resulting in low report generation efficiency and inability to generate reports efficiently. Summary of the invention

[0003] The embodiments of the present application provide a method and device for generating a target report, a program product, and a storage medium to at least solve the problem that reports cannot be generated efficiently in the related art.

[0004] According to an embodiment of the present application, a method for generating a target report is provided, comprising: extracting a word vector from each first text in a first text set, and generating a first matrix, a second matrix, and a third matrix using the word vector of each first text, wherein the plurality of first texts included in the first text set are financial information released by a first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category; determining the target subject category of each first text using the first matrix, the second matrix, and the third matrix; extracting the semantic relationship between the plurality of keywords included in each first text, and integrating the semantic relationship in each first text and the corresponding target subject category to obtain key information of the first text set; generating a target report based on the second text set and the key information, wherein the second text set includes transaction information generated by transactions between the first object and the second object within the historical time range, and the target report is used to indicate the transaction behavior of the first object within a future time range.

[0005] In an exemplary embodiment, before extracting the word vectors in each first text in the first text set and generating the first matrix, the second matrix and the third matrix using the word vectors of each of the first texts, the method further includes: obtaining an initial text set within the historical time range from a preset data channel, wherein the initial text set includes a first initial text set and a second initial text set; removing preset characters from the first initial text set and the second initial text set to obtain the first text set and the second text set.

[0006] In an exemplary embodiment, word vectors in each first text in a first text set are extracted, and a first matrix, a second matrix, and a third matrix are generated using the word vectors of each of the first texts, including: performing a word segmentation operation on the first texts to determine a plurality of word segments in the first texts; removing preset word segments from the plurality of word segments to obtain a plurality of keywords, and generating a keyword set of the first texts; performing a vector conversion operation on each of the keywords in the keyword set to obtain word vectors in the first texts; generating a first target matrix using the word vectors of each of the first texts, wherein one of the first texts corresponds to a row of elements in the target matrix, and the elements in the target matrix are used to represent the word vectors; and decomposing the first target matrix to obtain the first matrix, the second matrix, and the third matrix.

[0007] In an exemplary embodiment, the target topic category of each of the first texts is determined using the first matrix, the second matrix, and the third matrix, including: performing a normalization operation on the values ​​of each row of elements in the first matrix to obtain a fourth matrix, wherein the sum of the values ​​of each row of elements in the fourth matrix is ​​a preset probability value; calculating the product between the fourth matrix and a preset weight adjustment factor matrix to obtain a second target matrix, wherein the preset weight adjustment factor matrix is ​​used to adjust the values ​​of the elements in the fourth matrix, and the elements in the second target matrix are used to represent the association between the first text and the corresponding topic category in a preset scenario; determining N initial topic categories of each of the first texts based on the values ​​of the elements in the second target matrix, wherein the values ​​of the elements in the second target matrix corresponding to the initial topic categories are greater than a first preset threshold value, wherein N is a natural number greater than or equal to 1; and screening out the target topic category from the N initial topic categories based on the second matrix and the third matrix.

[0008] In an exemplary embodiment, the target topic category is selected from the N initial topic categories based on the second matrix and the third matrix, including: respectively determining the element values ​​corresponding to the N initial topic categories in the second matrix to obtain N first element values; respectively determining the element values ​​corresponding to the N initial topic categories in the third matrix to obtain N second element values; and determining the initial topic category corresponding to the element value greater than the second preset threshold value among the N first element values ​​and greater than the third preset threshold value among the N second element values ​​as the target topic category.

[0009] In an exemplary embodiment, a target report is generated based on the time series features in the second text set and the above-mentioned key information, including: extracting the time series features in the above-mentioned second text set, wherein the above-mentioned time series features are used to represent the feature changes of the above-mentioned transaction information within the above-mentioned historical time range; generating transaction trend information of the target subject category of the above-mentioned target text based on the above-mentioned time series features and the semantic relationship in the target text in the above-mentioned key information and the corresponding above-mentioned target subject category, wherein the above-mentioned target text is a text with the same target subject category in the above-mentioned first text, and the above-mentioned transaction trend information is used to represent the transaction features of the above-mentioned first object for the above-mentioned target subject category within the above-mentioned future time range; and generating the above-mentioned target report according to the above-mentioned transaction trend information.

