A system and method for automatically generating an evaluation report

Abstract

The present invention discloses an automatic generation system and method for an evaluation report, comprising the following steps: obtaining data keywords according to a survey data table; matching a report template to generate an evaluation data report; inputting the data keywords into a Genius model to obtain a text report, and using grammatical rules to detect and adjust the text report to obtain an evaluation text report; merging the evaluation data report and the evaluation text report to obtain an evaluation report; if any survey data is modified, updating other survey data according to a data association table to obtain an updated survey data table, and repeating steps S1 to S3 to obtain a final version of the evaluation report; users evaluate the final version of the evaluation report to obtain scoring information and adjust the report generation process; obtaining a text report through data keywords, and using grammatical rules to adjust the text report to make the content of the evaluation report more flexible and more targeted, saving manpower and material resources, ensuring the content quality of the evaluation report, and expanding the scope of application of the automatic report generation method.

Description

Technical Field

[0001] The present invention relates to the technical field of information technology, and in particular to a system and method for automatically generating an evaluation report. Background Art

[0002] In recent years, with the continuous advancement of my country's informatization construction, the amount of enterprise data has shown an explosive growth trend. Enterprise managers need to transform the phenomena or trends reflected behind the data into effective resources for enterprise management in order to make more accurate decisions. In this process, a good report can assist decision makers in making accurate decisions and improve work efficiency.

[0003] At present, the Chinese invention patent with application number 202211372432.6 provides a method and device for automatic report generation and report template configuration, which parses the report template to determine the variables in the report template; displays the list interface to be configured according to the variables; the list interface to be configured includes input controls corresponding to each variable; determines the method for obtaining data for each variable in response to the input operation for each input control; saves the report template and the corresponding methods for obtaining data to complete the configuration of the report template. Although this method completes the development of the report template in a configured manner, the configured report template can automatically render and generate a report. Although it improves the development efficiency of the report template, it only replaces and modifies the data part of the report, and does not generate a targeted report. The scope of application is limited, and there is no quality assessment of the generated report and adjustment of the report generation process, which is not conducive to the quality of report generation. Summary of the invention

[0004] The technical problem solved by the present invention is that the prior art only replaces and modifies the data part in the report, does not generate a targeted report, has a limited scope of application, and does not perform quality assessment on the generated report and adjust the report generation process, which is not conducive to the quality of report generation.

[0005] In order to solve the above technical problems, in a first aspect, the present invention provides a method for automatically generating an evaluation report, comprising the following steps:

[0006] Step S1, obtaining a survey data table, inputting the survey data table into a pre-trained sequence generation model, and obtaining data keywords;

[0007] Step S2, matching the report template in the template library according to the data keyword, and inputting the survey data in the survey data table into the corresponding position of the data template in the report template according to the data identifier, to generate an evaluation data report;

[0008] Step S3, inputting the data keywords into the Genius model to obtain a text report, and performing grammatical correction on the text report using the grammatical rules corresponding to the report template matching rule base to obtain an evaluation text report;

[0009] Step S4, merging the evaluation data report and the evaluation text report to obtain an evaluation report;

[0010] Step S5, if any survey data is modified, then update other associated survey data according to the data identifier and data association table of the modified survey data, obtain an updated survey data table, and repeat steps S1 to S3 according to the updated survey data table to obtain a final version of the evaluation report;

[0011] Step S6, the user performs template evaluation, grammar evaluation and data keyword evaluation on the final version of the evaluation report to obtain scoring information, and adjusts the report template, grammar rules and sequence generation model according to the scoring information, wherein the scoring information includes template scoring, grammar scoring and keyword scoring;

[0012] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0013] The step S1 specifically includes the following steps:

[0014] Step S101, obtaining a survey data table;

[0015] Step S102, inputting the survey data in the survey data table into an encoder of a pre-trained sequence generation model, and encoding the survey data into a feature vector of a fixed size according to a first mapping sequence;

[0016] Step S103, inputting the feature vector into the decoder of the sequence generation model, obtaining the keywords corresponding to each survey data according to the second mapping sequence, and obtaining the weight of each keyword by using the self-attention mechanism;

[0017] Step S104, sorting the keywords from large to small according to their weights, and extracting the first n keywords as data keywords of the survey data, where n is a natural number greater than zero;

[0018] Training a sequence generation model involves:

