Object rating method, device, equipment and storage medium

Abstract

The present application discloses an object rating method, device, equipment and storage medium. The method includes constructing a rating method set based on existing rating reports, and determining target scoring items of an object to be rated according to the rating method set; determining a first score of a first target scoring item in a first quantization category in the target scoring items based on a language model; determining a second score of a second target scoring item in a second quantization category in the target scoring items through a quantization method; and determining a rating level of the object to be rated according to the first score and the second score. With the rating method set as prior knowledge, the present application determines the scores of non-quantifiable scoring items through a large model, and calculates the scores of quantifiable scoring items through the quantization data in the rating method set. This can not only reduce the labor cost and time cost required for manual rating, but also improve the accuracy of enterprise object rating and the rating consistency with rated enterprise objects.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly relates to an object rating method, device, equipment, and storage medium. Background Art

[0002] With the continuous development of fintech, especially Internet technology finance, more and more technologies are applied in the financial field. However, the financial industry also poses higher requirements for technology. For example, the financial industry has higher requirements for enterprise rating. However, in the existing technology, it is necessary for financial practitioners to rate enterprises. This method not only has the problem of strong subjectivity, but also requires rating personnel to spend a lot of time screening and analyzing resources related to enterprise operations, affecting the efficiency and accuracy of enterprise rating.

[0003] Therefore, the existing technology still needs to be improved. Summary of the Invention

[0004] The technical problem to be solved by this application is to provide an object rating method, device, equipment, and storage medium in view of the deficiencies of the existing technology.

[0005] To solve the above technical problem, the first aspect of this application provides an object rating method, where the object rating method specifically includes:

[0006] Construct a rating method set based on existing rating reports. The rating method set includes several rating methods and several scoring items corresponding to each rating method. The scoring items include multiple scoring data for the target rating method;

[0007] Determine the target rating method corresponding to the object to be rated and the target scoring item corresponding to the target rating method according to the rating method set;

[0008] Obtain the first target scoring item of the first quantization category in the target scoring item, determine the first target scoring data of the first target scoring item, and score the object to be rated based on the first target scoring data and a preset large language model to determine the first score;

[0009] Obtain the second target scoring item of the second quantization category in the target scoring item, determine the second target scoring data of the second target scoring item, and score the object to be rated based on the second target scoring data to determine the second score;

[0010] Determine the scoring grade of the object to be rated according to the first score and the second score.

[0011] In the object rating method, the construction of the rating method set based on existing rating reports specifically includes:

[0012] Obtain a number of existing rating reports;

[0013] Respectively obtain the detection targets of each existing rating report and the detection data corresponding to the detection targets. Among them, the detection targets include titles, natural paragraphs, and tables, and the detection data includes category information, location information, and text content;

[0014] Based on the detection data of all detection targets of each existing rating report, determine the rating method adopted by each existing rating report, several scoring items corresponding to the rating method, and multiple scoring data of the scoring items;

[0015] Construct a rating method set according to the determined rating methods, several scoring items corresponding to each rating method, and multiple scoring data of the scoring items.

[0016] The object rating method described above, wherein the step of respectively obtaining the detection targets of each existing rating report and the detection data corresponding to the detection targets specifically includes:

[0017] Input each existing rating report into a trained text layout detection model, and output the detection targets in each existing rating report and the category information and location information of the detection targets through the text layout detection model;

[0018] Extract the text content of each detection target from the corresponding existing rating report based on the location information of each detection target;

[0019] Use the category information, location information, and text content as the detection data of the detection target to obtain the detection targets of each existing rating report and the detection data corresponding to the detection targets.

[0020] The object rating method described above, wherein the trained text layout detection model specifically includes:

[0021] Obtain a text layout data set, where the text layout data set includes a number of training texts and annotation target information corresponding to the training texts;

[0022] Use the training texts in the text layout data to train a preset object detection model to obtain a trained text layout detection model.

[0023] The object rating method described above, wherein the step of determining the rating method adopted by each existing rating report, several scoring items corresponding to the rating method, and multiple scoring data of the scoring items based on the detection data of all detection targets of each existing rating report specifically includes:

[0024] Based on the detection data of the natural paragraphs, determine the rating method adopted by the existing rating report;

[0025] Based on the detection data of the detection target, determine several scoring items corresponding to the rating method and multiple scoring data of the scoring items.

[0026] The object rating method described above, wherein the scoring data includes the scoring item name, the scoring value, the scoring basis, and the quantification data. Based on the detection data of the detection target, determining several scoring items corresponding to the rating method and the scoring data of the scoring items specifically includes:

[0027] Based on the detection data of the table, determine several scoring items corresponding to the rating method, the scoring item names, and the scoring values of each scoring item;

[0028] Based on the detection data of the title and the natural paragraph, determine the scoring basis of each scoring item;

[0029] Obtain the rating quantification method corresponding to each scoring item, and determine the quantification data of each scoring item based on the rating quantification method through a preset large language model.

[0030] The object rating method described above, wherein the determining the scoring basis of each scoring item based on the detection data of the title and the natural paragraph specifically includes:

[0031] Determine the target title corresponding to each scoring item, and obtain the target natural paragraph corresponding to the target title;

[0032] Based on the text content in the detection data of the target title and the text content in the detection data of the target natural paragraph, determine the scoring basis of the scoring item name.

[0033] The object rating method described above, wherein the determining the target title corresponding to each scoring item and obtaining the target natural paragraph corresponding to the target title specifically includes:

[0034] Obtain the structured table of contents of the existing rating report, where the structured table of contents consists of multi-level headings and the content included in each level of headings;

[0035] According to the scoring item names of each scoring item and the structured table of contents, determine the target title corresponding to each scoring item;

[0036] According to the report content included in the target title, determine the target natural paragraph corresponding to the target title.

[0037] The object rating method described above, wherein the construction process of the structured table of contents specifically includes:

[0038] Determine the title levels of each title in the existing rating report, and determine the subordinate relationship between titles and the content included in each title based on the title levels of each title and the position information of each title;

[0039] Construct a structured table of contents for the existing rating report based on the subordinate relationship between titles and the content included in each title.

[0040] The object rating method described above, wherein the determination of the title levels of each title in the existing rating report specifically includes:

[0041] Obtain the title font sizes and title numbers of each title;

[0042] Determine the title levels of each title based on the obtained title font sizes and title numbers of each title.

[0043] The object rating method described above, wherein the determination process of the content included in each title specifically includes:

[0044] Based on the position information of each detection target, sort each detection target in the order of appearance in the existing rating report to obtain a detection target sequence;

[0045] Take the detection targets and detection data between two adjacent same-level titles in the detection target sequence as the content included in the title with a higher order.

[0046] The object rating method described above, wherein the determination of the first score by scoring the object to be rated based on the first target score data and a preset large language model specifically includes:

[0047] Obtain a number of keywords corresponding to the first target scoring item;

[0048] Based on the object name of the object to be rated and the number of keywords, determine a summary text regarding the number of keywords;

[0049] Based on the summary text, select the first target score data for the first target scoring item in the rating method set;

[0050] Score the object to be rated through a preset large language model based on the summary text and the first target score data to determine the first score.

[0051] The object rating method described above, wherein the determination of the first score by scoring the object to be rated based on the summary text and the first target score data through a preset large language model specifically includes:

[0052] Construct a scoring prompt word for the first target scoring item;

[0053] Concatenate the scoring prompt words, the first target scoring item, the first target scoring data of the first target scoring item, and the summary text, and input them into a preset large language model, and output the first score of the first target scoring item through the preset large language model.

