Requirement document quality detection method, device, equipment and storage medium

Abstract

The present invention relates to artificial intelligence, and provides a method, device, equipment and storage medium for detecting the quality of demand documents. The method extracts text information from a document to be detected, and counts the amount of information. If the amount of information is greater than or equal to a preset word count threshold, the text information is divided to obtain background description text, domain description text, system description text and detail description text, calculate the text similarity between the background description text and the domain description text, match the system description text with the preset system name, detect the detail description text based on a text classification prediction model, obtain a probability vector, and accurately generate a quality score based on the text similarity, matching degree and probability vector. In addition, the present invention also relates to blockchain technology, and the quality score can be stored in the blockchain.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and storage medium for detecting quality of demand documents. Background Art

[0002] At present, the requirement documents are usually written by the requirement analysts after communicating with the business personnel. However, due to the limitation of the writing ability of the requirement analysts, the quality of the requirement documents written by the requirement analysts cannot be guaranteed, which leads to low R&D efficiency.

[0003] Therefore, how to accurately perform quality inspection on requirement documents has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of the above, it is necessary to provide a method, device, equipment and storage medium for quality detection of requirement documents, which can solve the technical problem of how to accurately perform quality detection on requirement documents.

[0005] In one aspect, the present invention provides a method for detecting quality of a requirement document, the method comprising:

[0006] When a quality inspection request is received, extracting text information from the document to be inspected according to the quality inspection request;

[0007] Counting the number of pieces of text information;

[0008] If the amount of information is greater than or equal to a preset word count threshold, the text information is divided to obtain a background description text, a domain description text, a system description text, and a detail description text;

[0009] Calculating the text similarity between the background description text and the field description text;

[0010] Matching the system description text with a preset system name to obtain a matching degree;

[0011] Detecting the detailed description text based on a pre-trained text classification prediction model, and obtaining a probability vector output by the text classification prediction model;

[0012] A quality score of the document to be detected is generated according to the text similarity, the matching degree and the probability vector.

[0013] According to a preferred embodiment of the present invention, extracting text information from the document to be detected according to the quality detection request includes:

[0014] Parsing the document storage path and document identifier from the quality inspection request;

[0015] Locating the document to be detected from the document storage path based on the document identifier;

[0016] Identifying a document format based on the attribute information of the document to be detected;

[0017] Obtaining an encoding method corresponding to the document format;

[0018] The text information is read from the document to be detected based on the encoding method.

[0019] According to a preferred embodiment of the present invention, the text information is divided to obtain background description text, field description text, system description text and detail description text, including:

[0020] Extracting multiple template tags from the text information;

[0021] Locating the label position of each template label in the document to be detected;

[0022] Determine the text information corresponding to the template tag whose tag position is the first preset position as the background description text, and determine the text information corresponding to the template tag whose tag position is the second preset position as the field description text;

[0023] The text information corresponding to the template tag whose tag position is the third preset position is determined as the system description text, and the text information corresponding to the template tag whose tag position is the fourth preset position is determined as the detail description text.

[0024] According to a preferred embodiment of the present invention, the calculating of the text similarity between the background description text and the field description text comprises:

[0025] Performing word segmentation processing on the background description text to obtain a plurality of initial background words, and performing word segmentation processing on the domain description text to obtain a plurality of initial domain words;

[0026] Eliminating stop words from the multiple initial background words based on a preset stop word list to obtain multiple target background words, and eliminating stop words from the multiple initial domain words based on the preset stop word list to obtain multiple target domain words;

[0027] Performing vectorization processing on each target background word to obtain a plurality of first vectors, and calculating an average value of the plurality of first vectors in each vector dimension to obtain a background vector;

[0028] Performing vectorization processing on each target domain vocabulary to obtain multiple second vectors, and calculating an average value of the multiple second vectors in each vector dimension to obtain a domain vector;

[0029] The text similarity is calculated based on the background vector and the domain vector. The calculation formula of the text similarity is:

[0030]

[0031] Wherein, sim(a,b) represents the text similarity, a represents the background vector, b represents the domain vector, n represents the total number of dimensions of the vector dimension, a i represents the i-th vector dimension in the background vector, b i Represents the i-th vector dimension in the domain vector.

[0032] According to a preferred embodiment of the present invention, matching the system description text with a preset system name to obtain a matching degree includes:

[0033] Traversing multiple system-to-be-tested vocabulary in the system description text, and locating the vocabulary position of each system-to-be-tested vocabulary in the system description text;

[0034] Selecting a preset system name that matches each system vocabulary and corresponding vocabulary position to be tested as a preliminary system name;

[0035] Counting the number of words in the preliminary selected system names that match the multiple system words to be tested as the number of matches, and counting the total number of words in the multiple system words to be tested;

[0036] The ratio of the maximum matching quantity to the total vocabulary is calculated as the matching degree.

[0037] According to a preferred embodiment of the present invention, the text classification prediction model includes an input network, an encoding network and a prediction network, and the detection of the detail description text based on the pre-trained text classification prediction model and obtaining the probability vector output by the text classification prediction model include:

[0038] Performing vectorization processing on the detailed description text based on the input network to obtain a text vector;

[0039] Acquire multiple encoding matrices from the encoding network;

[0040] Calculate the product of each encoding matrix and the text vector to obtain a plurality of initial vectors, wherein the plurality of initial vectors include a third vector, a fourth vector and a fifth vector;

[0041] Calculate the product of the transposed vector of the third vector and the fourth vector to obtain the correlation between any two detail words in the detail description text;

[0042] Normalizing the relevance to obtain encoding weights of any two detail words in the detail description text;

[0043] Calculating the product of the encoding weight and the fifth vector to obtain an encoding vector of the detailed description text;

[0044] Performing prediction processing on the encoding vector based on the prediction network to obtain a prediction vector;

[0045] A vector corresponding to the first prediction label is extracted from the prediction vector as the probability vector.