[0010] According to an embodiment of the present application, a device for generating a target report is also provided, comprising: a first extraction module, used to extract word vectors from each first text in a first text set, and generate a first matrix, a second matrix and a third matrix using the word vectors of each first text, wherein the plurality of first texts included in the first text set are financial information released by a first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category; The first determination module is used to determine the target topic category of each of the above-mentioned first texts by using the above-mentioned first matrix, the above-mentioned second matrix and the above-mentioned third matrix; the second extraction module is used to extract the semantic relationship between the multiple keywords included in each of the above-mentioned first texts, and integrate the semantic relationship in each of the above-mentioned first texts and the corresponding above-mentioned target topic category to obtain the key information of the above-mentioned first text set; the first generation module is used to generate a target report based on the second text set and the above-mentioned key information, wherein the above-mentioned second text set includes the transaction information generated by the transactions between the above-mentioned first object and the second object within the above-mentioned historical time range, and the above-mentioned target report is used to indicate the transaction behavior of the above-mentioned first object within the future time range.

[0011] In an exemplary embodiment, the above-mentioned device also includes: a first acquisition module, used to extract the word vector in each first text in the first text set, and before using the word vector of each of the above-mentioned first texts to generate the first matrix, the second matrix and the third matrix, obtain the initial text set within the above-mentioned historical time range from a preset data channel, wherein the above-mentioned initial text set includes the first initial text set and the second initial text set; a first removal module, used to remove preset characters in the above-mentioned first initial text set and the above-mentioned second initial text set to obtain the above-mentioned first text set and the above-mentioned second text set.

[0012] In an exemplary embodiment, the first extraction module includes: a first determination submodule, used to perform a word segmentation operation on the first text, and determine multiple word segments in the first text; a first removal submodule, used to remove preset word segments from the multiple word segments to obtain multiple keywords, and generate a keyword set for the first text; a first execution submodule, used to perform a vector conversion operation on each of the keywords in the keyword set to obtain a word vector in the first text; a first generation submodule, used to generate a first target matrix using the word vector of each first text, wherein one first text corresponds to a row of elements in the target matrix, and the elements in the target matrix are used to represent the word vector; a first decomposition submodule, used to decompose the first target matrix to obtain the first matrix, the second matrix and the third matrix.

[0013] In an exemplary embodiment, the first determination module includes: a second execution submodule, which is used to perform a normalization operation on the values ​​of each row of elements in the first matrix to obtain a fourth matrix, wherein the sum of the values ​​of each row of elements in the fourth matrix is ​​a preset probability value; a first calculation submodule, which is used to calculate the product between the fourth matrix and a preset weight adjustment factor matrix to obtain a second target matrix, wherein the preset weight adjustment factor matrix is ​​used to adjust the values ​​of the elements in the fourth matrix, and the elements in the second target matrix are used to represent the association relationship between the first text and the corresponding subject category in a preset scenario; a second determination submodule, which is used to determine the N initial subject categories of each of the first texts according to the values ​​of the elements in the second target matrix, wherein the values ​​of the elements in the second target matrix corresponding to the initial subject categories are greater than a first preset threshold value, wherein N is a natural number greater than or equal to 1; a first screening submodule, which is used to screen out the target subject category from the N initial subject categories based on the second matrix and the third matrix.

[0014] In an exemplary embodiment, the above-mentioned first screening submodule includes: a first determination unit, used to respectively determine the element values ​​corresponding to the N above-mentioned initial topic categories in the above-mentioned second matrix, and obtain N first element values; a second determination unit, used to respectively determine the element values ​​corresponding to the N above-mentioned initial topic categories in the above-mentioned third matrix, and obtain N second element values; a third determination unit, used to determine the initial topic category corresponding to the element value greater than the second preset threshold among the N above-mentioned first element values ​​and greater than the third preset threshold among the N above-mentioned second element values ​​as the above-mentioned target topic category.

[0015] In an exemplary embodiment, the first generation module includes: a first extraction submodule, used to extract time series features from the second text set, wherein the time series features are used to represent feature changes of the transaction information within the historical time range; a second generation submodule, used to generate transaction trend information of the target subject category of the target text based on the time series features and the semantic relationship in the target text in the key information and the corresponding target subject category, wherein the target text is the text with the same target subject category in the first text, and the transaction trend information is used to represent the transaction features of the first object for the target subject category within the future time range; and a third generation submodule, used to generate the target report based on the transaction trend information.