[0054] The object rating method described above, wherein, the specific steps of obtaining several keywords corresponding to the first target scoring item include:

[0055] Select all scoring data corresponding to the first target scoring item from the rating method set, and segment the scoring basis in the first target scoring data to obtain several words;

[0056] Calculate the similarity between each word and the name of the scoring item of the first target scoring item and the term frequency-inverse document frequency of each word respectively;

[0057] Based on the similarity and the term frequency-inverse document frequency of each word, determine the importance of each word;

[0058] Select several keywords from several words based on the importance of each word to obtain several keywords corresponding to the first target scoring item.

[0059] The object rating method described above, wherein, the specific steps of determining the summary text about several keywords based on the object name of the object to be rated and the several keywords include:

[0060] Search the text resources of the object to be rated based on the object name of the object to be rated and several keywords;

[0061] Divide the searched text resources into several natural paragraphs, and select a preset number of representative natural paragraphs from the several natural paragraphs based on the object name and several keywords;

[0062] Based on the preset number of representative natural paragraphs and the several keywords, determine the summary text about the several keywords through a preset large language model.

[0063] The object rating method described above, wherein, the specific steps of rating the object to be rated based on the second target scoring data and determining the second score include:

[0064] Obtain the quantization data in the second target scoring data, wherein the quantization data includes the code function corresponding to the rating quantization method and the quantization index required by the rating quantization method;

[0065] Obtain the quantization index data corresponding to the quantization index;

[0066] Use the quantization index data as the input of the code function to obtain the second score of the second target scoring item.

[0067] The object rating method described above, wherein determining the rating level of the object to be rated according to the first score and the second score specifically includes:

[0068] Weight the first score and the second score to obtain a final score;

[0069] According to the corresponding relationship between the preset rating levels and scores, determine the rating level corresponding to the final score to obtain the rating level of the object to be rated.

[0070] The second aspect of the present application provides an object rating device, which specifically includes:

[0071] A construction module for constructing a set of rating methods based on existing rating reports. The set of rating methods includes several rating methods and several scoring items corresponding to each rating method. The scoring items include multiple scoring data for the target rating method;

[0072] A rating method determination module for determining the target rating method corresponding to the object to be rated and the target scoring item corresponding to the target rating method according to the set of rating methods;

[0073] A first processing module for obtaining the first target scoring item of the first quantization category in the target scoring item, determining the first target scoring data of the first target scoring item, and scoring the object to be rated based on the first target scoring data and a preset large language model to determine the first score;

[0074] A second processing module for obtaining the second target scoring item of the second quantization category in the target scoring item, determining the second target scoring data of the second target scoring item, and scoring the object to be rated based on the second target scoring data to determine the second score;

[0075] A level determination module for determining the rating level of the object to be rated according to the first score and the second score.

[0076] The third aspect of the present application provides a computer-readable storage medium storing one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in any of the above object rating methods.

[0077] The fourth aspect of the present application provides a terminal device, which includes: a processor and a memory;

[0078] The memory stores a computer-readable program executable by the processor;

[0079] When the processor executes the computer-readable program, it implements the steps in any of the above-described object rating methods.

[0080] Beneficial effects:

[0081] 1. This application uses the object rating method extracted from the evaluation reports of existing objects (such as enterprises) as prior knowledge, and determines the score level of the object to be rated (such as an enterprise) through a large language model, reducing the dependence on the professional capabilities of rating personnel.

[0082] 2. This application screens and integrates the relevant materials of the object to be evaluated and rated (such as an enterprise) based on the scoring items to determine the key resource information for rating, which can reduce the time spent on resource screening and analysis related to the object's operation, and improve the efficiency and accuracy of object rating.

[0083] 3. This application selects the target scoring data of objects with the same operating status as the object to be rated from all scoring data using the summary text reflecting the operating status, and uses these target scoring data as the prior knowledge of the large model, which can ensure the consistency of the scoring results of objects with the same operating status and improve the rating accuracy of the object to be rated. Description of the drawings

[0084] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0085] Figure 1 It is a flowchart of the object rating method provided by the embodiment of the present application.

[0086] Figure 2 It is a schematic flowchart of the construction process of the rating method set.

[0087] Figure 3 It is a schematic flowchart of an implementation manner of step S30.

[0088] Figure 4 It is a schematic flowchart of an implementation manner of step S40.

[0089] Figure 5 It is a principle block diagram of the object rating device provided by the embodiment of the present application.

[0090] Figure 6 It is a principle block diagram of the terminal device provided by the embodiment of the present application. Detailed implementation manners

[0091] The embodiments of the present application provide an object rating method, device, equipment, and storage medium. To make the purpose, technical solution, and effects of the present application clearer and more explicit, the following further elaborates on the present application with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0092] Those skilled in the art of the present technology can understand that unless specifically stated otherwise, the singular forms "a", "an", "the", and "said" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of the present application means the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or their groups. It should be understood that when we say an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The phrase "and / or" used herein includes all or any unit and all combinations of one or more related listed items.

[0093] Those skilled in the art of the present technology can understand that unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those of ordinary skill in the art to which the present application belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as here.

[0094] It should be understood that the sequence numbers and magnitudes of the steps in this embodiment do not imply the order of execution. The order of execution of each process is determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0095] The following further illustrates the application content by describing the embodiments in conjunction with the accompanying drawings.

[0096] This embodiment provides an object rating method, as Figure 1 shown, the method includes:

[0097] S10. Construct a rating method set based on existing rating reports.

[0098] Specifically, the rating method set is used as a prior knowledge database for object rating methods, and a target rating method corresponding to the object to be rated can be selected from the rating method set. The rating method set includes several rating methods, each rating method includes several rating items, each rating item includes multiple rating data, and the several rating items included in the rating method are different. Different rating methods may include the same rating items. For example, the rating method set includes rating method A and rating method B. Rating method A includes rating item A1, rating item A2, and rating item A3. Rating method B includes rating item B1, rating item B2, and rating item B3. Rating item A1, rating item A2, and rating item A3 are different from each other. Rating item B1, rating item B2, and rating item B3 are different from each other. However, rating item A1 and rating item B2 are the same.

[0099] Furthermore, the multiple rating data may include a rating item name, a rating value, a rating basis, and quantification data. Among them, the rating item name is the unique identifier of the rating item, the rating value is the rating score of the rating item, the rating basis is the text explanation for the rating item to obtain its corresponding rating score, the quantification data includes the quantification category of the rating item, and when the quantification category is the first quantification category, the quantification data further includes the code function of the quantification method corresponding to the rating item and the required quantification indicators. That is to say, each evaluation rating method can rate an object based on multiple rating factor rating items (i.e., rating item names), and each rating item will have a specific rating score (i.e., rating value), the text explanation of the rating score (i.e., rating basis), as well as the quantification category of the rating item and the corresponding quantification method (i.e., quantification data).

[0100] In a specific implementation, as Figure 2 shown, the construction of the rating method set based on existing rating reports specifically includes:

[0101] S11. Obtain several existing rating reports;

[0102] S12. Respectively obtain the detection target of each existing rating report and the detection data corresponding to the detection target. Among them, the detection target includes a title, a natural paragraph, and a table, and the detection data includes category information, location information, and text content;

[0103] S13. Determine the rating method adopted by each existing rating report, the several rating items corresponding to the rating method, and the multiple rating data of the rating items based on the detection data of all detection targets of each existing rating report;

[0104] S14. Construct a rating method set according to the determined rating methods, the several rating items corresponding to each rating method, and the multiple rating data of the rating items.