[0046] According to a preferred embodiment of the present invention, generating the quality score of the document to be detected according to the text similarity, the matching degree and the probability vector includes:

[0047] Calculating a first score corresponding to the text similarity based on a first preset formula, and calculating a second score corresponding to the matching degree based on a second preset formula;

[0048] Mapping the probability vector based on the vector score mapping table to obtain a third score;

[0049] A weighted sum operation is performed on the first score, the second score, and the third score to obtain the quality score.

[0050] On the other hand, the present invention further provides a demand document quality detection device, the demand document quality detection device comprising:

[0051] An extraction unit, configured to extract text information from a document to be detected according to a quality detection request when a quality detection request is received;

[0052] A calculation unit, used for counting the amount of information in the text information;

[0053] A division unit, configured to divide the text information into background description text, domain description text, system description text and detail description text if the amount of the information is greater than or equal to a preset word count threshold;

[0054] A calculation unit, used for calculating the text similarity between the background description text and the field description text;

[0055] A matching unit, used for matching the system description text with a preset system name to obtain a matching degree;

[0056] A detection unit, used to detect the detail description text based on a pre-trained text classification prediction model, and obtain a probability vector output by the text classification prediction model;

[0057] A generating unit is used to generate a quality score of the document to be detected according to the text similarity, the matching degree and the probability vector.

[0058] On the other hand, the present invention further provides an electronic device, comprising:

[0059] a memory storing computer readable instructions; and

[0060] The processor executes the computer-readable instructions stored in the memory to implement the demand document quality detection method.

[0061] On the other hand, the present invention further proposes a computer-readable storage medium, in which computer-readable instructions are stored. The computer-readable instructions are executed by a processor in an electronic device to implement the requirement document quality detection method.

[0062] It can be seen from the above technical solutions that the present application can preliminarily detect whether the document to be detected meets the basic requirements by analyzing the amount of information in the text information. Then, on the premise that the document to be detected meets the basic requirements, since the background description text and the field description text are mutually echoing information, and the detailed description text should contain clearly described business processes and business rules, therefore, by combining the text similarity between the background description text and the field description text, the matching degree between the system description text and the preset system name, and the probability vector predicted by the text classification prediction model for the detailed description text, the quality of the document to be detected can be analyzed in combination with the characteristics of the demand document, thereby improving the accuracy of the quality score. The present application can ensure that the information in the demand document is clear, understandable, and has a pattern to follow by performing quality inspection on the document to be detected, so that R&D personnel can quickly understand the needs of the business, reduce communication costs, and greatly improve R&D efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of a preferred embodiment of the method for detecting quality of demand documents of the present invention.

[0064] Figure 2 It is a functional module diagram of a preferred embodiment of the demand document quality detection device of the present invention.

[0065] Figure 3 It is a structural schematic diagram of an electronic device of a preferred embodiment of the present invention for implementing the method for detecting quality of demand documents. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0067] like Figure 1 FIG. 1 is a flowchart of a preferred embodiment of the method for detecting quality of demand documents of the present invention. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0068] The requirement document quality detection method can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0069] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0070] The requirement document quality detection method is applied to one or more electronic devices, which are devices that can automatically perform numerical calculations and / or information processing according to pre-set or stored computer-readable instructions, and whose hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), programmable gate arrays (FPGA), digital signal processors (DSP), embedded devices, etc.

[0071] The electronic device may be any electronic product that can perform human-computer interaction with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0072] The electronic device may include a network device and / or a user device. The network device includes, but is not limited to, a single network electronic device, an electronic device group consisting of multiple network electronic devices, or a cloud consisting of a large number of hosts or network electronic devices based on cloud computing.

[0073] The network where the electronic device is located includes, but is not limited to: the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0074] S10: When a quality inspection request is received, text information is extracted from the document to be inspected according to the quality inspection request.

[0075] In at least one embodiment of the present invention, the quality inspection request may be generated by triggering a user responsible for demand document inspection, and the quality inspection request may also be generated when receiving upload feedback of the document to be inspected.

[0076] The document to be detected may include a technical requirement document, a product requirement document, etc. The document format of the document to be detected may be in a variety of formats, for example, the document format of the document to be detected may be in excel, word, txt, etc.

[0077] The text information refers to the information stored in the document to be detected.

[0078] In at least one embodiment of the present invention, the electronic device extracting text information from the document to be detected according to the quality detection request includes:

[0079] Parsing the document storage path and document identifier from the quality inspection request;

[0080] Locating the document to be detected from the document storage path based on the document identifier;

[0081] Identifying a document format based on the attribute information of the document to be detected;

[0082] Obtaining an encoding method corresponding to the document format;

[0083] The text information is read from the document to be detected based on the encoding method.

[0084] The document storage path refers to a path for storing the required document, and the document storage path may also refer to an upload address that triggers the generation of the quality inspection request.

[0085] The document identifier refers to an identification code that uniquely identifies the document to be detected.

[0086] The encoding method refers to the method of writing the text information in the document to be detected.

[0087] By combining the document storage path and document identification, the document to be detected can be accurately located, and the document format can be quickly identified based on the attribute information, thereby improving the reading efficiency of the text information. At the same time, reading the text information based on the encoding method corresponding to the document format can improve the reading accuracy of the text information.

[0088] S11, counting the information quantity of the text information.

[0089] In at least one embodiment of the present invention, the information quantity refers to the total number of words in the text information.

[0090] In at least one embodiment of the present invention, after counting the information quantity of the text information, the method further includes:

[0091] Comparing the information quantity with a preset word count threshold;

[0092] If the amount of information is less than the preset word count threshold, a first warning message is generated based on the amount of information.