[0019] Obtain a survey data set, perform data preprocessing on the survey data set, divide the preprocessed survey data set into a survey data training set and a survey data test set according to a certain ratio, perform data labeling on the survey data training set, input the labeled survey data training set into the first bidirectional recurrent neural network of the training generation model, train with the survey data as input and the data features of the data labeling as output, and test the trained first bidirectional recurrent neural network using the survey data test set until the fit reaches an expected threshold, then stop training, obtain the encoder of the sequence generation model and generate a first mapping sequence;

[0020] Acquire data features of the data annotation, perform keyword annotation on the data features, obtain data keywords, train a second bidirectional recurrent neural network using the data features as input and the data keywords as output until the degree of fit reaches an expected threshold, then stop training, obtain a decoder of the sequence generation model and generate a second mapping sequence;

[0021] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0022] The step S2 specifically includes the following steps:

[0023] Step S201, using the n data keywords to match corresponding templates in the template library according to the template keywords of each template, and taking the template with the highest keyword matching degree as the report template;

[0024] Step S202, matching the data identifier of the survey data with each placeholder in the report template according to the placeholder identifier, and replacing the successfully matched survey data with the corresponding placeholder in the report template, wherein the data identifier and the placeholder identifier are in a one-to-one correspondence;

[0025] Step S203, using the replaced report template as an evaluation data report;

[0026] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0027] The step S3 specifically comprises the following steps:

[0028] Step S301, input the data keywords into the Genius model to obtain a text report;

[0029] Step S302, using the pySBD library to perform sentence processing on the text report, obtain each text sentence, and number the text sentences;

[0030] Step S303, obtaining the template number of the report template, and matching the corresponding grammar rule in the rule base according to the template number, performing grammar detection on each text sentence respectively using the grammar rule, obtaining the grammar error of the text sentence, and adjusting the text sentence according to the grammar rule;

[0031] Step S304, merging the text sentences according to the order of the numbers to obtain an evaluation text report;

[0032] The training of the Genius model includes:

[0033] Obtain a high-quality text report, extract keywords from the text report using the Yake method, use the Mask method to mask the text report content other than the processed keywords, obtain a text draft, input the text draft into a deep learning model, train the text draft as the model input and the text report as the model output until the model fit reaches an expected threshold, then stop training and obtain a Genius model;

[0034] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0035] The step S6 specifically comprises the following steps:

[0036] Step S601, the user evaluates the fitness of the report template of the final version of the evaluation report to obtain a template score, the user evaluates the sentence grammar of the text report to obtain a grammar score, and the user evaluates the core report content of the text report to obtain a keyword score;

[0037] Step S602, if the template score is less than or equal to the expected threshold, the user adjusts the template keywords of the report template;

[0038] Step S603, if the grammar score is less than or equal to the expected threshold, obtaining the grammar rule usage record of the text report, obtaining the grammar rules used in the text report according to the grammar rule usage record, and the user adjusts the grammar rules according to the grammar dislocation of the text report;

[0039] Step S604, if the keyword score is less than or equal to the expected threshold, adjusting the first mapping sequence of the sequence generation model;

[0040] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0041] The updating of other associated survey data according to the data identifier and data association table of the modified survey data includes:

[0042] Acquire a data identifier of the modified survey data, acquire other data identifiers associated with the data identifier in a data association table and the association relationship therebetween according to the data identifier, and modify other associated survey data according to the modified survey data and the association relationship, wherein the data association table includes the association relationship between the data identifier and each identifier;

[0043] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0044] The step S303 specifically includes the following steps:

[0045] Step S3031, dividing the text sentence into a plurality of participles by using syntactic analysis, and constructing an abstract syntax tree of the text sentence according to the relationship between the participles;

[0046] Step S3032, obtaining the template number of the report template, matching the corresponding grammar rule in the rule base according to the template number, generating the corresponding regular expression according to the grammar rule, and using the regular expression to detect the abstract syntax tree, and obtaining the text sentence corresponding to the abstract syntax tree that fails the detection;

[0047] Step S3033, using the regular expression to adjust several word segments of the text sentence to obtain an adjusted text sentence;

[0048] As a preferred solution of the method for automatically generating an evaluation report according to the present invention, wherein:

[0049] The keywords of the text report extracted by using the Yake method include:

[0050] Preprocessing the text report to remove non-semantic characters, wherein the non-semantic characters include expression symbols and numbers;

[0051] Defining keyword features of the text report, the keyword features including the position of the segmentation, the frequency of occurrence of the segmentation in the sentence, the frequency of occurrence of the segmentation in the text report and the contextual relationship of the segmentation, wherein the contextual relationship of the segmentation is determined by the number of different candidate words of the segmentation;

[0052] Obtaining the position score, sentence frequency score, text frequency score and candidate word score of each participle according to the keyword features, and obtaining the key weight of each participle according to the position score, sentence frequency score, text frequency score and candidate word score;

[0053] Arrange each word in descending order according to the key weight, remove duplicate words, extract the first m words, and define the words as keywords of the text report, where m is a natural number greater than zero.