[0105] Specifically, in step S11, the existing rating reports are reports formed by rating existing objects. For example, reports formed by evaluating and rating enterprise objects, etc. Among them, the rated objects corresponding to each existing rating report may be different, and some of the rated objects may belong to the same industry. At the same time, some of the existing rating reports among several existing rating reports may adopt the same rating method, and some of the existing rating reports may adopt different rating methods. Among them, the existing rating reports can be in pdf format, or in word format, etc.

[0106] In addition, after obtaining several existing rating reports, the industry to which the object corresponding to each existing rating report belongs can be obtained, and this industry can be associated with the rating method adopted by this existing rating report, so as to configure an industry category for the evaluation rating method, so as to facilitate subsequent searching for the target rating method corresponding to the object to be rated among the existing rating reports when constructing a rating method set. Of course, in practical applications, other category labels can also be configured for the rating methods adopted by the existing rating reports. For example, the operating status category can be configured for the rating methods adopted by the existing rating reports according to the operating status of the objects corresponding to the existing rating reports, or the industry category and operating status category can be configured for the objects corresponding to the existing rating reports. Then, when searching for the target rating method corresponding to the object to be rated, the search can be first performed according to the industry category, and then the search can be performed according to the operating status category among the search results, etc.

[0107] Further, in step S12, the detection targets include titles, natural paragraphs, and tables, and the detection data includes category information, position information, and text content. The category information is used to reflect that the detection target is a title, a natural paragraph, or a table, the position information is used to reflect the text position of the detection target in the existing rating report, and the text content is used to reflect the text content included in the detection target. Specifically, the position information may include the upper left coordinate and the lower right coordinate of the detection target, and since the existing rating report may include multiple pages of text, the position information may also include the page number used to clarify the text page where the detection target is located. Of course, the position coordinates can also be represented in other ways. For example, the position coordinates can be the center point coordinates, the width of the detection target, and the height of the detection target, etc. The text content is the text content included in the detection target, that is, the text content may include literal content. In addition, since in practical applications, titles and the main text generally use different font sizes, the text content may also include the font size, so as to facilitate the subsequent construction of the table of contents structure.

[0108] In one implementation manner, the step of respectively obtaining the detection targets of each existing rating report and the detection data corresponding to the detection targets specifically includes:

[0109] S121. Input each of the existing rating reports into a trained text layout detection model, and output the detection targets in each existing rating report, as well as the category information and location information of the detection targets through the text layout detection model;

[0110] S122. Extract the text content of each detection target from the corresponding existing rating report based on the location information of each detection target;

[0111] S123. Use the category information, the location information, and the text content as the detection data of the detection target to obtain the detection targets of each existing rating report and the detection data corresponding to the detection targets.

[0112] Specifically, the text layout detection model is a trained neural network model. Through the text layout detection model, each detection target in the existing rating report and the location information of each detection target can be detected. That is to say, the input item of the text layout detection model is the existing rating report, and the output item is the detection target information. The detection target information includes the category information and the location information of the detection target. The category information includes titles, natural paragraphs, and tables. The location information includes page numbers, upper left coordinates, and lower right coordinates. That is, the detection target information can be expressed as (page_num, class, x1, y1, x2, y2), where page_num represents the page number, class represents the category information of the detection target, (x1, y1) represents the upper left coordinates of the detection target, and (x2, y2) represents the lower right coordinates of the detection target.

[0113] Exemplarily, the training process of the text layout detection model can be as follows:

[0114] Obtain a text layout data set;

[0115] Use the training texts in the text layout data to train a preset object detection model to obtain a trained text layout detection model.

[0116] Specifically, the text layout data set includes a number of training texts and the corresponding labeled target information. The labeled target information includes labeled category information and labeled location information, and each training text includes one or more of titles, natural paragraphs, and tables. For example, the training text is a single-page financial pdf text, and the labeled target information is (class, x1, y1, x2, y2). Class represents the labeled category information of the detection target (title, table, or natural paragraph), (x1, y1) represents the upper left coordinates of the detection target, and (x2, y2) represents the lower right coordinates of the detection target. The origin of the coordinates is the upper left corner of the single-page financial pdf text, and the x and y axes are the short side and the long side of the single-page financial pdf text, respectively.

[0117] Furthermore, when training a preset object detection model using the training text in the text layout data, the training text is input into the preset object detection model, and the predicted target information corresponding to the training text is output through the preset object detection model. The predicted target information includes predicted category information and predicted location information. Then, a loss term (such as a contrast loss term, an L1 loss term, etc.) is constructed based on the predicted target information and the annotated target information, and the preset object detection model is trained based on this loss term to obtain a trained text layout detection model. In addition, in practical applications, the preset object detection model can adopt a pre-trained open-source object detection model, such as the YOLO series, etc., and then the pre-trained object detection model is fine-tuned using the text layout data set.

[0118] It should be noted that when performing object detection on an existing rating report through the trained text layout detection model, the existing rating report can be split page by page first, and the page numbers of each text page obtained by splitting in the existing rating report are recorded. Then, each text page is subjected to object detection through the text layout detection model to obtain the category information and detection location information of the detection targets in each text page. Then, the page numbers of each text page are added to the detection location information to obtain the location information of the detection targets.

[0119] After detecting the location information of the detection target, the target document corresponding to the detection target can be determined based on the location information. Then, by parsing the target document (such as optical character recognition, etc.), the text content and font size can be obtained, and the text content and font size are used as the text content of the detection target. Of course, in practical applications, since existing rating reports generally adopt the pdf format, after obtaining the target document corresponding to the detection target, a pdf parsing tool can be directly used to parse the target document to obtain the text content and font size corresponding to the detection target. Finally, the category information, location information, and content information of the detection target are used as the detection data of the detection target, and the detection targets of each existing rating report and the detection data corresponding to the detection targets are obtained.

[0120] Furthermore, in step S13, after obtaining the detection data of the detection targets included in the existing rating report, the rating method used in the existing rating report can be searched in the detection data of the detection targets included in the existing rating report, and several scoring items corresponding to each evaluation rating method and multiple scoring data for each scoring item can be determined.

[0121] Exemplarily, determining the rating method adopted by each existing rating report, several scoring items corresponding to the rating method, and multiple scoring data for the scoring items based on the detection data of all detection targets of each existing rating report specifically includes:

[0122] S131. Determine the rating method adopted in the existing rating report based on the detection data of the natural paragraph;

[0123] S132. Determine several scoring items corresponding to the rating method and multiple scoring data of the scoring items based on the detection data of the detection target.

[0124] In step S131, the rating method is generally recorded in the existing rating report in text form. Therefore, when determining the rating method adopted in the existing rating report, search for the rating method adopted in the existing rating report in the detection data of the natural paragraph. That is to say, the rating method can be used as the search keyword to search for the rating method adopted in the existing rating report in the text content corresponding to the natural paragraph. In addition, in practical applications, there can be different versions of the rating method. Therefore, when determining the rating method, the version number of the rating method can also be searched synchronously to more accurately determine the rating method adopted in the existing rating report. Correspondingly, when determining the search keyword, the rating method and the version number can be used as the search keyword, and then according to this search keyword, search for the rating method and its version number adopted in the existing rating report in the text content corresponding to the natural paragraph, and use the found rating method and its version number as the rating method of the existing rating report. Of course, in practical applications, the rating method and its version number adopted in the existing rating report can also be directly searched according to the search keyword in the text content of all detection targets.