[0093] The preset word count threshold may be set according to requirements, and the preset word count threshold may also be determined according to the average word count of qualified requirement documents.

[0094] The first warning information is used to indicate that the document to be detected is unqualified in terms of word count and that the document to be detected needs to be revised and rewritten.

[0095] Through the above implementation, when the amount of information is less than the preset word count threshold, the first warning information can be generated to indicate that the document to be detected is unqualified in terms of word count.

[0096] S12, if the amount of information is greater than or equal to the preset word count threshold, the text information is divided to obtain background description text, field description text, system description text and detail description text.

[0097] In at least one embodiment of the present invention, the background description text mainly introduces the source of demand and the business status and problems.

[0098] The field description text mainly introduces the specific field information corresponding to the project.

[0099] The system description text mainly introduces the system information of the specific application required.

[0100] The detailed description text mainly introduces the business process and business rules.

[0101] In at least one embodiment of the present invention, the electronic device divides the text information to obtain the background description text, the field description text, the system description text and the detail description text, including:

[0102] Extracting multiple template tags from the text information;

[0103] Locating the label position of each template label in the document to be detected;

[0104] Determine the text information corresponding to the template tag whose tag position is the first preset position as the background description text, and determine the text information corresponding to the template tag whose tag position is the second preset position as the field description text;

[0105] The text information corresponding to the template tag whose tag position is the third preset position is determined as the system description text, and the text information corresponding to the template tag whose tag position is the fourth preset position is determined as the detail description text.

[0106] The multiple template tags include, but are not limited to: background tags, domain tags, system tags, and detail tags.

[0107] The first preset position refers to the position of the background tag in the document to be detected, the second preset position is the position of the field tag in the document to be detected, the third preset position refers to the position of the system tag in the document to be detected, and the fourth preset position is the position of the details tag in the document to be detected.

[0108] By identifying the label position of each template label in the document to be detected, the background description text, the field description text, the system description text and the detail description text can be accurately extracted by utilizing the correspondence between the label position and the first preset position, the second preset position, the third preset position and the fourth preset position.

[0109] S13, calculating the text similarity between the background description text and the field description text.

[0110] In at least one embodiment of the present invention, the text similarity is used to indicate the matching status of the background description text and the field description text. The higher the text similarity, the higher the quality score of the document to be detected.

[0111] In at least one embodiment of the present invention, the electronic device calculating the text similarity between the background description text and the field description text includes:

[0112] Performing word segmentation processing on the background description text to obtain a plurality of initial background words, and performing word segmentation processing on the domain description text to obtain a plurality of initial domain words;

[0113] Eliminating stop words from the multiple initial background words based on a preset stop word list to obtain multiple target background words, and eliminating stop words from the multiple initial domain words based on the preset stop word list to obtain multiple target domain words;

[0114] Performing vectorization processing on each target background word to obtain a plurality of first vectors, and calculating an average value of the plurality of first vectors in each vector dimension to obtain a background vector;

[0115] Performing vectorization processing on each target domain vocabulary to obtain multiple second vectors, and calculating an average value of the multiple second vectors in each vector dimension to obtain a domain vector;

[0116] The text similarity is calculated based on the background vector and the domain vector. The calculation formula of the text similarity is:

[0117]

[0118] Wherein, sim(a,b) represents the text similarity, a represents the background vector, b represents the domain vector, n represents the total number of dimensions of the vector dimension, a i represents the i-th vector dimension in the background vector, b i Represents the i-th vector dimension in the domain vector.

[0119] The preset stop word list may be a stop word list directly obtained from the Internet.

[0120] The preset stop word list can eliminate the interference of stop words on the background description text and the domain description text, and by calculating the average value of the first vector of each target background vocabulary in each vector dimension, it can be ensured that each vector dimension in the background vector contains information about each target background vocabulary, and by calculating the average value of the second vector of each target domain vocabulary in each vector dimension, it can be ensured that each vector dimension in the domain vector contains information about each target domain vocabulary, thereby doubly improving the accuracy of the text similarity.

[0121] S14, matching the system description text with a preset system name to obtain a matching degree.

[0122] In at least one embodiment of the present invention, the matching degree refers to the matching situation between the system used by the document to be detected and the existing system name. The higher the matching degree, the higher the quality score of the document to be detected.

[0123] The preset system name refers to an existing system name.

[0124] In at least one embodiment of the present invention, the electronic device matches the system description text with a preset system name to obtain a matching degree, including:

[0125] Traversing multiple system-to-be-tested vocabulary in the system description text, and locating the vocabulary position of each system-to-be-tested vocabulary in the system description text;

[0126] Selecting a preset system name that matches each system vocabulary and corresponding vocabulary position to be tested as a preliminary system name;

[0127] Counting the number of words in the preliminary selected system names that match the multiple system words to be tested as the number of matches, and counting the total number of words in the multiple system words to be tested;

[0128] The ratio of the maximum matching quantity to the total vocabulary is calculated as the matching degree.

[0129] For example, the system description text is ABCD, and the preset system names include the name of system 1: ACDF and the name of system 2: ABEF. Since the name of system 1 contains A which is the same as A and the corresponding vocabulary position in the system description text, and the name of system 2 contains A and B which are the same as A and B and the corresponding vocabulary positions in the system description text, the name of system 1 and the name of system 2 are both the preliminary system names. According to statistics, since the number of matches corresponding to the name of system 2 is 2, which is greater than the number of matches corresponding to the name of system 1, the matching degree is 2 / 4=0.5.