[0054] In a second aspect, the present invention provides an automatic evaluation report generation system, including a report generation module, a report modification module and a report evaluation module;

[0055] The report generation module is used to obtain a survey data table, obtain data keywords using a sequence generation model according to the survey data table, match a report template according to the data keywords, generate an evaluation data report and an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report;

[0056] The report modification module is used to update the generated evaluation report according to the modified survey data to obtain the final version of the evaluation report;

[0057] The report evaluation module is used by the user to perform template evaluation, grammar evaluation and data keyword evaluation on the final version of the evaluation report, obtain scoring information, and adjust the report template, grammar rules and sequence generation model according to the scoring information.

[0058] As a preferred solution of the automatic generation system of an evaluation report described in the present invention, wherein:

[0059] The report generation module includes a data report generation unit and a text report generation unit;

[0060] The data report generating unit is used to obtain a survey data table, obtain data keywords according to the survey data table, match the report template with the data keywords and generate an evaluation data report;

[0061] The text report generation unit is used to input the data keywords into the Genius model to obtain a text report, and use grammatical rules to detect and adjust the text report to obtain an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report.

[0062] The beneficial effects of the present invention are as follows: the present invention obtains data keywords according to survey data by utilizing a sequence generation model, and inputs the data keywords into a Genius model to obtain a text report, and obtains an evaluation text report by performing grammar detection and adjustment on the text report by utilizing grammar rules, thereby making the content of the evaluation report more flexible and more targeted, saving a lot of manpower and material resources, ensuring the content quality of the evaluation report, and expanding the scope of application of the automatic report generation method.

[0063] By matching report templates with data keywords and replacing corresponding placeholders in the report templates with survey data according to data identifiers, the efficiency of generating evaluation reports is improved. The matching of different templates also ensures the pertinence of evaluation reports and expands the scope of application of automatic report generation methods.

[0064] When the survey data needs to be modified, you only need to update any corresponding survey data, and then you can adjust other related survey data according to the data association table, and update the evaluation data report and evaluation text report in time, making the update process more convenient and simple, saving a lot of manpower and material resources.

[0065] Users conduct grammar evaluation, data keyword evaluation and template evaluation on the final version of the evaluation report, and adjust the grammar rules, template keywords and sequence generation model according to the evaluation results, which improves the generation quality of the evaluation report and ensures a good user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A basic flow chart of a method for automatically generating an evaluation report provided by an embodiment of the present invention;

[0067] Figure 2 A schematic diagram of the basic flow of a system for automatically generating an evaluation report provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0069] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for automatically generating an evaluation report, comprising the following steps:

[0070] Step S1, obtaining a survey data table, inputting the survey data table into a pre-trained sequence generation model, and obtaining data keywords;

[0071] Step S2, matching the report template in the template library according to the data keyword, and inputting the survey data in the survey data table into the corresponding position of the data template in the report template according to the data identifier, to generate an evaluation data report;

[0072] Step S3, inputting the data keywords into the Genius model to obtain a text report, and performing grammatical correction on the text report using the grammatical rules corresponding to the report template matching rule base to obtain an evaluation text report;

[0073] Step S4, merging the evaluation data report and the evaluation text report to obtain an evaluation report;

[0074] Step S5, if any survey data is modified, then update other associated survey data according to the data identifier and data association table of the modified survey data, obtain an updated survey data table, and repeat steps S1 to S3 according to the updated survey data table to obtain a final version of the evaluation report;

[0075] In step S6, the user performs template evaluation, grammar evaluation and data keyword evaluation on the final evaluation report to obtain scoring information, and adjusts the report template, grammar rules and sequence generation model according to the scoring information. The scoring information includes template scoring, grammar scoring and keyword scoring.