[0016] According to another embodiment of the present application, a computer program product is provided, including a computer program, wherein the computer program is configured for a processor to execute the steps of any of the above method embodiments.

[0017] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to enable a processor to execute the steps in any of the above method embodiments.

[0018] According to another embodiment of the present application, an electronic device is also provided, including a memory, a processor, and a computer program stored on the above-mentioned memory and executable on the above-mentioned processor, wherein the above-mentioned processor is configured to execute the above-mentioned computer program to perform the steps in any one of the above-mentioned method embodiments.

[0019] Through this application, the word vectors in each first text in the first text set are used to generate a first matrix representing the association relationship between the first text and the corresponding subject category, a second matrix for representing the importance of the subject category, and a third matrix representing the association relationship between the keywords corresponding to the word vector and the subject category, and the first matrix, the second matrix and the third matrix are used to determine the target subject category of each first text. The semantic relationship between the multiple keywords included in the first text and the corresponding target subject category are determined as the key information of the first text set, and finally the target report is generated based on the second text set and the key information. Since the present application adopts a multi-matrix analysis method to quickly and accurately identify and classify text data and obtain the key information of the first text set, the problem of being unable to efficiently generate reports in the related art is solved, and the effect of efficiently generating reports is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a hardware structure block diagram of a mobile terminal of a method for generating a target report in an embodiment of the present application;

[0021] Figure 2 is a flow chart of a method for generating a target report according to an embodiment of the present application;

[0022] Figure 3 It is a flow chart of a method for generating a target report according to a specific embodiment of the present application;

[0023] Figure 4 It is a structural block diagram of a target report generation device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0026] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 1 is a hardware structure block diagram of a mobile terminal of a method for generating a target report in an embodiment of the present application. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1Only one is shown in the figure) a processor 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) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. 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 mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0027] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to a method for generating a target report in an embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. 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 mobile 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.

[0028] 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 mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as 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.

[0029] In this embodiment, a method for generating a target report is provided. Figure 2 is a flow chart of a method for generating a target report according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0030] Step S202, extracting the word vectors in each first text in the first text set, and using the word vectors of each first text to generate a first matrix, a second matrix and a third matrix, wherein the first texts included in the first text set are financial information released by the first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category;

[0031] Optionally, the first object includes but is not limited to financial institutions. The subject categories include but are not limited to monetary policy, stock market analysis, credit risk, macroeconomic trends, and industry analysis.

[0032] Optionally, each row of the first matrix corresponds to a first text, and each column corresponds to a subject category. The second matrix can be a diagonal matrix, and the values ​​of the elements on the diagonal represent the importance of the subject category. Each row of the third matrix corresponds to a keyword, and each column corresponds to a subject category.

[0033] Step S204, determining the target topic category of each of the first texts using the first matrix, the second matrix and the third matrix;

[0034] Step S206, extracting the semantic relationship between the multiple keywords included in each of the first texts, and integrating the semantic relationship in each of the first texts and the corresponding target subject category to obtain key information of the first text set;

[0035] Optionally, a first text includes one or more target topic categories, a first text includes multiple semantic relationships, the key information of the first text includes the target topic category of the first text and the semantic relationship of the first text, and the key information of the first text set includes the target topic category of each first text and the semantic relationship corresponding to the first text.

[0036] Step S208, generating a target report based on the second text set and the above key information, wherein the above second text set includes transaction information generated by transactions between the above first object and the second object within the above historical time range, and the above target report is used to indicate the transaction behavior of the above first object within a future time range.

[0037] Optionally, the second object may be a customer of the first object. Transaction information includes, but is not limited to, full-caliber deposit information, general deposit information, various loan information, and financial product information.

[0038] In this embodiment, the execution subject of the above steps can be a terminal, a server, a specific processor set in the terminal or server, or a processor or processing device set relatively independently from the terminal or server, but is not limited thereto.

[0039] Through the above steps, the word vectors in each first text in the first text set are used to generate a first matrix representing the association relationship between the first text and the corresponding subject category, a second matrix for representing the importance of the subject category, and a third matrix representing the association relationship between the keywords corresponding to the word vector and the subject category, and the first matrix, the second matrix and the third matrix are used to determine the target subject category of each first text. The semantic relationship between the multiple keywords included in the first text and the corresponding target subject category are determined as the key information of the first text set, and finally the target report is generated based on the second text set and the key information. Since the present application adopts a multi-matrix analysis method to quickly and accurately identify and classify text data and obtain the key information of the first text set, the problem of being unable to efficiently generate reports in the related art is solved, and the effect of efficiently generating reports is achieved.