[0130] By detecting the preset system name in combination with each system vocabulary to be tested and the corresponding vocabulary position, the preliminary system name can be accurately screened out and the number of matches can be accurately counted, thereby improving the accuracy of the matching degree. At the same time, since only the number of matches in the preliminary system name needs to be counted, the efficiency of determining the matching degree can be improved.

[0131] S15, detecting the detail description text based on a pre-trained text classification prediction model, and obtaining a probability vector output by the text classification prediction model.

[0132] In at least one embodiment of the present invention, the text classification prediction model includes an input network, an encoding network and a prediction network, wherein the input network is used to convert the detailed description text into information that can be recognized by the encoding network. The encoding network is used to extract feature information in the text vector output by the input network. The prediction network is used to predict the probability value of the encoding vector output by the encoding network on the first prediction label and the second prediction label. The first prediction label can be a label representing good quality, and the second prediction label can be a label representing poor quality.

[0133] In at least one embodiment of the present invention, the electronic device detects the detail description text based on a pre-trained text classification prediction model, and obtains the probability vector output by the text classification prediction model, including:

[0134] Performing vectorization processing on the detailed description text based on the input network to obtain a text vector;

[0135] Acquire multiple encoding matrices from the encoding network;

[0136] Calculate the product of each encoding matrix and the text vector to obtain a plurality of initial vectors, wherein the plurality of initial vectors include a third vector, a fourth vector and a fifth vector;

[0137] Calculate the product of the transposed vector of the third vector and the fourth vector to obtain the correlation between any two detail words in the detail description text;

[0138] Normalizing the relevance to obtain encoding weights of any two detail words in the detail description text;

[0139] Calculating the product of the encoding weight and the fifth vector to obtain an encoding vector of the detailed description text;

[0140] Performing prediction processing on the encoding vector based on the prediction network to obtain a prediction vector;

[0141] A vector corresponding to the first prediction label is extracted from the prediction vector as the probability vector.

[0142] The multiple encoding matrices are matrix information corresponding to when the text classification prediction model is trained.

[0143] By calculating the product of the transposed vector of the third vector and the fourth vector, the correlation between any two detail words in the detail description text can be accurately identified, and then the detail description text can be re-encoded using the correlation, which can improve the representation ability of the encoding vector and thus improve the accuracy of the probability vector.

[0144] S16, generating a quality score of the document to be detected according to the text similarity, the matching degree and the probability vector.

[0145] It should be emphasized that in order to further ensure the privacy and security of the above quality scores, the above quality scores can also be stored in a blockchain node.

[0146] In at least one embodiment of the present invention, the quality score refers to the score obtained after the document to be detected undergoes intelligent quality detection. The quality score is used to measure the quality of the document to be detected.

[0147] In at least one embodiment of the present invention, the electronic device generates the quality score of the document to be detected according to the text similarity, the matching degree and the probability vector, including:

[0148] Calculating a first score corresponding to the text similarity based on a first preset formula, and calculating a second score corresponding to the matching degree based on a second preset formula;

[0149] Mapping the probability vector based on the vector score mapping table to obtain a third score;

[0150] A weighted sum operation is performed on the first score, the second score, and the third score to obtain the quality score.

[0151] Among them, the first preset formula refers to the relationship between the text similarity and the score, and the text similarity and the score are in direct proportion. Therefore, the first preset formula can be y=kx, where k is a constant greater than 1. The first preset formula can also be y=kx+b, where k is a constant greater than 1 and b is a constant.

[0152] The second preset formula refers to the relationship between the matching degree and the score, and the matching degree and the score are also in direct proportion.

[0153] The vector-score mapping table stores a mapping relationship between vectors and scores.

[0154] The quality score can be accurately detected by combining the text similarity, the matching degree and the probability vector.

[0155] In at least one embodiment of the present invention, after generating the quality score of the document to be detected according to the text similarity, the matching degree and the probability vector, the method further includes:

[0156] If the quality score is less than a preset score threshold, identifying the abnormal description text in the document to be detected according to the text similarity, the matching degree and the probability vector;

[0157] Generate second warning information according to the abnormal description text.

[0158] The abnormal description text may refer to the module text with the lowest score value. For example, if the second score corresponding to the matching degree is the lowest, then the system description text is the abnormal description text.

[0159] The second warning information is used to remind the user of the module text that needs to be rewritten.

[0160] Through the above implementation, the abnormal description text can be accurately identified, thereby reminding the user of the module text that needs to be rewritten, avoiding the user from rewriting the entire document to be detected, and improving the user experience.

[0161] It can be seen from the above technical solutions that the present application can preliminarily detect whether the document to be detected meets the basic requirements by analyzing the amount of information in the text information. Then, on the premise that the document to be detected meets the basic requirements, since the background description text and the field description text are mutually echoing information, and the detailed description text should contain clearly described business processes and business rules, therefore, by combining the text similarity between the background description text and the field description text, the matching degree between the system description text and the preset system name, and the probability vector predicted by the text classification prediction model for the detailed description text, the quality of the document to be detected can be analyzed in combination with the characteristics of the demand document, thereby improving the accuracy of the quality score. The present application can ensure that the information in the demand document is clear, understandable, and has a pattern to follow by performing quality inspection on the document to be detected, so that R&D personnel can quickly understand the needs of the business, reduce communication costs, and greatly improve R&D efficiency.

[0162] like Figure 2 , which is a functional module diagram of a preferred embodiment of the demand document quality detection device of the present invention. The demand document quality detection device 11 includes an extraction unit 110, a calculation unit 111, a division unit 112, a matching unit 113, a detection unit 114, a generation unit 115, a comparison unit 116 and an identification unit 117. The module / unit referred to in the present invention refers to a series of computer-readable instruction segments that can be acquired by the processor 13 and can complete a fixed function, which is stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0163] When receiving the quality detection request, the extraction unit 110 extracts text information from the document to be detected according to the quality detection request.