[0125] In step S132, after obtaining the rating method, search for several scoring items corresponding to the rating method and multiple scoring data of each scoring item in the detection data of all detection targets corresponding to the existing rating report. For example, first determine the scoring item names of each scoring item included in the rating method, then determine the scoring values of each scoring item based on the scoring item names, and then determine the scoring basis and quantification data of each scoring item.

[0126] Based on this, determining several scoring items corresponding to the rating method and the scoring data of the scoring items based on the detection data of the detection target specifically includes:

[0127] S1321. Determine several scoring items corresponding to the rating method, the scoring item names, and the scoring values of each scoring item based on the detection data of the table;

[0128] S1322. Determine the scoring basis of each scoring item based on the detection data of the title and the natural paragraph;

[0129] S1323. Obtain the rating quantification method corresponding to each scoring item, and determine the quantification data of each scoring item based on the rating quantification method through a preset large language model.

[0130] In step S1321, in the existing rating report, the rating elements and scores are generally presented in the form of a table, so that the name of the scoring item included in the rating method and the corresponding score value for each scoring item name can be found in the text content of the table. Specifically, the rating elements and scores can be used as search keywords to find the name of each scoring item included in the rating method and the score value in the text content of the table. Among them, the column where "rating element" is located is the name of the scoring item of the rating method, and each scoring item name corresponds to a score. The column where "score" is located is the score value corresponding to the scoring item.

[0131] S1322. After obtaining the scoring item name and the score value, in order to enable the large model to better understand the reason for rating this score for this scoring item, the scoring basis for each scoring item can be obtained, and a text explanation of the scoring reason can be given through this scoring basis. Among them, the scoring basis can be obtained by searching the text content of the detection target. In practical applications, the scoring basis will be recorded in natural paragraphs. Therefore, when obtaining the scoring basis, the text content of the natural paragraph used to explain the scoring item can be used to determine the scoring basis.

[0132] Exemplarily, the determination of the scoring basis for each scoring item based on the detection data of the title and the natural paragraph specifically includes:

[0133] Determine the target title corresponding to each scoring item, and obtain the target natural paragraph corresponding to the target title;

[0134] Based on the text content in the detection data of the target title and the text content in the detection data of the target natural paragraph, determine the scoring basis for the scoring item name.

[0135] Specifically, the target title is a title in the existing rating report, and the target title includes the scoring item name corresponding to the scoring item. That is to say, the name of the scoring item can be used as a keyword to find the target title corresponding to each scoring item in all the titles of the existing rating report, and then obtain the target natural paragraph corresponding to the target title, where the target natural paragraph is the natural paragraph subordinate to the target title.

[0136] Exemplarily, the determination of the target title corresponding to each scoring item and the acquisition of the target natural paragraph corresponding to the target title specifically include:

[0137] Obtain the structured directory of the existing rating report;

[0138] According to the scoring item name of each scoring item and the structured directory, determine the target title corresponding to each scoring item;

[0139] Determine the target natural paragraph corresponding to the target title according to the report content included in the target title.

[0140] Specifically, the structured table of contents is composed of multi-level headings and the content included in each level of headings. Among them, the structured table of contents structure can adopt a tree structure in JSON format.

[0141] Illustrative example: The table of contents structure can be as follows:

[0143] "Name of the first-level heading":{

[0144] "Content of the natural paragraph":...,

[0145] "Content of the table": ...,

[0146] "Sub-heading":

[0147] "Name of the second-level heading":{

[0148] "Content of the natural paragraph":...,

[0149] "Content of the table":...,

[0150] "Sub-heading":

[0151] "Name of the third-level heading":{

[0152] "Content of the natural paragraph": ...,

[0153] "Content of the table": ...,

[0154] },

[0156] },

[0158] },

[0160] Exemplarily, the construction process of the structured table of contents specifically includes:

[0161] N10. Determine the heading levels of each heading in the existing rating report, and determine the subordination relationship between each heading and the content included in each heading based on the heading levels of each heading and the position information of each heading;

[0162] N20. Construct the structured table of contents of the existing rating report based on the subordination relationship between each heading and the content included in each heading.

[0163] In step N10, the heading level is obtained by grading the headings and is used to reflect which level of heading the heading is. For example, the heading levels include first-level headings, second-level headings, third-level headings, etc. Among them, the heading level can be determined according to the font size and serial number in the text content of the heading. ​​​​

[0164] Exemplarily, the determination of the title levels of each title in the existing rating report specifically includes:

[0165] Obtain the title font size and title serial number of each title;

[0166] Based on the obtained title font size and title serial number of each title, determine the title level of each title.

[0167] Specifically, the title font size and title serial number can be obtained by reading the text content of the title. The title serial number is the text at the very front in the text content of the title, such as "(1)", "1.", etc. The title font size is the font size used for the title. After obtaining the title font size and title serial number, the title level of each title can be determined according to the preset title level conditions. Among them, the title level conditions can be: if the title font size is greater than the average font size of the natural paragraph and there is no serial number, the title level of this title is the first-level; if the title font size is equal to the average font size of the natural paragraph and the title serial number is the serial number in the first preset form (such as the serial number in the form of "(1)"), the title level of this title is the second-level; if the title font size is equal to the average font size of the natural paragraph and the title serial number is the serial number in the first preset form (such as the serial number in the form of "1."), the title level of this title is the third-level.

[0168] It should be noted that in actual applications, other methods can also be used to determine the title level. For example, input the text content of the title into a large model and output the title level through the large model.

[0169] Furthermore, the subordinate relationship between titles refers to the title (denoted as the second title) that is located between two adjacent titles of the same title level (denoted as the first title) and is one level lower than the title level of the first title is the sub-title of the first title in the front among the two first titles. For example, the title of the second level between two adjacent titles of the first level is the sub-title of the first title of the first level in the front, and the title of the third level between two adjacent titles of the second level is the sub-title of the first title of the second level in the front.

[0170] The content included in each title includes the text content included in the title itself, the text content included in the sub-titles that have a subordinate relationship with each title, and the natural paragraph content located after each title and its sub-titles. That is to say, the content included in each title includes the text content of all detection targets between this title and a specific title, as well as the text content included in this title itself, where the specific title has the same title level as this title and is the adjacent title located after this title.

[0171] Exemplarily, the determination process of the content included in each title specifically includes:

[0172] Based on the location information of each detection target, sort each detection target in the order of appearance in the existing rating report to obtain a detection target sequence;

[0173] Take the detection targets and detection data between two adjacent same-level headings in the detection target sequence as the content included in the heading sorted earlier.

[0174] Specifically, the detection target sequence includes all detection targets, and each detection target is arranged in the order of appearance in the existing rating report. That is, for detection target a and detection target b in the detection target sequence, when detection target a is before detection target b, it means that detection target a appears in the existing rating report before detection target b. Among them, the detection target sequence can be obtained by sorting the detection targets with the page number as the first keyword and the y-direction coordinate in the upper left corner coordinate as the second keyword, or, with the page number as the first keyword and the y-direction coordinate in the lower right corner coordinate as the second keyword, etc. In addition, after obtaining the order of appearance of each detection target in the existing rating report, attribute the detection targets between two adjacent same-heading-level headings to the heading content of the previous heading to obtain the content included in each heading, where the heading content can include natural paragraphs and / or tables.

[0175] Further, in step N20, after obtaining the subordinate relationship between each heading and the content included in each heading, establish a heading structure according to the subordinate relationship for each heading, and then add the content included in each heading behind each heading to obtain a structured table of contents. The above completes the description of the process of constructing the structured table of contents.