[0040] In an exemplary embodiment, before extracting the word vectors in each first text in the first text set and generating the first matrix, the second matrix and the third matrix using the word vectors of each of the first texts, the method further includes: obtaining an initial text set within the historical time range from a preset data channel, wherein the initial text set includes a first initial text set and a second initial text set; removing preset characters from the first initial text set and the second initial text set to obtain the first text set and the second text set.

[0041] Optionally, preset data channels refer to those sources that have been determined and can provide the required data. Preset data channels include but are not limited to: the internal system or data lake of the first object, website news, announcements, and articles published by the first object. Preset characters are characters that do not carry semantic information or interfere with subsequent word vector extraction and topic analysis, including but not limited to punctuation marks, special characters, and tags. Data collection technologies for obtaining the initial text set include but are not limited to web crawlers, application program interface access, and database queries. Text preprocessing technologies for removing preset characters include but are not limited to regular expressions, natural language processing libraries, and string operations.

[0042] By removing preset characters, this embodiment can significantly reduce the noise in the text and reduce the dimension of the text, thereby achieving the purpose of improving data quality and reducing the dimension and computing cost.

[0043] In an exemplary embodiment, word vectors in each first text in a first text set are extracted, and a first matrix, a second matrix, and a third matrix are generated using the word vectors of each of the first texts, including: performing a word segmentation operation on the first texts to determine a plurality of word segments in the first texts; removing preset word segments from the plurality of word segments to obtain a plurality of keywords, and generating a keyword set of the first texts; performing a vector conversion operation on each of the keywords in the keyword set to obtain word vectors in the first texts; generating a first target matrix using the word vectors of each of the first texts, wherein one of the first texts corresponds to a row of elements in the target matrix, and the elements in the target matrix are used to represent the word vectors; and decomposing the first target matrix to obtain the first matrix, the second matrix, and the third matrix.

[0044] Optionally, the word segmentation operation is used to divide the continuous sequence of characters in the first text into independent words or phrases to obtain multiple word segments. The matrix decomposition technology for decomposing the first target matrix includes but is not limited to singular value decomposition, non-negative matrix decomposition, and principal component analysis. The preset word segments are words or phrases that appear frequently in the first text but do not carry specific information. The number of rows of the first target matrix is ​​equal to the number of first texts in the first text set, and the number of columns is equal to the number of word vectors.

[0045] Optionally, the first target matrix is ​​generated using the word vectors of each of the above-mentioned first texts, including: creating an empty matrix, wherein the number of rows of the empty matrix is ​​equal to the number of first texts in the first text set, and the number of columns is equal to the number of word vectors; for each first text, the following operations are performed: unifying the number of word vectors in each first text in the first text set, and filling the unified word vectors into corresponding rows of the empty matrix in a preset order to obtain the first target matrix.

[0046] This embodiment organizes the word vectors of each text into rows by constructing a first target matrix, which not only retains the semantic features of the original text, but also forms a structured data format, thereby achieving the purpose of efficiently representing the multi-dimensional features of the text.

[0047] In an exemplary embodiment, the target topic category of each of the first texts is determined using the first matrix, the second matrix, and the third matrix, including: performing a normalization operation on the values ​​of each row of elements in the first matrix to obtain a fourth matrix, wherein the sum of the values ​​of each row of elements in the fourth matrix is ​​a preset probability value; calculating the product between the fourth matrix and a preset weight adjustment factor matrix to obtain a second target matrix, wherein the preset weight adjustment factor matrix is ​​used to adjust the values ​​of the elements in the fourth matrix, and the elements in the second target matrix are used to represent the association between the first text and the corresponding topic category in a preset scenario; determining N initial topic categories of each of the first texts based on the values ​​of the elements in the second target matrix, wherein the values ​​of the elements in the second target matrix corresponding to the initial topic categories are greater than a first preset threshold value, wherein N is a natural number greater than or equal to 1; and screening out the target topic category from the N initial topic categories based on the second matrix and the third matrix.