[0164] In at least one embodiment of the present invention, the quality inspection request may be generated by triggering a user responsible for demand document inspection, and the quality inspection request may also be generated when receiving upload feedback of the document to be inspected.

[0165] The document to be detected may include a technical requirement document, a product requirement document, etc. The document format of the document to be detected may be in a variety of formats, for example, the document format of the document to be detected may be in excel, word, txt, etc.

[0166] The text information refers to the information stored in the document to be detected.

[0167] In at least one embodiment of the present invention, the extracting unit 110 extracts text information from the document to be detected according to the quality detection request, including:

[0168] Parsing the document storage path and document identifier from the quality inspection request;

[0169] Locating the document to be detected from the document storage path based on the document identifier;

[0170] Identifying a document format based on the attribute information of the document to be detected;

[0171] Obtaining an encoding method corresponding to the document format;

[0172] The text information is read from the document to be detected based on the encoding method.

[0173] The document storage path refers to a path for storing the required document, and the document storage path may also refer to an upload address that triggers the generation of the quality inspection request.

[0174] The document identifier refers to an identification code that uniquely identifies the document to be detected.

[0175] The encoding method refers to the method of writing the text information in the document to be detected.

[0176] By combining the document storage path and document identification, the document to be detected can be accurately located, and the document format can be quickly identified based on the attribute information, thereby improving the reading efficiency of the text information. At the same time, reading the text information based on the encoding method corresponding to the document format can improve the reading accuracy of the text information.

[0177] The calculation unit 111 counts the information quantity of the text information.

[0178] In at least one embodiment of the present invention, the information quantity refers to the total number of words in the text information.

[0179] In at least one embodiment of the present invention, after counting the information quantity of the text information, the comparison unit 116 compares the information quantity with a preset word count threshold;

[0180] If the number of messages is less than the preset word count threshold, the generating unit 115 generates first warning information based on the number of messages.

[0181] The preset word count threshold may be set according to requirements, and the preset word count threshold may also be determined according to the average word count of qualified requirement documents.

[0182] The first warning information is used to indicate that the document to be detected is unqualified in terms of word count and that the document to be detected needs to be revised and rewritten.

[0183] Through the above implementation, when the amount of information is less than the preset word count threshold, the first warning information can be generated to indicate that the document to be detected is unqualified in terms of word count.

[0184] If the amount of information is greater than or equal to the preset word count threshold, the dividing unit 112 divides the text information into background description text, domain description text, system description text and detail description text.

[0185] In at least one embodiment of the present invention, the background description text mainly introduces the source of demand and the business status and problems.

[0186] The field description text mainly introduces the specific field information corresponding to the project.

[0187] The system description text mainly introduces the system information of the specific application required.

[0188] The detailed description text mainly introduces the business process and business rules.

[0189] In at least one embodiment of the present invention, the dividing unit 112 divides the text information to obtain the background description text, the field description text, the system description text and the detail description text, including:

[0190] Extracting multiple template tags from the text information;

[0191] Locating the label position of each template label in the document to be detected;

[0192] Determine the text information corresponding to the template tag whose tag position is the first preset position as the background description text, and determine the text information corresponding to the template tag whose tag position is the second preset position as the field description text;

[0193] The text information corresponding to the template tag whose tag position is the third preset position is determined as the system description text, and the text information corresponding to the template tag whose tag position is the fourth preset position is determined as the detail description text.

[0194] The multiple template tags include, but are not limited to: background tags, domain tags, system tags, and detail tags.

[0195] The first preset position refers to the position of the background tag in the document to be detected, the second preset position is the position of the field tag in the document to be detected, the third preset position refers to the position of the system tag in the document to be detected, and the fourth preset position is the position of the details tag in the document to be detected.

[0196] By identifying the label position of each template label in the document to be detected, the background description text, the field description text, the system description text and the detail description text can be accurately extracted by utilizing the correspondence between the label position and the first preset position, the second preset position, the third preset position and the fourth preset position.

[0197] The calculation unit 111 calculates the text similarity between the background description text and the field description text.

[0198] In at least one embodiment of the present invention, the text similarity is used to indicate the matching status of the background description text and the field description text. The higher the text similarity, the higher the quality score of the document to be detected.

[0199] In at least one embodiment of the present invention, the calculating unit 111 calculates the text similarity between the background description text and the field description text, including:

[0200] Performing word segmentation processing on the background description text to obtain a plurality of initial background words, and performing word segmentation processing on the domain description text to obtain a plurality of initial domain words;

[0201] Eliminating stop words from the multiple initial background words based on a preset stop word list to obtain multiple target background words, and eliminating stop words from the multiple initial domain words based on the preset stop word list to obtain multiple target domain words;

[0202] Performing vectorization processing on each target background word to obtain a plurality of first vectors, and calculating an average value of the plurality of first vectors in each vector dimension to obtain a background vector;

[0203] Performing vectorization processing on each target domain vocabulary to obtain multiple second vectors, and calculating an average value of the multiple second vectors in each vector dimension to obtain a domain vector;

[0204] The text similarity is calculated based on the background vector and the domain vector. The calculation formula of the text similarity is:

[0205]

[0206] Wherein, sim(a,b) represents the text similarity, a represents the background vector, b represents the domain vector, n represents the total number of dimensions of the vector dimension, a i represents the i-th vector dimension in the background vector, b i Represents the i-th vector dimension in the domain vector.

[0207] The preset stop word list may be a stop word list directly obtained from the Internet.

[0208] The preset stop word list can eliminate the interference of stop words on the background description text and the domain description text, and by calculating the average value of the first vector of each target background vocabulary in each vector dimension, it can be ensured that each vector dimension in the background vector contains information about each target background vocabulary, and by calculating the average value of the second vector of each target domain vocabulary in each vector dimension, it can be ensured that each vector dimension in the domain vector contains information about each target domain vocabulary, thereby doubly improving the accuracy of the text similarity.