[0176] Further, after obtaining the structured table of contents, search for the target headings corresponding to each scoring item with the scoring item name as the keyword in the structured table of contents. That is, find the heading with the highest-level heading containing the scoring item target in the structured table of contents, and use the found heading as the target heading corresponding to the scoring item. After obtaining the target heading, read the content included in the heading, and then obtain the natural paragraph in the content included in the heading to obtain the target natural paragraph corresponding to the target heading.

[0177] Further, after obtaining the target heading and the target natural paragraph, use the text content in the detection data of the target heading and the text content in the detection data of the target natural paragraph as the scoring basis for the scoring item name. Specifically, splice the text content of the target natural paragraph and the text content of the target heading, and use the spliced text content as the scoring basis for the scoring item, where the scoring basis can include the business conditions of the enterprise object, etc. In addition, it should be noted that when the target heading includes sub-headings, the natural paragraphs included in the sub-headings are also the target natural paragraphs corresponding to the target heading.

[0178] Further, in step S1323, after obtaining the name, score value, and scoring basis of the scoring item, it is also possible to determine whether the scoring item is quantifiable. When it is quantifiable, it is also possible to obtain the rating quantification method of the scoring item, and use this rating quantification method as an element item in the scoring item, so that the rating quantification method can be directly used subsequently to determine the score of the scoring item. The rating quantification method can be obtained by parsing the rating quantification method file corresponding to the rating method. The rating quantification method file includes the rating quantification methods of each scoring item in the object rating method. Among them, the parsing process can be implemented through the above steps S11 - S13, which will not be elaborated here. In addition, after obtaining the rating quantification method, the rating quantification method can be input into a preset large language model, and the code function and quantification index corresponding to the rating quantification method can be output through the preset large language model to obtain the quantification data of the scoring item.

[0179] It should be noted that when determining several scoring items corresponding to the rating method and the scoring data of the scoring items, it is also possible to first determine the name of the scoring item, and then determine the scoring basis, score value, and quantification data. Or, first determine the name of the scoring item, and then determine the quantification data, scoring basis, and score value, etc. That is to say, the acquisition order of the scoring basis, score value, and quantification data can be adjusted according to user needs, and no specific restrictions are imposed here.

[0180] In the embodiment of the present application, a rating method set is formed by parsing the existing rating report, and each rating method in the rating method set corresponds to several scoring items including the name of the scoring item, score value, scoring basis, and quantification data. In this way, the rating method set can be used as a prior knowledge set to provide reference for the large model through the score value and scoring basis of the scoring item, so that the large model can more accurately score the scoring items of the object to be rated, thereby improving the rating accuracy. At the same time, for quantifiable scoring items, the scoring items can be directly quantified based on the quantification data, which can improve the calculation speed of the scoring of the scoring items, and further improve the rating efficiency of enterprise objects.

[0181] S20. Determine the target rating method corresponding to the object to be rated and the target scoring items corresponding to the target rating method according to the rating method set.

[0182] Specifically, the target rating method is included in the rating method set. That is, when it is necessary to rate the object to be rated, a target rating method can be selected for the object to be rated from the rating method set. Among them, when selecting a target rating method for the object to be rated, the target rating method can be selected for the object to be rated according to the industry in which the object to be rated is located, or a prompt word can be constructed for the object to be rated, and then based on the prompt word and the object name of the object to be rated, the target rating method can be selected for the object to be rated through a large model. It can also be to select the target rating method by performing string matching or embedding similarity matching between the enterprise object profile of the object to be rated and the usage instructions of the rating method, etc. In addition, after obtaining the target rating method, all rating items included in the target rating method can be read, and the quantitative category of each rating item can be determined based on the quantitative data of each rating item, so as to determine the scoring acquisition method of the rating item based on the quantitative category of the rating item (including the first quantitative category and the second quantitative category).

[0183] S30. Obtain the first target rating item in the first quantitative category of the target rating items, determine the first target rating data of the first target rating item, and based on the first target rating data and a preset large language model, rate the object to be rated to determine the first score.

[0184] Specifically, the first quantitative category indicates that the first score of the first target rating item cannot be obtained through quantitative calculation. For example, the first target rating item is the regional layout of branches, etc. For the first target rating item in the first quantitative category, the first score of the first target rating item can be determined through a large model based on the object materials of the object to be rated with reference to the first target rating data corresponding to the first target rating item included in the rating method set. Among them, the first target rating data corresponding to the first target rating item is used to provide prior knowledge for the large model, so that the large model learns the scoring method of the first target rating item, and then rates the first target rating item of the object to be rated. In this way, on the one hand, the accuracy of the first score can be improved, and on the other hand, the consistency and comparability of the scores between the object to be rated and the rated enterprise objects can be ensured.

[0185] Further, each rating method includes several scoring items, and different rating methods may include the same scoring items. Thus, the first target scoring data corresponding to the first target scoring item can be found among the several scoring items of each rating method (i.e., based on the same or similar scoring item names for searching, etc.), so as to obtain several first target scoring data corresponding to the first target scoring item. Among them, all the first target scoring data corresponding to the first target scoring item can be included in the existing rating reports of multiple rated enterprise objects. For example, the rating method set includes rating method A and rating method B. Rating method A includes scoring item A1, scoring item A2, and scoring item A3. Rating method B includes scoring item B1, scoring item B2, and scoring item B3. Scoring item A1 and scoring item B1 both match the first target scoring item. Then, the first target scoring data of the first target scoring item is determined based on scoring item A1 and scoring item B1. For example, the scoring basis in scoring item A1 and the scoring basis in scoring item B1 are used as the first target scoring data of the first target scoring item.

[0186] Exemplarily, as Figure 3 shown, the process of determining the first score by scoring the object to be rated based on the first target scoring data and a preset large language model specifically includes:

[0187] S31. Obtain several keywords corresponding to the first target scoring item;

[0188] S32. Based on the object name of the object to be rated and the several keywords, determine a summary text regarding the several keywords;

[0189] S33. Based on the summary text, select the first target scoring data for the first target scoring item in the rating method set;

[0190] S34. Based on the summary text and the first target scoring data, score the object to be rated through the preset large language model to determine the first score.

[0191] Specifically, in step S31, the several keywords corresponding to the first target scoring item are used as the basis for selecting the summary text of the first target scoring item. That is to say, the several keywords are used as search keywords to determine the summary text regarding the several keywords. Among them, the summary text includes the business conditions of the object to be rated, etc.

[0192] Exemplarily, the process of obtaining the several keywords corresponding to the first target scoring item specifically includes:

[0193] S311. Segment the scoring basis in the first target scoring data to obtain several words;

[0194] S312. Calculate the similarity between each word and the name of the first target scoring item, and the term frequency-inverse document frequency of each word respectively;

[0195] S313. Determine the importance of each word based on the similarity and the term frequency-inverse document frequency of each word;

[0196] S314. Select a number of keywords from a number of words based on the importance of each word to obtain a number of keywords corresponding to the first target scoring item.

[0197] Specifically, the similarity is used to reflect the similarity between the word and the first target scoring item. Among them, the similarity can be the cosine similarity between the word embedding of the word and the word embedding of the first target scoring item. The term frequency-inverse document frequency is used to reflect the importance of the word for all scoring bases of the first target scoring item. Among them, the term frequency refers to the ratio of the number of occurrences of the word in all scoring bases of the first target scoring item to the total number of words in all scoring bases of the first target scoring item, and the inverse document frequency refers to the logarithm of the ratio of the number of scoring bases in which the word appears to the total number of all scoring bases. The term frequency-inverse document frequency is equal to the product of the term frequency and the inverse document frequency.