[0076] The sequence generation model is a deep learning model that generates meaningful new sequences by learning the inherent rules in ordered data;

[0077] In this embodiment, the data identifier represents letters, numbers or other characters used to identify survey data;

[0078] The Genius model is a pre-trained model that generates complete and fluent text based on fixed key information;

[0079] The data association table is a database table used to record the association relationship between various survey data.

[0080] In this embodiment, a sequence generation model is used to obtain data keywords based on survey data, and the data keywords are input into the Genius model to obtain a text report. The grammar rules are used to perform grammar detection and adjustment on the text report to obtain an evaluation text report. This makes the content of the evaluation report more flexible and targeted, saves a lot of manpower and material resources, ensures the content quality of the evaluation report, and expands the scope of application of the automatic report generation method.

[0081] By matching report templates with data keywords and replacing corresponding placeholders in the report templates with survey data according to data identifiers, the efficiency of generating evaluation reports is improved. The matching of different templates also ensures the pertinence of evaluation reports and expands the scope of application of automatic report generation methods.

[0082] When the survey data needs to be modified, you only need to update any corresponding survey data, and then you can adjust other related survey data according to the data association table, and update the evaluation data report and evaluation text report in time, making the update process more convenient and simple, saving a lot of manpower and material resources.

[0083] Users conduct grammar evaluation, data keyword evaluation and template evaluation on the final version of the evaluation report, and adjust the grammar rules, template keywords and sequence generation model according to the evaluation results, which improves the generation quality of the evaluation report and ensures a good user experience.

[0084] The step S1 specifically includes the following steps:

[0085] Step S101, obtaining a survey data table;

[0086] Step S102, inputting the survey data in the survey data table into an encoder of a pre-trained sequence generation model, and encoding the survey data into a feature vector of a fixed size according to a first mapping sequence;

[0087] Step S103, inputting the feature vector into the decoder of the sequence generation model, obtaining the keywords corresponding to each survey data according to the second mapping sequence, and obtaining the weight of each keyword by using the self-attention mechanism;

[0088] Step S104, sorting the keywords from large to small according to their weights, and extracting the first n keywords as data keywords of the survey data, where n is a natural number greater than zero;

[0089] Training a sequence generation model involves:

[0090] Obtain a survey data set, perform data preprocessing on the survey data set, divide the preprocessed survey data set into a survey data training set and a survey data test set according to a certain ratio, perform data labeling on the survey data training set, input the labeled survey data training set into the first bidirectional recurrent neural network of the training generation model, train with the survey data as input and the data features of the data labeling as output, and test the trained first bidirectional recurrent neural network using the survey data test set until the fit reaches an expected threshold, then stop training, obtain the encoder of the sequence generation model and generate a first mapping sequence;

[0091] The data features of the data annotation are obtained, keywords are annotated on the data features, data keywords are obtained, and a second bidirectional recurrent neural network is trained using the data features as input and the data keywords as output until the degree of fit reaches an expected threshold, then the training is stopped to obtain a decoder of the sequence generation model and generate a second mapping sequence.

[0092] The encoder is used to map the input data to a hidden state in a high-dimensional space that captures key information of the data;

[0093] The decoder is used to gradually generate a target sequence according to the hidden state;

[0094] The data preprocessing refers to a series of processing of raw data in the process of data mining, data analysis and machine learning to improve data quality, reduce noise and extract useful information, including data cleaning, data conversion, data normalization, data reduction and data enhancement;

[0095] Data annotation refers to the process of processing unprocessed voice, images, text and video to transform them into machine-recognizable information;

[0096] The bidirectional recurrent neural network is an artificial neural network that can process input data in both forward and backward directions;

[0097] The degree of fit is used to indicate the degree of fit between the output data and the expected data.

[0098] In this embodiment, a sequence generation model is obtained through training, and survey data is input into the sequence generation model to obtain data keywords, which provides a detailed keyword reference for the subsequent matching report template and ensures the accuracy and high quality of the core content of the generated text report.

[0099] The step S2 specifically includes the following steps:

[0100] Step S201, using the n data keywords to match corresponding templates in the template library according to the template keywords of each template, and taking the template with the highest keyword matching degree as the report template;

[0101] Step S202, matching the data identifier of the survey data with each placeholder in the report template according to the placeholder identifier, and replacing the successfully matched survey data with the corresponding placeholder in the report template, wherein the data identifier and the placeholder identifier are in a one-to-one relationship;

[0102] Step S203: Using the replaced report template as an evaluation data report.