[0209] The matching unit 113 matches the system description text with the preset system name to obtain a matching degree.

[0210] In at least one embodiment of the present invention, the matching degree refers to the matching situation between the system used by the document to be detected and the existing system name. The higher the matching degree, the higher the quality score of the document to be detected.

[0211] The preset system name refers to an existing system name.

[0212] In at least one embodiment of the present invention, the matching unit 113 matches the system description text with the preset system name to obtain a matching degree including:

[0213] Traversing multiple system-to-be-tested vocabulary in the system description text, and locating the vocabulary position of each system-to-be-tested vocabulary in the system description text;

[0214] Selecting a preset system name that matches each system vocabulary and corresponding vocabulary position to be tested as a preliminary system name;

[0215] Counting the number of words in the preliminary selected system names that match the multiple system words to be tested as the number of matches, and counting the total number of words in the multiple system words to be tested;

[0216] The ratio of the maximum matching quantity to the total vocabulary is calculated as the matching degree.

[0217] For example, the system description text is ABCD, and the preset system names include the name of system 1: ACDF and the name of system 2: ABEF. Since the name of system 1 contains A which is the same as A and the corresponding vocabulary position in the system description text, and the name of system 2 contains A and B which are the same as A and B and the corresponding vocabulary positions in the system description text, the name of system 1 and the name of system 2 are both the preliminary system names. According to statistics, since the number of matches corresponding to the name of system 2 is 2, which is greater than the number of matches corresponding to the name of system 1, the matching degree is 2 / 4=0.5.

[0218] By detecting the preset system name in combination with each system vocabulary to be tested and the corresponding vocabulary position, the preliminary system name can be accurately screened out and the number of matches can be accurately counted, thereby improving the accuracy of the matching degree. At the same time, since only the number of matches in the preliminary system name needs to be counted, the efficiency of determining the matching degree can be improved.

[0219] The detection unit 114 detects the detail description text based on a pre-trained text classification prediction model, and obtains a probability vector output by the text classification prediction model.

[0220] In at least one embodiment of the present invention, the text classification prediction model includes an input network, an encoding network and a prediction network, wherein the input network is used to convert the detailed description text into information that can be recognized by the encoding network. The encoding network is used to extract feature information in the text vector output by the input network. The prediction network is used to predict the probability value of the encoding vector output by the encoding network on the first prediction label and the second prediction label. The first prediction label can be a label representing good quality, and the second prediction label can be a label representing poor quality.

[0221] In at least one embodiment of the present invention, the detection unit 114 detects the detail description text based on a pre-trained text classification prediction model, and obtains the probability vector output by the text classification prediction model, including:

[0222] Performing vectorization processing on the detailed description text based on the input network to obtain a text vector;

[0223] Acquire multiple encoding matrices from the encoding network;

[0224] Calculate the product of each encoding matrix and the text vector to obtain a plurality of initial vectors, wherein the plurality of initial vectors include a third vector, a fourth vector and a fifth vector;

[0225] Calculate the product of the transposed vector of the third vector and the fourth vector to obtain the correlation between any two detail words in the detail description text;

[0226] Normalizing the relevance to obtain encoding weights of any two detail words in the detail description text;

[0227] Calculating the product of the encoding weight and the fifth vector to obtain an encoding vector of the detailed description text;

[0228] Performing prediction processing on the encoding vector based on the prediction network to obtain a prediction vector;

[0229] A vector corresponding to the first prediction label is extracted from the prediction vector as the probability vector.

[0230] The multiple encoding matrices are matrix information corresponding to when the text classification prediction model is trained.

[0231] By calculating the product of the transposed vector of the third vector and the fourth vector, the correlation between any two detail words in the detail description text can be accurately identified, and then the detail description text can be re-encoded using the correlation, which can improve the representation ability of the encoding vector and thus improve the accuracy of the probability vector.

[0232] The generating unit 115 generates a quality score of the document to be detected according to the text similarity, the matching degree and the probability vector.

[0233] It should be emphasized that in order to further ensure the privacy and security of the above quality scores, the above quality scores can also be stored in a blockchain node.

[0234] In at least one embodiment of the present invention, the quality score refers to the score obtained after the document to be detected undergoes intelligent quality detection. The quality score is used to measure the quality of the document to be detected.

[0235] In at least one embodiment of the present invention, the generating unit 115 generates the quality score of the document to be detected according to the text similarity, the matching degree and the probability vector, including:

[0236] Calculating a first score corresponding to the text similarity based on a first preset formula, and calculating a second score corresponding to the matching degree based on a second preset formula;

[0237] Mapping the probability vector based on the vector score mapping table to obtain a third score;

[0238] A weighted sum operation is performed on the first score, the second score, and the third score to obtain the quality score.

[0239] Among them, the first preset formula refers to the relationship between the text similarity and the score, and the text similarity and the score are in direct proportion. Therefore, the first preset formula can be y=kx, where k is a constant greater than 1. The first preset formula can also be y=kx+b, where k is a constant greater than 1 and b is a constant.

[0240] The second preset formula refers to the relationship between the matching degree and the score, and the matching degree and the score are also in direct proportion.

[0241] The vector-score mapping table stores a mapping relationship between vectors and scores.

[0242] The quality score can be accurately detected by combining the text similarity, the matching degree and the probability vector.

[0243] In at least one embodiment of the present invention, after generating the quality score of the document to be detected according to the text similarity, the matching degree and the probability vector, if the quality score is less than a preset score threshold, the recognition unit 117 recognizes the abnormal description text in the document to be detected according to the text similarity, the matching degree and the probability vector;

[0244] The generating unit 115 generates second warning information according to the abnormal description text.