[0198] After obtaining the similarity and the term frequency-inverse document frequency of the word, the product of the similarity and the term frequency-inverse document frequency can be used as the importance of the word, and then a number of words are selected as the keywords of the first target scoring item in the order from high to low importance. Of course, in practical applications, the importance of the word can also be determined by weighting the similarity and the term frequency-inverse document frequency, etc.

[0199] In the embodiment of the present application, a number of keywords are selected for the first target scoring item based on the importance. Subsequently, a summary text used as the rating basis for the object to be rated can be retrieved based on the number of keywords, improving the interpretability of the summary text for the scoring item, so as to highlight the scoring basis for scoring.

[0200] It should be noted that the number of keywords can also be determined in other ways. For example, it can be set artificially, or the similarity between each word and the name of the first target scoring item can be directly used as the importance, or the term frequency-inverse document frequency can be directly used as the importance, etc.

[0201] Further, in step S32, the summary text is the text material used to reflect the enterprise object status of the object to be rated. For example, the summary text may include the text material reflecting the operating status of the object to be rated, etc. Among them, the summary material can be obtained by searching with several keywords as search keywords. For example, several keywords can be used as search keywords to search the local database materials, or several keywords can be used as search keywords to conduct an online search, or several keywords can be used as search keywords to synchronously search the local database materials and the online search.

[0202] Exemplarily, determining the summary text regarding several keywords based on the object name of the object to be rated and the several keywords specifically includes:

[0203] S321. Search the text resources of the object to be rated based on the object name of the object to be rated and several keywords;

[0204] S322. Divide the searched text resources into several natural paragraphs, and select a preset number of representative natural paragraphs from the several natural paragraphs based on the object name and several keywords;

[0205] S323. Determine the summary text regarding the several keywords based on the preset number of representative natural paragraphs and the several keywords through a large language model.

[0206] Specifically, the object name of the object to be rated and several keywords are used as search keywords to search for the text resources including several keywords regarding the object to be rated, and this text material is used as the basis for scoring the first target scoring item of the object to be rated. Among them, the search process of the text resources can adopt a word matching method or a word embedding similarity method, etc.

[0207] After obtaining the text material, the text material can be divided into several natural paragraphs, and then the several natural paragraphs can be screened based on the similarity between the natural paragraphs and several keywords. Specifically, the keywords can be concatenated with the object name of the object to be rated to form a sentence, and then the sentence embedding of each natural paragraph and the sentence embedding of this sentence are calculated (such as cosine similarity, etc.), and then a preset number of representative natural paragraphs are selected in the order from high to low similarity. Of course, in practical applications, several target natural paragraphs can also be selected according to the number of keywords included in the natural paragraphs.

[0208] After obtaining a preset number of representative natural paragraphs, prompt engineering can be used to utilize a large model to summarize several target natural paragraphs around several keywords to obtain a summary text. Among them, the prompt words used in prompt engineering are used to instruct the large model to summarize around several keywords. For example, the prompt words can be "Please summarize according to the above materials around the following keywords". That is to say, when determining the summary text, a summary prompt word for instructing the large model to summarize around several keywords can be constructed first, and then the summary prompt word, several target natural paragraphs, and several keywords are concatenated and then input into the large model to output the summary text through the large model.

[0209] In this application, by screening the text materials of the object to be rated, and then forming a summary text around several keywords based on the selected target natural paragraphs, the summary text can reflect the enterprise object status (such as business status, etc.) of the object to be rated. In this way, when determining the score of the first target scoring item, the scoring basis and score of the rated enterprise object similar to the enterprise object status of the object to be rated can be selected based on the summary text as a reference to determine the score of the first target scoring item. This can not only improve the scoring accuracy, but also ensure the score consistency and comparability between the object to be rated and the rated enterprise object.

[0210] Of course, in practical applications, the found text materials can be directly used as the summary text, or several selected target natural paragraphs can be used as the summary text, etc.

[0211] Furthermore, in step S33, several target scoring data are selected from the first target scoring data corresponding to the first target scoring item based on the summary text, and are used as the prior knowledge for determining the first score, so that the large model can determine the score corresponding to the summary text based on this prior knowledge. When selecting several target scoring data from the first target scoring data corresponding to the first target scoring item, it can be selected based on the similarity between the scoring basis in the scoring data and the summary text, or it can be randomly selected, etc.

[0212] In the embodiment of this application, after selecting all the first target scoring data corresponding to the first target scoring item, calculate the similarity (such as cosine similarity, etc.) between the text embedding of the summary text and the text embedding of the scoring basis in each first target scoring data, and then select several target scoring data in descending order of similarity to obtain several target scoring data most similar to the summary text.

[0213] Further, in step S33, after obtaining a number of target rating data, extract the rating basis and rating in each target rating data, and use prompt engineering to determine the first rating of the first target rating item through a large model. Exemplarily, rating the object to be rated through a preset large language model based on the summary text and the first target rating data to determine the first rating specifically includes:

[0214] Construct a rating prompt for the first target rating item;

[0215] Concatenate the rating prompt, the first target rating item, the first target rating data of the first target rating item, and the summary text and input them into the preset large language model, and output the first rating of the first target rating item through the preset large language model.

[0216] Specifically, the rating prompt is used to prompt the large model to output the first rating of the first target rating item. For example, the rating prompt is "The following is a description of some enterprise objects regarding {fill in the name of the rating item} and the corresponding scores. Please use this as a reference to rate the enterprise objects described below". After determining the rating prompt, concatenate the rating basis and rating in each target rating data as an example, then concatenate the rating prompt with the summary text and input them into the large model, and output the first rating of the first target rating item through the large model.

[0217] It should be noted that in actual applications, the object to be rated, the first target rating item, and the first target rating data corresponding to the first target rating item can also be directly used as the input items of the large model. The large model obtains the enterprise object information of the object to be rated, and determines the rating of the first target rating item based on the obtained enterprise object information, the first target rating item, and the first target rating data corresponding to the first target rating item. Among them, the rating of the first target rating item given for the enterprise object information of the object to be rated, that is, the rating of the first target rating item is a rating of the object to be rated.

[0218] S40. Obtain the second target rating item of the second quantitative category in the target rating items, determine the second target rating data of the second target rating item, and rate the object to be rated based on the second target rating data to determine the second rating.

[0219] Specifically, the second target scoring item of the second quantization category is the scoring item for which the second score can be directly calculated by the rating quantization method. That is to say, for the second target scoring item of the second quantization category, the corresponding rating quantization method of the second target scoring item can be directly determined, and then the second score of the second target scoring item can be calculated through the rating quantization method. For this purpose, for the second target scoring item of the second quantization category, the second target scoring data corresponding to the second target scoring item can be directly selected from the rating method set, and then the second score of the second target scoring item can be calculated based on the second target scoring data. That is to say, the second target scoring data can be quantization data. Of course, the second target scoring data can also include the scoring item name, score value, scoring basis, and quantization data corresponding to the scoring item, and then the quantization data can be read from the second target scoring data. In addition, when there are multiple second target scoring data corresponding to the second target scoring item, one second target scoring data can be randomly selected from the multiple second target scoring data, and then the quantization data in the second target scoring data can be used as the quantization data for determining the second score of the second target scoring item.

[0220] Exemplarily, as Figure 4 shown, the process of scoring the object to be rated based on the second target scoring data and determining the second score specifically includes:

[0221] S41. Obtain the quantization data in the second target scoring data, where the quantization data includes the code function corresponding to the rating quantization method and the quantization indicators required by the rating quantization method;

[0222] S42. Use the quantization index data as the input of the code function, and execute the code function to obtain the second score of the second target scoring item.