[0048] Optionally, the preset weight adjustment factor matrix is ​​used to adjust the values ​​of the elements in the fourth matrix to increase the importance or weight of different subject categories in preset scenarios. The preset scenarios include but are not limited to financial scenarios, such as financial market trend analysis scenarios. A normalization operation is performed on the values ​​of each row of elements in the first matrix to ensure that the sum of the values ​​of each row of elements is a preset probability value (usually 1). This step converts the first matrix into a fourth matrix, in which each row can be regarded as a probability distribution.

[0049] Optionally, the larger the value of the element in the second target matrix, the stronger the ability of the corresponding topic category to explain the first text, and the more it can represent the topic of the first text. Based on the value of the element in the second target matrix, the topic category is preliminarily screened to obtain the initial topic category, which can improve the efficiency of determining the target topic category.

[0050] This embodiment uses a matrix operation method to determine the target topic category of each first text, which can provide accurate topic classification for text analysis and report generation, thereby achieving the purpose of automated and refined analysis of text data.

[0051] In an exemplary embodiment, the target topic category is selected from the N initial topic categories based on the second matrix and the third matrix, including: respectively determining the element values ​​corresponding to the N initial topic categories in the second matrix to obtain N first element values; respectively determining the element values ​​corresponding to the N initial topic categories in the third matrix to obtain N second element values; and determining the initial topic category corresponding to the element value greater than the second preset threshold value among the N first element values ​​and greater than the third preset threshold value among the N second element values ​​as the target topic category.

[0052] Optionally, for example, the initial topic categories of the first text include monetary policy, stock market analysis, credit risk, product recommendations, the "monetary policy" topic category has the highest weight value in the second matrix, and the weight value of this topic category in the third matrix is ​​also the highest, then the target topic category of the first text is "monetary policy".

[0053] This embodiment screens out initial topic categories whose element values ​​in the second matrix and the third matrix are greater than a preset threshold, thereby ensuring that only those topic categories that show high correlation in multiple dimensions will be selected as the final target topic categories, effectively reducing the impact of misjudgment and noise, and achieving the purpose of accurately determining the text topic.

[0054] In an exemplary embodiment, a target report is generated based on the time series features in the second text set and the above-mentioned key information, including: extracting the time series features in the above-mentioned second text set, wherein the above-mentioned time series features are used to represent the feature changes of the above-mentioned transaction information within the above-mentioned historical time range; generating transaction trend information of the target subject category of the above-mentioned target text based on the above-mentioned time series features and the semantic relationship in the target text in the above-mentioned key information and the corresponding above-mentioned target subject category, wherein the above-mentioned target text is a text with the same target subject category in the above-mentioned first text, and the above-mentioned transaction trend information is used to represent the transaction features of the above-mentioned first object for the above-mentioned target subject category within the above-mentioned future time range; and generating the above-mentioned target report according to the above-mentioned transaction trend information.

[0055] Optionally, one or more second texts related to the target subject category are screened from the second text set to obtain a target second text set, and time series features in the target second text set are extracted.

[0056] Optionally, after determining the target subject category, a second initial text set related to the target subject category is obtained, and preset characters in the second initial text set are removed to obtain a second text set.

[0057] Optionally, the time series features include but are not limited to moving average, exponential weighted average, cyclical changes, and seasonal patterns. The transaction trend information includes but is not limited to transaction volume trend information, product sales trend information, and deposit trend information.

[0058] Optionally, taking the analysis of sales data (second text set) of a series of financial products (such as a certain insurance or investment product) and related market news articles (first text set) as an example, the following time series features are extracted from the historical sales data of financial products, including daily or monthly sales volume, sales amount, customer feedback, etc.: Moving average: short-term (such as 7 days) and long-term (such as 30 days) moving average of sales volume; Seasonal pattern: seasonal changes in sales data, such as holiday effects or quarterly patterns. The semantic relationships in Articles 1, 2, and 6 with the target topic category of transaction volume include but are not limited to "new product launch", "market reaction to an event", and "competitor's strategy change". Using the trained model, the transaction trend information "sales volume of financial product A is expected to increase in the next month" is obtained based on the above information.

[0059] Optionally, the target subject category of the target text and the transaction trend information corresponding to the target text are input into a preset report template to obtain a target report.