[0245] The abnormal description text may refer to the module text with the lowest score value. For example, if the second score corresponding to the matching degree is the lowest, then the system description text is the abnormal description text.

[0246] The second warning information is used to remind the user of the module text that needs to be rewritten.

[0247] Through the above implementation, the abnormal description text can be accurately identified, thereby reminding the user of the module text that needs to be rewritten, avoiding the user from rewriting the entire document to be detected, and improving the user experience.

[0248] It can be seen from the above technical solutions that the present application can preliminarily detect whether the document to be detected meets the basic requirements by analyzing the amount of information in the text information. Then, on the premise that the document to be detected meets the basic requirements, since the background description text and the field description text are mutually echoing information, and the detailed description text should contain clearly described business processes and business rules, therefore, by combining the text similarity between the background description text and the field description text, the matching degree between the system description text and the preset system name, and the probability vector predicted by the text classification prediction model for the detailed description text, the quality of the document to be detected can be analyzed in combination with the characteristics of the demand document, thereby improving the accuracy of the quality score. The present application can ensure that the information in the demand document is clear, understandable, and has a pattern to follow by performing quality inspection on the document to be detected, so that R&D personnel can quickly understand the needs of the business, reduce communication costs, and greatly improve R&D efficiency.

[0249] like Figure 3 , which is a schematic diagram of the structure of an electronic device of a preferred embodiment of the method for detecting quality of demand documents according to the present invention.

[0250] In one embodiment of the present invention, the electronic device 1 includes, but is not limited to, a memory 12, a processor 13, and computer-readable instructions stored in the memory 12 and executable on the processor 13, such as a requirement document quality detection program.

[0251] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. The electronic device 1 may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the electronic device 1 may also include input and output devices, network access devices, buses, etc.

[0252] The processor 13 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 13 is the computing core and control center of the electronic device 1, and uses various interfaces and lines to connect various parts of the entire electronic device 1, and executes the operating system of the electronic device 1 and various installed applications, program codes, etc.

[0253] Exemplarily, the computer-readable instructions may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present invention. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer-readable instructions in the electronic device 1. For example, the computer-readable instructions may be divided into an extraction unit 110, a calculation unit 111, a division unit 112, a matching unit 113, a detection unit 114, a generation unit 115, a comparison unit 116, and an identification unit 117.

[0254] The memory 12 can be used to store the computer-readable instructions and / or modules, and the processor 13 implements various functions of the electronic device 1 by running or executing the computer-readable instructions and / or modules stored in the memory 12, and calling the data stored in the memory 12. The memory 12 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the electronic device, etc. The memory 12 can include non-volatile and volatile memories, such as: a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other storage devices.

[0255] The memory 12 may be an external memory and / or an internal memory of the electronic device 1. Furthermore, the memory 12 may be a memory in a physical form, such as a memory stick, a TF card (Trans-flash Card), and the like.

[0256] If the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also instruct the relevant hardware to complete it through computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium. When the computer-readable instructions are executed by the processor, the steps of the above-mentioned various method embodiments can be implemented.

[0257] The computer-readable instructions include computer-readable instruction codes, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer-readable instruction code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM).

[0258] The blockchain referred to in the present invention is a new application mode of computer technologies such as distributed demand document quality detection, point-to-point transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, platform product service layer, and application service layer.

[0259] Combination Figure 1 The memory 12 in the electronic device 1 stores computer-readable instructions to implement a method for detecting quality of a demand document, and the processor 13 can execute the computer-readable instructions to implement:

[0260] When a quality inspection request is received, extracting text information from the document to be inspected according to the quality inspection request;

[0261] Counting the number of pieces of text information;

[0262] If the amount of information is greater than or equal to a preset word count threshold, the text information is divided to obtain a background description text, a domain description text, a system description text, and a detail description text;

[0263] Calculating the text similarity between the background description text and the field description text;

[0264] Matching the system description text with a preset system name to obtain a matching degree;

[0265] Detecting the detailed description text based on a pre-trained text classification prediction model, and obtaining a probability vector output by the text classification prediction model;

[0266] A quality score of the document to be detected is generated according to the text similarity, the matching degree and the probability vector.

[0267] Specifically, the specific implementation method of the processor 13 for the above-mentioned computer readable instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0268] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0269] The computer-readable storage medium stores computer-readable instructions, wherein the computer-readable instructions are used to implement the following steps when executed by the processor 13:

[0270] When a quality inspection request is received, extracting text information from the document to be inspected according to the quality inspection request;

[0271] Counting the number of pieces of text information;

[0272] If the amount of information is greater than or equal to a preset word count threshold, the text information is divided to obtain a background description text, a domain description text, a system description text, and a detail description text;

[0273] Calculating the text similarity between the background description text and the field description text;

[0274] Matching the system description text with a preset system name to obtain a matching degree;

[0275] Detecting the detailed description text based on a pre-trained text classification prediction model, and obtaining a probability vector output by the text classification prediction model;

[0276] A quality score of the document to be detected is generated according to the text similarity, the matching degree and the probability vector.