[0223] Specifically, the quantization indicators can be financial indicators, enterprise object scale indicators, etc. for calculating the second score. After the quantization indicators are read, the corresponding quantization index data of the quantization indicators will be queried from the enterprise object materials of the object to be rated, and then the quantization index data will be used as the input of the code function. By executing the code function, the second score of the second target scoring item can be obtained, where the code function can adopt Python functions, etc.

[0224] In this application, by pre-constructing the code function, and then after obtaining the quantization index data, inputting the quantization index data into the code function can quickly calculate the second score of the second target scoring item, making the scoring process simple and accurate results can be obtained.

[0225] S50. Determine the scoring grade of the object to be rated according to the first score and the second score.

[0226] Specifically, after obtaining all the first scores and all the second scores, the first scores and the second scores can be directly weighted and summed to obtain the final score of the object to be rated of the enterprise to be evaluated, and finally, according to the corresponding relationship between the final score and the rating of the enterprise object, the score level of the object to be rated is determined. Of course, the average of all the first scores and all the second scores can also be used as the final score of the object to be rated of the enterprise to be evaluated, etc.

[0227] Exemplarily, determining the score level of the object to be rated according to the first score and the second score specifically includes:

[0228] Weight the first score and the second score to obtain the final score;

[0229] According to the preset corresponding relationship between the score level and the score, determine the score level corresponding to the final score to obtain the score level of the object to be rated.

[0230] Specifically, weight coefficients can be pre-configured for each scoring item, and then based on the weight coefficients of each scoring item, the scores corresponding to each scoring item (the first score or the second score) are weighted. This can adjust the weight coefficients of each scoring item according to actual needs to make the final score more accurate. The corresponding relationship between the score level and the score of the enterprise object is pre-set. For example, several score levels are pre-set, and then the score intervals corresponding to each score level are set to form the corresponding relationship between the score level and the score of the enterprise object. Of course, the corresponding relationship between the score level and the score of the enterprise object can also be adjusted according to the actual situation.

[0231] In summary, this embodiment provides an object rating method, which includes obtaining a preset rating method set; determining a target rating method corresponding to the object to be rated and target scoring items corresponding to the target rating method according to the rating method set; obtaining first target scoring items of a first quantization category in the target scoring items, determining first target scoring data of the first target scoring items, and rating the object to be rated based on the first target scoring data and a preset large language model to determine a first score; for second target scoring items that can be quantified in the object rating method, determining a second score of the second target scoring items based on quantization data in second target scoring data corresponding to the second target scoring items; and determining a rating level of the object to be rated according to the first score and the second score. In this application, a rating method set formed based on existing rating reports of rated enterprise objects is used as prior knowledge. The score of non-quantifiable scoring items corresponding to the object to be rated is determined through a large model, and the score of quantifiable scoring items is directly calculated through quantization data in the rating method set. This can not only reduce the labor cost and time cost required for manual rating, but also save the time cost spent in the process of screening enterprise object materials of the object to be rated. At the same time, by using the rating method set as prior knowledge, the accuracy of enterprise object rating and the rating consistency with rated enterprise objects can also be improved.

[0232] Based on the above object rating method, this embodiment provides an object rating device, as Figure 5 shown. The object rating device specifically includes:

[0233] A construction module 100, configured to construct a rating method set based on existing rating reports. The rating method set includes several rating methods and several scoring items corresponding to each rating method. The scoring items include multiple scoring data target rating methods;

[0234] A rating method determination module 200, configured to determine a target rating method corresponding to the object to be rated and target scoring items corresponding to the target rating method according to the rating method set;

[0235] A first processing module 300, configured to obtain first target scoring items of a first quantization category in the target scoring items, determine first target scoring data of the first target scoring items, and rate the object to be rated based on the first target scoring data and a preset large language model to determine a first score;

[0236] A second processing module 400, configured to obtain second target scoring items of a second quantization category in the target scoring items, determine second target scoring data of the second target scoring items, and rate the object to be rated based on the second target scoring data to determine a second score;

[0237] A rating determination module 500 is configured to determine the rating level of the object to be rated according to the first score and the second score.

[0238] Based on the above object rating method, this embodiment provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the object rating method as described in the above embodiment.

[0239] Based on the above object rating method, this application also provides a terminal device, as Figure 6 shown, which includes at least one processor 20; a display screen 21; and a memory 22. It may further include a communication interface 23 and a bus 24. Among them, the processor 20, the display screen 21, the memory 22, and the communication interface 23 can communicate with each other through the bus 24. The display screen 21 is set to display a user guidance interface preset in the initial setting mode. The communication interface 23 can transmit information. The processor 20 can call the logical instructions in the memory 22 to execute the method in the above embodiment.

[0240] In addition, when the logical instructions in the above memory 22 are implemented in the form of a software functional unit and sold or used as an independent product, they can be stored in a computer-readable storage medium.

[0241] The memory 22, as a computer-readable storage medium, can be set to store software programs and computer-executable programs, such as the program instructions or modules corresponding to the method in the embodiment of the present disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 22, that is, to implement the method in the above embodiment.

[0242] The memory 22 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the terminal device, etc. In addition, the memory 22 may include a high-speed random access memory and may also include a non-volatile memory. For example, various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes can also be a transient storage medium.

[0243] In addition, the specific processes of loading and executing the multiple instructions by the instruction processor in the above storage medium and terminal device have been described in detail in the above method, and will not be repeated here one by one.

[0244] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An object rating method, characterized in that, The described object rating method specifically includes: Constructing a rating method set based on existing rating reports. The rating method set includes several rating methods and several scoring items corresponding to each rating method. The scoring items include multiple scoring data; Determining the target rating method corresponding to the object to be rated and the target scoring items corresponding to the target rating method according to the rating method set; Obtaining the first target scoring item of the first quantization category in the target scoring items, determining the first target scoring data of the first target scoring item, and scoring the object to be rated based on the first target scoring data and a preset large language model to determine the first score, where the first quantization category represents a target scoring item that cannot be calculated by quantization; Obtaining the second target scoring item of the second quantization category in the target scoring items, determining the second target scoring data of the second target scoring item, and scoring the object to be rated based on the second target scoring data to determine the second score; Determining the scoring level of the object to be rated according to the first score and the second score; Among them, the specific process of constructing the rating method set based on existing rating reports includes: Obtaining several existing rating reports; Respectively obtaining the detection targets of each existing rating report and the detection data corresponding to the detection targets. The detection targets include titles, natural paragraphs, and tables, and the detection data includes category information, location information, and text content; Determining the rating methods adopted by each existing rating report, the several scoring items corresponding to the rating methods, and the multiple scoring data of the scoring items based on the detection data of all detection targets of each existing rating report; Constructing a rating method set according to the determined rating methods, the several scoring items corresponding to each rating method, and the multiple scoring data of the scoring items; The specific process of respectively obtaining the detection targets of each existing rating report and the detection data corresponding to the detection targets includes: Inputting each existing rating report into a trained text layout detection model, and outputting the detection targets in each existing rating report and the category information and location information of the detection targets through the text layout detection model; Extracting the text content of each detection target in the corresponding existing rating report based on the location information of each detection target; Taking the category information, the location information, and the text content as the detection data of the detection target to obtain the detection targets of each existing rating report and the detection data corresponding to the detection targets.