[0060] The present invention will be described below in conjunction with specific embodiments:

[0061] Figure 3 This is a flowchart of a method for generating a target report according to a specific embodiment of the present application, comprising the following steps:

[0062] S302, using a crawler to obtain multiple financial articles published by the bank in the past six months through web pages to obtain a first initial text set. Removing preset characters from the first initial text set to obtain a first text set.

[0063] S304, perform text preprocessing on the first text in the first text set, supplement the missing text data, and perform format conversion to obtain consistent text data. Then perform a word segmentation operation on the first text, determine multiple word segments in the first text, remove useless word segments from the multiple word segments to obtain a keyword set of the first text, perform a vector conversion operation on each keyword in the keyword set, and obtain a word vector in the first text. Create an empty matrix, where the number of rows of the empty matrix is ​​equal to the number of first texts in the first text set, and the number of columns is equal to the number of word vectors; for each first text, perform the following operations: unify the number of word vectors in each first text in the first text set, and fill the unified word vectors into the corresponding rows of the empty matrix in a preset order to obtain a first target matrix of the number of first texts*the number of word vectors. The first target matrix is ​​decomposed by the latent semantic analysis model, and the first text set is represented as a linear combination of topic categories and topic category weights: the first target matrix is ​​decomposed into a first matrix, a second matrix and a third matrix by singular value decomposition, wherein the elements in the first matrix are used to represent the association relationship between the first text and the corresponding topic category, the elements in the second matrix are used to represent the importance of the topic category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the topic category. The values ​​of each row of elements in the first matrix are normalized, and the topic weights in the first matrix are readjusted in combination with the preset weight adjustment factor matrix to obtain the second target matrix. The topic categories corresponding to the weights in the second target matrix that are greater than the preset threshold are used to determine the N initial topic categories of each first text, and the initial topic categories whose element values ​​in the second matrix and the third matrix are both greater than the preset threshold are determined as the target topic categories. For example, the target topic categories of the first text 1, the first text 2, and the first text 3 in the first text set are general deposits, and the target topic categories of the first text 4, the first text 5, and the first text 6 are trading volume. Then, entity relations are extracted from the first texts, and the semantic relations and corresponding target topic categories in each first text are integrated to obtain key information of the first text set.

[0064] S306, after determining the target subject category, obtain a second initial text set related to the target subject category, remove the preset characters in the second initial text set, and obtain the second text set, for example, use a crawler to obtain the non-real-time financial data on general deposits and transaction volumes released by banks in the past six months through web pages, and access the corresponding service database or access the project data service to obtain quasi-real-time financial data on general deposits and transaction volumes, and obtain the second initial text set. Remove the preset characters in the second initial text set to obtain the second text set.

[0065] The second text in the second text set is preprocessed to supplement the missing text data, and the format is converted to obtain consistent text data. The text data of general deposits and transaction volume are respectively extracted to obtain time series features. For example, time series features are extracted from deposit data: moving average: short-term (e.g., 7 days) and long-term (e.g., 30 days) moving averages of deposit balances; seasonal patterns: identify seasonal changes in deposit data, such as holiday effects or quarterly patterns. The semantic relationships in the first text 1, the first text 2, and the first text 3 with the target topic category of general deposits include but are not limited to "interest rate hikes", "economic policy changes", "enhanced market confidence", etc. Using the trained model, based on the above information, the transaction trend information about general deposits is obtained, "the general deposit balance is expected to increase in the next month."

[0066] S308, generating a target report based on the transaction trend information, the contents of the report including but not limited to the target subject category and its corresponding transaction trend information. After the target report is generated, it is sent by calling a simple mail transfer protocol server.

[0067] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0068] In this embodiment, a device for generating a target report is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0069] Figure 4 is a structural block diagram of a target report generation device according to an embodiment of the present application, such as Figure 4 As shown, the device comprises:

[0070] A first extraction module 402 is used to extract the word vector of each first text in the first text set, and generate a first matrix, a second matrix and a third matrix using the word vector of each first text, wherein the plurality of first texts included in the first text set are financial information released by the first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category;

[0071] A first determination module 404 is used to determine a target topic category of each of the first texts using the first matrix, the second matrix and the third matrix;

[0072] A second extraction module 406 is used to extract the semantic relationship between the multiple keywords included in each of the first texts, and integrate the semantic relationship in each of the first texts and the corresponding target topic category to obtain key information of the first text set;

[0073] The first generating module 408 is used to generate a target report based on the second text set and the above-mentioned key information, wherein the above-mentioned second text set includes transaction information generated by transactions between the above-mentioned first object and the second object within the above-mentioned historical time range, and the above-mentioned target report is used to indicate the transaction behavior of the above-mentioned first object within the future time range.