[0277] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0278] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0279] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0280] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices mentioned can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0281] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution 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 solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for detecting quality of a requirement document, characterized in that: The requirement document quality detection method comprises: When a quality inspection request is received, extracting text information from the document to be inspected according to the quality inspection request; Counting the number of pieces of text information; If the amount of information is greater than or equal to a preset word count threshold, the text information is divided to obtain a background description text, a domain description text, a system description text and a detail description text, including: extracting a plurality of template tags from the text information; locating the tag position of each template tag in the document to be detected; determining the text information corresponding to the template tag whose tag position is a first preset position as the background description text, and determining the text information corresponding to the template tag whose tag position is a second preset position as the domain description text; determining the text information corresponding to the template tag whose tag position is a third preset position as the system description text, and determining the text information corresponding to the template tag whose tag position is a fourth preset position as the detail description text; Calculating the text similarity between the background description text and the field description text; Matching the system description text with a preset system name to obtain a matching degree; Detecting the detailed description text based on a pre-trained text classification prediction model, and obtaining a probability vector output by the text classification prediction model; The quality score of the document to be detected is generated according to the text similarity, the matching degree and the probability vector, including: calculating a first score corresponding to the text similarity based on a first preset formula, and calculating a second score corresponding to the matching degree based on a second preset formula; mapping the probability vector based on a vector score mapping table to obtain a third score; and performing a weighted sum operation on the first score, the second score and the third score to obtain the quality score.

2. The method for detecting quality of demand documents as claimed in claim 1, characterized in that: The extracting text information from the document to be detected according to the quality detection request comprises: Parsing the document storage path and document identifier from the quality inspection request; Locating the document to be detected from the document storage path based on the document identifier; Identifying a document format based on the attribute information of the document to be detected; Obtaining an encoding method corresponding to the document format; The text information is read from the document to be detected based on the encoding method.

3. The method for detecting quality of demand documents as claimed in claim 1, characterized in that: The calculating the text similarity between the background description text and the field description text comprises: Performing word segmentation processing on the background description text to obtain a plurality of initial background words, and performing word segmentation processing on the domain description text to obtain a plurality of initial domain words; Eliminating stop words from the multiple initial background words based on a preset stop word list to obtain multiple target background words, and eliminating stop words from the multiple initial domain words based on the preset stop word list to obtain multiple target domain words; Performing vectorization processing on each target background word to obtain a plurality of first vectors, and calculating an average value of the plurality of first vectors in each vector dimension to obtain a background vector; Performing vectorization processing on each target domain vocabulary to obtain multiple second vectors, and calculating an average value of the multiple second vectors in each vector dimension to obtain a domain vector; The text similarity is calculated based on the background vector and the domain vector. The calculation formula of the text similarity is: ； in, represents the text similarity, represents the background vector, represents the field vector, represents the total number of dimensions of the vector, Represents the background vector vector dimensions, Represents the first vector dimensions.

4. The method for detecting quality of demand documents as claimed in claim 1, characterized in that: The matching of the system description text with the preset system name to obtain the matching degree includes: Traversing multiple system-to-be-tested vocabulary in the system description text, and locating the vocabulary position of each system-to-be-tested vocabulary in the system description text; Selecting a preset system name that matches each system vocabulary and corresponding vocabulary position to be tested as a preliminary system name; Counting the number of words in the preliminary selected system names that match the multiple system words to be tested as the number of matches, and counting the total number of words in the multiple system words to be tested; The ratio of the maximum matching quantity to the total vocabulary is calculated as the matching degree.

5. The method for detecting quality of demand documents as claimed in claim 1, characterized in that: The text classification prediction model includes an input network, an encoding network and a prediction network. The detection of the detail description text based on the pre-trained text classification prediction model and obtaining the probability vector output by the text classification prediction model include: Performing vectorization processing on the detailed description text based on the input network to obtain a text vector; Acquire multiple encoding matrices from the encoding network; Calculate the product of each encoding matrix and the text vector to obtain a plurality of initial vectors, wherein the plurality of initial vectors include a third vector, a fourth vector and a fifth vector; Calculate the product of the transposed vector of the third vector and the fourth vector to obtain the correlation between any two detail words in the detail description text; Normalizing the relevance to obtain encoding weights of any two detail words in the detail description text; Calculating the product of the encoding weight and the fifth vector to obtain an encoding vector of the detailed description text; Performing prediction processing on the encoding vector based on the prediction network to obtain a prediction vector; A vector corresponding to the first prediction label is extracted from the prediction vector as the probability vector.

6. A device for detecting quality of demand documents, characterized in that: The demand document quality detection device comprises: An extraction unit, configured to extract text information from a document to be detected according to a quality detection request when a quality detection request is received; A calculation unit, used for counting the amount of information in the text information; A division unit, for dividing the text information if the amount of the information is greater than or equal to a preset word count threshold, to obtain a background description text, a domain description text, a system description text and a detail description text, including: extracting a plurality of template tags from the text information; locating the tag position of each template tag in the document to be detected; determining the text information corresponding to the template tag with the tag position being a first preset position as the background description text, and determining the text information corresponding to the template tag with the tag position being a second preset position as the domain description text; determining the text information corresponding to the template tag with the tag position being a third preset position as the system description text, and determining the text information corresponding to the template tag with the tag position being a fourth preset position as the detail description text; A calculation unit, used for calculating the text similarity between the background description text and the field description text; A matching unit, used for matching the system description text with a preset system name to obtain a matching degree; A detection unit, used to detect the detail description text based on a pre-trained text classification prediction model, and obtain a probability vector output by the text classification prediction model; A generating unit is used to generate a quality score of the document to be detected according to the text similarity, the matching degree and the probability vector, including: calculating a first score corresponding to the text similarity based on a first preset formula, and calculating a second score corresponding to the matching degree based on a second preset formula; mapping the probability vector based on a vector score mapping table to obtain a third score; performing a weighted sum operation on the first score, the second score and the third score to obtain the quality score.

7. An electronic device, characterized in that: The electronic device comprises: a memory storing computer-readable instructions; and A processor executes computer-readable instructions stored in the memory to implement the method for detecting quality of demand documents as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions are executed by a processor in an electronic device to implement the demand document quality detection method as described in any one of claims 1 to 5.

Images