[0103] The placeholder identifier refers to numbers, letters or other characters used to distinguish between various placeholders.

[0104] In this embodiment, the report template is matched according to data keywords to provide an accurate template for obtaining a standardized evaluation report, and the placeholder is replaced according to the data identifier to obtain the evaluation data report, thereby ensuring accurate and detailed data support for the evaluation report.

[0105] The step S3 specifically comprises the following steps:

[0106] Step S301, input the data keywords into the Genius model to obtain a text report;

[0107] Step S302, using the pySBD library to perform sentence processing on the text report, obtain each text sentence, and number the text sentences;

[0108] Step S303, obtaining the template number of the report template, and matching the corresponding grammar rule in the rule base according to the template number, performing grammar detection on each text sentence respectively using the grammar rule, obtaining the grammar error of the text sentence, and adjusting the text sentence according to the grammar rule;

[0109] Step S304, merging the text sentences according to the order of the numbers to obtain an evaluation text report;

[0110] The training of the Genius model includes:

[0111] A high-quality text report is obtained, keywords of the text report are extracted using the Yake method, the text report content other than the keywords is masked using the Mask method, a text draft is obtained, the text draft is input into a deep learning model, and the text draft is used as the model input and the text report is used as the model output for training until the model fit reaches the expected threshold, then the training is stopped to obtain a Genius model.

[0112] The pySBD is a Python library for automatic segmentation of paragraphs and sentences;

[0113] The Yake method is an unsupervised keyword extraction method for extracting keywords from a single document without relying on a training corpus;

[0114] The Mask method means covering the original document with a mask, thereby shielding or selecting some specific elements.

[0115] In this embodiment, the existing technology Genius model is used to obtain a text report based on keywords, the text report is segmented into sentences, and the sentences are grammatically checked using grammatical rules to obtain an evaluation text report. Obtaining the text report based on keywords saves a lot of manpower and material resources, and using grammatical rules for testing ensures the text quality of the evaluation text report.

[0116] The step S6 specifically comprises the following steps:

[0117] Step S601, the user evaluates the fitness of the report template of the final version of the evaluation report to obtain a template score, the user evaluates the sentence grammar of the text report to obtain a grammar score, and the user evaluates the core report content of the text report to obtain a keyword score;

[0118] Step S602, if the template score is less than or equal to the expected threshold, the user adjusts the template keywords of the report template;

[0119] Step S603, if the grammar score is less than or equal to the expected threshold, obtaining the grammar rule usage record of the text report, obtaining the grammar rules used in the text report according to the grammar rule usage record, and the user adjusts the grammar rules according to the grammar dislocation of the text report;

[0120] Step S604: if the keyword score is less than or equal to the expected threshold, adjust the first mapping sequence of the sequence generation model.

[0121] In this embodiment, the user performs grammar evaluation, data keyword evaluation and template evaluation on the final evaluation report, and adjusts the grammar rules, template keywords and sequence generation model according to the evaluation results, thereby improving the generation quality of the evaluation report and ensuring a good user experience.

[0122] The updating of other associated survey data according to the data identifier and data association table of the modified survey data includes:

[0123] Obtain a data identifier of the modified survey data, obtain other data identifiers associated with the data identifier in a data association table and the association relationships therebetween based on the data identifier, and modify other associated survey data based on the modified survey data and the association relationships, wherein the data association table includes data identifiers and the association relationships between each identifier.

[0124] In this embodiment, when any survey data is modified, other related survey data can be modified directly and conveniently through the data association table, thereby ensuring the convenience of data change and saving a lot of manpower and material resources.

[0125] The step S303 specifically includes the following steps:

[0126] Step S3031, dividing the text sentence into a plurality of participles by using syntactic analysis, and constructing an abstract syntax tree of the text sentence according to the relationship between the participles;

[0127] Step S3032, obtaining the template number of the report template, matching the corresponding grammar rule in the rule base according to the template number, generating the corresponding regular expression according to the grammar rule, and using the regular expression to detect the abstract syntax tree, and obtaining the text sentence corresponding to the abstract syntax tree that fails the detection;

[0128] Step S3033: Use the regular expression to adjust several word segments of the text sentence to obtain an adjusted text sentence.

[0129] The syntactic analysis refers to the process of analyzing the input text sentence to obtain the syntactic structure of the sentence;

[0130] The abstract syntax tree is a structure that represents syntax in a tree-like form;

[0131] The regular expression is a text pattern used to find, replace, verify and extract text data.