2. The object rating method according to claim 1, characterized in that The trained text layout detection model specifically includes: Obtaining a text layout data set, where the text layout data set includes several training texts and the labeled target information corresponding to the training texts; Training a preset target detection model using the training texts in the text layout data to obtain a trained text layout detection model.

3. The object rating method according to claim 1, characterized in that, The specific process of determining the rating methods adopted by each existing rating report, the several scoring items corresponding to the rating methods, and the multiple scoring data of the scoring items based on the detection data of all detection targets of each existing rating report includes: Based on the detection data of the natural paragraph, determine the rating method adopted by the existing rating report; Based on the detection data of the detection target, determine several scoring items corresponding to the rating method and multiple scoring data of the scoring items.

4. The object rating method according to claim 3, characterized in that, The scoring data includes the scoring item name, scoring value, scoring basis, and quantification data. Based on the detection data of the detection target, determining several scoring items corresponding to the rating method and the scoring data of the scoring items specifically includes: Based on the detection data of the table, determine several scoring items corresponding to the rating method, the scoring item name, and the scoring value of each scoring item; Based on the detection data of the title and the natural paragraph, determine the scoring basis of each scoring item; Obtain the rating quantification method corresponding to each scoring item, and based on the rating quantification method, determine the quantification data of each scoring item through a preset large language model.

5. The object rating method according to claim 4, characterized in that The specific process of determining the scoring basis of each scoring item based on the detection data of the title and the natural paragraph includes: Determine the target title corresponding to each scoring item, and obtain the target natural paragraph corresponding to the target title; Based on the text content in the detection data of the target title and the text content in the detection data of the target natural paragraph, determine the scoring basis of the scoring item name.

6. The object rating method according to claim 5, characterized in that, The specific process of determining the target title corresponding to each scoring item and obtaining the target natural paragraph corresponding to the target title includes: Obtain the structured table of contents of the existing rating report, where the structured table of contents is composed of multi-level headings and the content included in each level of headings; According to the scoring item name of each scoring item and the structured table of contents, determine the target title corresponding to each scoring item; According to the report content included in the target title, determine the target natural paragraph corresponding to the target title.

7. The object rating method according to claim 6, characterized in that, The specific process of constructing the structured table of contents includes: Determine the title level of each heading in the existing rating report, and based on the title level of each heading and the position information of each heading, determine the subordinate relationship between each heading and the content included in each heading; Based on the subordinate relationship between each heading and the content included in each heading, construct the structured table of contents of the existing rating report.

8. The object rating method according to claim 7, characterized in that, The specific process of determining the title level of each heading in the existing rating report includes: Obtain the title font size and title serial number of each heading; Based on the obtained title font size and title serial number of each heading, determine the title level of each heading.

9. The object rating method according to claim 7, characterized in that, The specific process of determining the content included in each heading includes: Based on the position information of each detection target, sort each detection target in the order of appearance in the existing rating report to obtain a detection target sequence; Take the detection targets and detection data between two adjacent same-level headings in the detection target sequence as the content included in the heading sorted earlier.

10. The object rating method according to claim 1, characterized in that, The specific process of scoring the object to be rated based on the first target scoring data and the preset large language model to determine the first score includes: Obtain several keywords corresponding to the first target scoring item; Based on the object name of the object to be rated and the several keywords, determine a summary text about the several keywords; Based on the summary text, select the first target scoring data for the first target scoring item from the rating method set; Based on the summary text and the first target scoring data, use a preset large language model to score the object to be rated and determine the first score.

11. The object rating method according to claim 10, characterized in that, The specific steps of using a preset large language model to score the object to be rated based on the summary text and the first target scoring data to determine the first score include: Construct a scoring prompt for the first target scoring item; Concatenate the scoring prompt, the first target scoring item, the first target scoring data of the first target scoring item, and the summary text, and then input them into the preset large language model. Output the first score of the first target scoring item through the preset large language model.

12. The object rating method according to claim 10, characterized in that, The specific steps of obtaining several keywords corresponding to the first target scoring item include: Segment the scoring basis in the first target scoring data to obtain several words; Calculate the similarity between each word and the name of the scoring item of the first target scoring item and the term frequency-inverse document frequency of each word respectively; Based on the similarity and the term frequency-inverse document frequency of each word, determine the importance of each word; Select several keywords from several words based on the importance of each word to obtain several keywords corresponding to the first target scoring item.

13. The object rating method according to claim 10, characterized in that, The specific steps of determining the summary text about several keywords based on the object name of the object to be rated and the several keywords include: Search for the text resources of the object to be rated based on the object name of the object to be rated and several keywords; Divide the searched text resources into several natural paragraphs, and select a preset number of representative natural paragraphs from the several natural paragraphs based on the object name and several keywords; Based on the preset number of representative natural paragraphs and the several keywords, determine the summary text about the several keywords through a preset large language model.

14. The object rating method according to claim 1, wherein The specific steps of using the second target scoring data to score the object to be rated and determine the second score include: Obtain the quantitative data in the second target scoring data, where the quantitative data includes the code function corresponding to the rating quantification method and the quantitative indicators required for the rating quantification method; Obtain the quantitative indicator data corresponding to the quantitative indicators; Use the quantitative indicator data as the input of the code function to obtain the second score of the second target scoring item.

15. The object rating method according to claim 1, characterized in that, The specific steps of determining the rating level of the object to be rated according to the first score and the second score include: Weight the first score and the second score to obtain the final score; According to the corresponding relationship between the preset rating level and the score, determine the rating level corresponding to the final score to obtain the rating level of the object to be rated.

16. An object rating device, characterized in that, The object rating device specifically includes: A construction module for constructing a rating method set based on existing rating reports. The rating method set includes several rating methods and several scoring items corresponding to each rating method. The scoring items include multiple scoring data target rating methods; A rating method determination module for determining the target rating method corresponding to the object to be rated and the target scoring item corresponding to the target rating method according to the rating method set; The first processing module is configured to obtain the first target scoring item of the first quantitative category in the target scoring items, determine the first target scoring data of the first target scoring item, and score the object to be rated based on the first target scoring data and a preset large language model to determine a first score, where the first quantitative category represents the target scoring item that cannot be calculated quantitatively; The second processing module is configured to obtain the second target scoring item of the second quantitative category in the target scoring items, determine the second target scoring data of the second target scoring item, and score the object to be rated based on the second target scoring data to determine a second score; The grade determination module is configured to determine the scoring grade of the object to be rated according to the first score and the second score; Wherein, the constructing the rating method set based on the existing rating reports specifically includes: Obtain a number of existing rating reports; Respectively obtain the detection target of each existing rating report and the detection data corresponding to the detection target, where the detection target includes the title, natural paragraph and table, and the detection data includes category information, location information and text content; Based on the detection data of all detection targets of each existing rating report, determine the rating method adopted by each existing rating report, several scoring items corresponding to the rating method, and multiple scoring data of the scoring items; Construct a rating method set according to the determined rating methods, several scoring items corresponding to each rating method, and multiple scoring data of the scoring items; The respectively obtaining the detection target of each existing rating report and the detection data corresponding to the detection target specifically includes: Input each existing rating report into a trained text layout detection model, and output the detection target in each existing rating report and the category information and location information of the detection target through the text layout detection model; Extract the text content of each detection target in the corresponding existing rating report respectively based on the location information of each detection target; Use the category information, the location information and the text content as the detection data of the detection target to obtain the detection target of each existing rating report and the detection data corresponding to the detection target.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the object rating method according to any one of claims 1-15.

18. A terminal device, characterized in that, Including: A processor and a memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, it implements the steps in the object rating method according to any one of claims 1-15.

Images