[0074] In an exemplary embodiment, the above-mentioned device also includes: a first acquisition module, used to extract the word vector in each first text in the first text set, and before using the word vector of each of the above-mentioned first texts to generate the first matrix, the second matrix and the third matrix, obtain the initial text set within the above-mentioned historical time range from a preset data channel, wherein the above-mentioned initial text set includes the first initial text set and the second initial text set; a first removal module, used to remove preset characters in the above-mentioned first initial text set and the above-mentioned second initial text set to obtain the above-mentioned first text set and the above-mentioned second text set.

[0075] In an exemplary embodiment, the first extraction module includes: a first determination submodule, used to perform a word segmentation operation on the first text, and determine multiple word segments in the first text; a first removal submodule, used to remove preset word segments from the multiple word segments to obtain multiple keywords, and generate a keyword set for the first text; a first execution submodule, used to perform a vector conversion operation on each of the keywords in the keyword set to obtain a word vector in the first text; a first generation submodule, used to generate a first target matrix using the word vector of each first text, wherein one first text corresponds to a row of elements in the target matrix, and the elements in the target matrix are used to represent the word vector; a first decomposition submodule, used to decompose the first target matrix to obtain the first matrix, the second matrix and the third matrix.

[0076] In an exemplary embodiment, the first determination module includes: a second execution submodule, which is used to perform a normalization operation on the values ​​of each row of elements in the first matrix to obtain a fourth matrix, wherein the sum of the values ​​of each row of elements in the fourth matrix is ​​a preset probability value; a first calculation submodule, which is used to calculate the product between the fourth matrix and a preset weight adjustment factor matrix to obtain a second target matrix, wherein the preset weight adjustment factor matrix is ​​used to adjust the values ​​of the elements in the fourth matrix, and the elements in the second target matrix are used to represent the association relationship between the first text and the corresponding subject category in a preset scenario; a second determination submodule, which is used to determine the N initial subject categories of each of the first texts according to the values ​​of the elements in the second target matrix, wherein the values ​​of the elements in the second target matrix corresponding to the initial subject categories are greater than a first preset threshold value, wherein N is a natural number greater than or equal to 1; a first screening submodule, which is used to screen out the target subject category from the N initial subject categories based on the second matrix and the third matrix.

[0077] In an exemplary embodiment, the above-mentioned first screening submodule includes: a first determination unit, used to respectively determine the element values ​​corresponding to the N above-mentioned initial topic categories in the above-mentioned second matrix, and obtain N first element values; a second determination unit, used to respectively determine the element values ​​corresponding to the N above-mentioned initial topic categories in the above-mentioned third matrix, and obtain N second element values; a third determination unit, used to determine the initial topic category corresponding to the element value greater than the second preset threshold among the N above-mentioned first element values ​​and greater than the third preset threshold among the N above-mentioned second element values ​​as the above-mentioned target topic category.

[0078] In an exemplary embodiment, the first generation module includes: a first extraction submodule, used to extract time series features from the second text set, wherein the time series features are used to represent feature changes of the transaction information within the historical time range; a second generation submodule, used to generate transaction trend information of the target subject category of the target text based on the time series features and the semantic relationship in the target text in the key information and the corresponding target subject category, wherein the target text is the text with the same target subject category in the first text, and the transaction trend information is used to represent the transaction features of the first object for the target subject category within the future time range; and a third generation submodule, used to generate the target report based on the transaction trend information.

[0079] An embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0080] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0081] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0082] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0083] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0084] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0085] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0086] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for generating a target report, characterized in that: include: Extracting the word vector of each first text in the first text set, and using the word vector of each first text to generate a first matrix, a second matrix and a third matrix, wherein the first texts included in the first text set are financial information released by the first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category; Determine a target topic category for each of the first texts using the first matrix, the second matrix, and the third matrix; Extracting semantic relations between a plurality of keywords included in each of the first texts, and integrating the semantic relations in each of the first texts and the corresponding target subject category to obtain key information of the first text set; A target report is generated based on the second text set and the key information, wherein the second text set includes transaction information generated by transactions between the first object and the second object within the historical time range, and the target report is used to indicate the transaction behavior of the first object within a future time range.