[0132] In this embodiment, syntactic analysis is used to obtain the abstract syntax tree of the text clause, and the abstract syntax tree of the clause is detected and adjusted using regular expressions of grammatical rules, thereby ensuring the grammatical standardization and text quality of the evaluation text report and expanding the scope of application of automatic report generation.

[0133] The keywords of the text report extracted by using the Yake method include:

[0134] Preprocessing the text report to remove non-semantic characters, wherein the non-semantic characters include expression symbols and numbers;

[0135] Defining keyword features of the text report, the keyword features including the position of the segmentation, the frequency of occurrence of the segmentation in the sentence, the frequency of occurrence of the segmentation in the text report and the contextual relationship of the segmentation, wherein the contextual relationship of the segmentation is determined by the number of different candidate words of the segmentation;

[0136] Obtaining the position score, sentence frequency score, text frequency score and candidate word score of each participle according to the keyword features, and obtaining the key weight of each participle according to the position score, sentence frequency score, text frequency score and candidate word score;

[0137] Arrange each word in descending order according to the key weight, remove duplicate words, extract the first m words, and define the words as keywords of the text report, where m is a natural number greater than zero.

[0138] The candidate words represent other participles on the left and right sides of the participle;

[0139] In this embodiment, the key weight of each word is obtained through the keyword features of the text report, and the top m words with the highest key weight are extracted, which provides an accurate and specific keyword basis for training the Genius model and ensures the content quality of the text report generated based on the keywords.

[0140] Example 2, reference Figure 2 , which is another embodiment of the present invention, and which is different from the first embodiment in that it provides an automatic generation system for an evaluation report, including a report generation module, a report modification module and a report evaluation module;

[0141] The report generation module is used to obtain a survey data table, obtain data keywords using a sequence generation model according to the survey data table, match a report template according to the data keywords, generate an evaluation data report and an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report;

[0142] The report modification module is used to update the generated evaluation report according to the modified survey data to obtain the final version of the evaluation report;

[0143] The report evaluation module is used by the user to perform template evaluation, grammar evaluation and data keyword evaluation on the final version of the evaluation report, obtain scoring information, and adjust the report template, grammar rules and sequence generation model according to the scoring information.

[0144] In this embodiment, a sequence generation model is used to obtain data keywords based on survey data, and the data keywords are input into the Genius model to obtain a text report. The grammar rules are used to perform grammar detection and adjustment on the text report to obtain an evaluation text report. This makes the content of the evaluation report more flexible and targeted, saves a lot of manpower and material resources, ensures the content quality of the evaluation report, and expands the scope of application of the automatic report generation method.

[0145] By matching report templates with data keywords and replacing corresponding placeholders in the report templates with survey data according to data identifiers, the efficiency of generating evaluation reports is improved. The matching of different templates also ensures the pertinence of evaluation reports and expands the scope of application of automatic report generation methods.

[0146] When the survey data needs to be modified, you only need to update any corresponding survey data, and then you can adjust other related survey data according to the data association table, and update the evaluation data report and evaluation text report in time, making the update process more convenient and simple, saving a lot of manpower and material resources.

[0147] Users conduct grammar evaluation, data keyword evaluation and template evaluation on the final version of the evaluation report, and adjust the grammar rules, template keywords and sequence generation model according to the evaluation results, which improves the generation quality of the evaluation report and ensures a good user experience.

[0148] The report generation module includes a data report generation unit and a text report generation unit;

[0149] The data report generating unit is used to obtain a survey data table, obtain data keywords according to the survey data table, match the report template with the data keywords and generate an evaluation data report;

[0150] The text report generation unit is used to input the data keywords into the Genius model to obtain a text report, and use grammatical rules to detect and adjust the text report to obtain an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report.

[0151] In this embodiment, a sequence generation model is used to obtain data keywords, match report templates and generate evaluation data reports, thereby ensuring accurate and detailed data support for the evaluation report and saving a lot of manpower and material resources. A text report is generated based on data keywords, and the text report is adjusted using grammatical rules to obtain an evaluation text report, thereby ensuring the content quality and grammatical specifications of the text report.

[0152] In this embodiment, a sequence generation model is used to obtain data keywords based on survey data, and the data keywords are input into the Genius model to obtain a text report. The grammar rules are used to perform grammar detection and adjustment on the text report to obtain an evaluation text report. This makes the content of the evaluation report more flexible and targeted, saves a lot of manpower and material resources, ensures the content quality of the evaluation report, and expands the scope of application of the automatic report generation method.

[0153] By matching report templates with data keywords and replacing corresponding placeholders in the report templates with survey data according to data identifiers, the efficiency of generating evaluation reports is improved. The matching of different templates also ensures the pertinence of evaluation reports and expands the scope of application of automatic report generation methods.

[0154] When the survey data needs to be modified, you only need to update any corresponding survey data, and then you can adjust other related survey data according to the data association table, and update the evaluation data report and evaluation text report in time, making the update process more convenient and simple, saving a lot of manpower and material resources.

[0155] Users conduct grammar evaluation, data keyword evaluation and template evaluation on the final version of the evaluation report, and adjust the grammar rules, template keywords and sequence generation model according to the evaluation results, which improves the generation quality of the evaluation report and ensures a good user experience.

[0156] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for automatically generating an evaluation report, characterized in that: The following steps are involved: Step S1, obtaining a survey data table, inputting the survey data table into a pre-trained sequence generation model, and obtaining data keywords; Step S2, matching the report template in the template library according to the data keyword, and inputting the survey data in the survey data table into the corresponding position of the data template in the report template according to the data identifier, to generate an evaluation data report; Step S3, inputting the data keywords into the Genius model to obtain a text report, and according to the corresponding grammatical rules in the report template matching rule library, performing grammatical correction on the text report using the grammatical rules to obtain an evaluation text report; Step S4, merging the evaluation data report and the evaluation text report to obtain an evaluation report; Step S5, if any survey data is modified, update other associated survey data according to the data identifier and data association table of the modified survey data, obtain an updated survey data table, and repeat steps S1 to S3 according to the updated survey data table to obtain a final version of the evaluation report; Step S6, the user performs template evaluation, grammar evaluation and data keyword evaluation on the final version of the evaluation report to obtain scoring information, and adjusts the report template, grammar rules and sequence generation model according to the scoring information, wherein the scoring information includes template scoring, grammar scoring and keyword scoring; The step S1 specifically includes the following steps: Step S101, obtaining a survey data table; Step S102, inputting the survey data in the survey data table into an encoder of a pre-trained sequence generation model, and encoding the survey data into a feature vector of a fixed size according to a first mapping sequence; Step S103, inputting the feature vector into the decoder of the sequence generation model, obtaining the keywords corresponding to each survey data according to the second mapping sequence, and obtaining the weight of each keyword by using the self-attention mechanism; Step S104, sorting the keywords from large to small according to their weights, and extracting the first n keywords as data keywords of the survey data, where n is a natural number greater than zero; Training a sequence generation model involves: Obtain a survey data set, perform data preprocessing on the survey data set, divide the preprocessed survey data set into a survey data training set and a survey data test set according to a certain ratio, perform data labeling on the survey data training set, input the labeled survey data training set into the first bidirectional recurrent neural network of the training generation model, train with the survey data as input and the data features of the data labeling as output, and test the trained first bidirectional recurrent neural network using the survey data test set until the fit reaches an expected threshold, then stop training, obtain the encoder of the sequence generation model and generate a first mapping sequence; Acquire data features of the data annotation, perform keyword annotation on the data features, obtain data keywords, train a second bidirectional recurrent neural network using the data features as input and the data keywords as output until the fit reaches an expected threshold, then stop training, obtain a decoder of the sequence generation model and generate a second mapping sequence; The step S3 specifically comprises the following steps: Step S301, input the data keywords into the Genius model to obtain a text report; Step S302, using the pySBD library to perform sentence processing on the text report, obtain each text sentence, and number the text sentences; Step S303, obtaining the template number of the report template, and matching the corresponding grammar rule in the rule base according to the template number, performing grammar detection on each text sentence respectively using the grammar rule, obtaining the grammar error of the text sentence, and adjusting the text sentence according to the grammar rule; Step S304, merging the text sentences according to the sequence of the numbers to obtain an evaluation text report; The training of the Genius model includes: A high-quality text report is obtained, keywords of the text report are extracted using the Yake method, the text report content other than the keywords is masked using the Mask method, a text draft is obtained, the text draft is input into a deep learning model, and the text draft is used as the model input and the text report is used as the model output for training until the model fit reaches the expected threshold, then the training is stopped to obtain a Genius model.