2. The method according to claim 1, characterized in that Before extracting the word vectors from each first text in the first text set and generating the first matrix, the second matrix and the third matrix using the word vectors of each first text, the method further includes: Acquire an initial text set within the historical time range from a preset data channel, wherein the initial text set includes a first initial text set and a second initial text set; The preset characters in the first initial text set and the second initial text set are removed to obtain the first text set and the second text set.

3. The method according to claim 1, characterized in that Extracting a word vector from each first text in the first text set, and using the word vector of each first text to generate a first matrix, a second matrix, and a third matrix, including: Performing a word segmentation operation on the first text to determine a plurality of word segments in the first text; Removing preset segmented words from the plurality of segmented words to obtain the plurality of keywords, and generating a keyword set of the first text; Performing a vector conversion operation on each of the keywords in the keyword set to obtain a word vector in the first text; Generate a first target matrix using the word vector of each of the first texts, wherein one of the first texts corresponds to a row of elements in the target matrix, and the elements in the target matrix are used to represent the word vector; The first target matrix is ​​decomposed to obtain the first matrix, the second matrix and the third matrix.

4. The method according to claim 1, characterized in that Determining a target topic category of each of the first texts using the first matrix, the second matrix, and the third matrix includes: Performing a normalization operation on the values ​​of each row of elements in the first matrix to obtain a fourth matrix, wherein the sum of the values ​​of each row of elements in the fourth matrix is ​​a preset probability value; Calculate the product between the fourth matrix and a preset weight adjustment factor matrix to obtain a second target matrix, wherein the preset weight adjustment factor matrix is ​​used to adjust the values ​​of elements in the fourth matrix, and the elements in the second target matrix are used to represent the association relationship between the first text and the corresponding subject category in a preset scenario; Determine N initial topic categories of each of the first texts according to the values ​​of the elements in the second target matrix, wherein the values ​​of the elements in the second target matrix corresponding to the initial topic categories are greater than a first preset threshold, wherein N is a natural number greater than or equal to 1; The target topic category is screened out from the N initial topic categories based on the second matrix and the third matrix.

5. The method according to claim 4, characterized in that Screening out the target topic category from the N initial topic categories based on the second matrix and the third matrix includes: Respectively determine the element values ​​corresponding to the N initial topic categories in the second matrix to obtain N first element values; Respectively determine the element values ​​corresponding to the N initial topic categories in the third matrix to obtain N second element values; The initial topic category corresponding to the element value greater than the second preset threshold among the N first element values ​​and greater than the third preset threshold among the N second element values ​​is determined as the target topic category.

6. The method according to claim 1, characterized in that Based on the second text set and the key information, a target report is generated, including: Extracting time series features from the second text set, wherein the time series features are used to represent feature changes of the transaction information within the historical time range; Based on the time series feature and the semantic relationship in the target text in the key information and the corresponding target subject category, generating transaction trend information of the target subject category of the target text, wherein the target text is the text of the same target subject category in the first text, and the transaction trend information is used to represent the transaction characteristics of the first object for the target subject category within the future time range; The target report is generated according to the transaction trend information.

7. A device for generating a target report, characterized in that: include: A first extraction module is used to extract the word vector of each first text in the first text set, and generate a first matrix, a second matrix and a third matrix using the word vector of each first text, wherein the first texts included in the first text set are financial information released by the first object within a historical time range, the elements in the first matrix are used to represent the association relationship between the first text and the corresponding subject category, the elements in the second matrix are used to represent the importance of the subject category, and the elements in the third matrix are used to represent the association relationship between the keywords corresponding to the word vector and the subject category; A first determination module, configured to determine a target topic category of each of the first texts using the first matrix, the second matrix, and the third matrix; A second extraction module is used to extract the semantic relationship between the multiple keywords included in each of the first texts, and integrate the semantic relationship in each of the first texts and the corresponding target subject category to obtain key information of the first text set; The first generating module is used to generate a target report based on the second text set and the key information, wherein the second text set includes transaction information generated by transactions between the first object and the second object within the historical time range, and the target report is used to indicate the transaction behavior of the first object within a future time range.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the method described in any one of claims 1 to 6 when executed by a processor.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 6 are implemented.

Images