2. The method for automatically generating an evaluation report according to claim 1, wherein: The step S2 specifically includes the following steps: Step S201, using the n data keywords to match corresponding templates in the template library according to the template keywords of each template, and taking the template with the highest keyword matching degree as the report template; Step S202, matching the data identifier of the survey data with each placeholder in the report template according to the placeholder identifier, and replacing the successfully matched survey data with the corresponding placeholder in the report template, wherein the data identifier and the placeholder identifier are in a one-to-one correspondence; Step S203: Using the replaced report template as an evaluation data report.

3. The method for automatically generating an evaluation report according to claim 1, wherein: The step S6 specifically comprises the following steps: Step S601, the user evaluates the fitness of the report template of the final version of the evaluation report to obtain a template score, the user evaluates the sentence grammar of the text report to obtain a grammar score, and the user evaluates the core report content of the text report to obtain a keyword score; Step S602, if the template score is less than or equal to the expected threshold, the user adjusts the template keywords of the report template; Step S603, if the grammar score is less than or equal to the expected threshold, obtaining the grammar rule usage record of the text report, obtaining the grammar rules used in the text report according to the grammar rule usage record, and the user adjusts the grammar rules according to the grammar dislocation of the text report; Step S604: if the keyword score is less than or equal to the expected threshold, adjust the first mapping sequence of the sequence generation model.

4. The method for automatically generating an evaluation report according to claim 1, wherein: The updating of other associated survey data according to the data identifier and data association table of the modified survey data includes: Obtain a data identifier of the modified survey data, obtain other data identifiers associated with the data identifier in a data association table and the association relationships therebetween based on the data identifier, and modify other associated survey data based on the modified survey data and the association relationships, wherein the data association table includes data identifiers and the association relationships between each identifier.

5. The method for automatically generating an evaluation report according to claim 4, characterized in that: The step S303 specifically includes the following steps: Step S3031, dividing the text sentence into a plurality of participles by using syntactic analysis, and constructing an abstract syntax tree of the text sentence according to the relationship between the participles; Step S3032, obtaining the template number of the report template, matching the corresponding grammar rule in the rule base according to the template number, generating the corresponding regular expression according to the grammar rule, and using the regular expression to detect the abstract syntax tree, and obtaining the text sentence corresponding to the abstract syntax tree that fails the detection; Step S3033: Use the regular expression to adjust several word segments of the text sentence to obtain an adjusted text sentence.

6. The method for automatically generating an evaluation report according to claim 4, characterized in that: The keywords of the text report extracted by using the Yake method include: Preprocessing the text report to remove non-semantic characters, wherein the non-semantic characters include expression symbols and numbers; Defining keyword features of the text report, the keyword features including the position of the segmentation, the frequency of occurrence of the segmentation in the sentence, the frequency of occurrence of the segmentation in the text report and the contextual relationship of the segmentation, wherein the contextual relationship of the segmentation is determined by the number of different candidate words of the segmentation; Obtaining the position score, sentence frequency score, text frequency score and candidate word score of each participle according to the keyword features, and obtaining the key weight of each participle according to the position score, sentence frequency score, text frequency score and candidate word score; Arrange each word in descending order according to the key weight, remove duplicate words, extract the first m words, and define the words as keywords of the text report, where m is a natural number greater than zero.

7. An automatic generation system based on the automatic generation method of the evaluation report according to claim 1, characterized in that: It includes report generation module, report modification module and report evaluation module; The report generation module is used to obtain a survey data table, obtain data keywords using a sequence generation model according to the survey data table, match a report template according to the data keywords, generate an evaluation data report and an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report; The report modification module is used to update the generated evaluation report according to the modified survey data to obtain the final version of the evaluation report; The report evaluation module is used for the user to perform template evaluation, grammar evaluation and data keyword evaluation on the final version of the evaluation report, obtain scoring information, and adjust the report template, grammar rules and sequence generation model according to the scoring information; The report generation module includes a data report generation unit and a text report generation unit; The data report generating unit is used to obtain a survey data table, obtain data keywords according to the survey data table, match the report template with the data keywords and generate an evaluation data report; The text report generation unit is used to input the data keywords into the Genius model to obtain a text report, and use grammatical rules to detect and adjust the text report to obtain an evaluation text report, and merge the evaluation data report and the evaluation text report to obtain an evaluation report.

Images