An information processing method and device, and a storage medium

Abstract

The embodiment of the application discloses an information processing method and device, and a storage medium, comprising: in the case that the to-be-tested business information of a to-be-tested object is acquired, inputting the to-be-tested business information into a decay conversion coding model to obtain to-be-tested coding information; acquiring to-be-diffused coding information corresponding to a to-be-diffused object; diffusing the target information corresponding to the to-be-tested coding information from the to-be-diffused coding information by using a diffusion model; determining a target object corresponding to the target information from the to-be-diffused object, and taking the target object as a related object related to the to-be-tested object.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and in particular to an information processing method, apparatus, and storage medium. Background Technology

[0002] In various businesses within the internet finance sector, it is often necessary to identify companies in the same industry or sector as the target company from a small sample of target enterprises. Target companies can be high-performing companies in a particular business category, or they can be high-quality potential clients for developing new businesses or exploring new sectors. Based on this sample of target companies, similar companies in the same sector can be identified to facilitate related business development.

[0003] In related technologies, existing classifications, such as national standard classifications at various levels, are used to build models by combining industry business data and the company's own data to identify peer companies. However, many companies do not have national standard classification values, and there is often overlap between different categories that do have classification values. Furthermore, some companies have broad business scopes, making it difficult to accurately classify them using a single classification attribute. Additionally, a company's main business scope may change over time, and national standard classifications cannot provide timely updates. Therefore, the accuracy of peer companies found based on national standard classifications is low, thus reducing the accuracy of identifying relevant targets. Summary of the Invention

[0004] To address the aforementioned technical problems, embodiments of this application aim to provide an information processing method, apparatus, and storage medium that can improve the accuracy of identifying relevant objects.

[0005] The technical solution of this application is implemented as follows:

[0006] This application provides an information processing method, the information processing method including:

[0007] Once the test operation information of the test object is obtained, the test operation information is input into the attenuation conversion coding model to obtain the test coding information;

[0008] Obtain the diffusion encoding information corresponding to the object to be diffused;

[0009] Using a diffusion model, target information corresponding to the coding information to be tested is diffused from the coding information to be diffused.

[0010] The target object corresponding to the target information is determined from the objects to be diffused, and the target object is regarded as a related object related to the object to be tested.

[0011] This application provides an information processing apparatus, the apparatus comprising:

[0012] The input unit is used to input the test operation information into the attenuation conversion coding model to obtain the test coding information when the test operation information of the test object is obtained.

[0013] The acquisition unit is used to acquire the diffusion encoding information corresponding to the object to be diffused;

[0014] A diffusion unit is used to diffuse target information corresponding to the coding information to be tested from the coding information to be diffused using a diffusion model.

[0015] The determining unit is used to determine the target object corresponding to the target information from the object to be diffused, and to regard the target object as a related object related to the object to be tested.

[0016] This application provides an information processing apparatus, the apparatus comprising:

[0017] The system includes a memory, a processor, and a communication bus. The memory communicates with the processor via the communication bus. The memory stores information processing programs executable by the processor. When the information processing programs are executed, the processor performs the information processing method described above.

[0018] This application provides a storage medium storing a computer program for use in an information processing device, characterized in that the computer program, when executed by a processor, implements the information processing method described above.

[0019] This application provides an information processing method, apparatus, and storage medium. The information processing method includes: upon obtaining the business information to be tested of a test object, inputting the business information to be tested into an attenuation conversion coding model to obtain the code information to be tested; obtaining the code information to be diffused corresponding to a target object; using a diffusion model to diffuse target information corresponding to the code information to be tested from the code information to be diffused; determining the target object corresponding to the target information from the target objects to be diffused, and using the target object as a related object associated with the test object. Using the above method, the information processing apparatus uses an attenuation conversion coding model to perform attenuation conversion coding on the business information to be tested of the test object to obtain the code information to be tested; uses a diffusion model to diffuse target information corresponding to the code information to be diffused, thereby determining the target object corresponding to the target information from the target objects to be diffused, i.e., obtaining the related object associated with the test object. This improves the accuracy of determining related objects, even when the business information to be tested of the test object changes over time, by directly using the attenuation conversion coding model and the diffusion model based on the changed business information of the test object. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of an information processing method provided in an embodiment of this application;

[0021] Figure 2 An exemplary schematic diagram of business information to be tested is provided for an embodiment of this application;

[0022] Figure 3 This is an exemplary schematic diagram of the business scope information to be tested, provided as an embodiment of this application.

[0023] Figure 4 This application provides an exemplary preprocessed intent of the business scope information to be tested, as shown in the embodiments of this application.

[0024] Figure 5 This is a schematic diagram of an exemplary information processing structure provided in an embodiment of this application;

[0025] Figure 6 This application provides an exemplary schematic diagram of the object scope.

[0026] Figure 7 An exemplary national standard classification diagram provided for embodiments of this application. Figure 1 ;

[0027] Figure 8 An exemplary national standard classification diagram provided for embodiments of this application. Figure 2 ;

[0028] Figure 9 An exemplary national standard classification ratio diagram provided for embodiments of this application;

[0029] Figure 10 A schematic diagram illustrating the average distance ratio of adjacent categories under a national standard level 4 classification, provided for embodiments of this application;

[0030] Figure 11 An exemplary image aggregation diagram provided for an embodiment of this application;

[0031] Figure 12 A flowchart illustrating an exemplary information processing method provided in this application embodiment;

[0032] Figure 13 A schematic diagram of the composition structure of an information processing device provided in this application embodiment. Figure 1 ;

[0033] Figure 14 A schematic diagram of the composition structure of an information processing device provided in this application embodiment. Figure 2 . Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0035] This application provides an information processing method, which is applied to an information processing device. Figure 1 A flowchart of an information processing method provided in an embodiment of this application is shown below. Figure 1 As shown, information processing methods may include:

[0036] S101. After obtaining the test operation information of the test object, input the test operation information into the attenuation conversion coding model to obtain the test coding information.

[0037] The information processing method provided in this application embodiment is applicable to scenarios where related objects are identified in relation to the object under test.

[0038] In the embodiments of this application, the information processing device can be implemented in various forms. For example, the information processing device described in this application may include devices such as mobile phones, cameras, tablet computers, laptops, handheld computers, personal digital assistants (PDAs), portable media players (PMPs), navigation devices, wearable devices, smart bracelets, pedometers, etc., as well as devices such as digital TVs, desktop computers, servers, etc.

[0039] In this embodiment of the application, the test object can be an enterprise, a product, or other items; the specific test object can be determined according to the actual situation, and this embodiment of the application does not limit it.

[0040] In this embodiment of the application, if the object to be tested is an enterprise, the business information to be tested may include the business scope information and the business product information of the object to be tested; specifically, the business product information may be the main product information of the object to be tested.

[0041] In this embodiment, the attenuation conversion coding model can be a model configured in the information processing device; the attenuation conversion coding model can also be a model transmitted to the information processing device by other devices; the specific way in which the information processing device obtains the attenuation conversion coding model can be determined according to the actual situation, and this embodiment does not limit it.

[0042] In this embodiment, the attenuation conversion coding model can be the Ernie-decay model; the attenuation conversion coding model can also be other models that can process the business information to be tested to obtain the coding information to be tested; the specific attenuation conversion coding model can be determined according to the actual situation, and this embodiment does not limit it.

[0043] In this embodiment, the encoding information to be tested can be numerical information or character information; the encoding information to be tested can also be other forms of information; the specific encoding information to be tested can be determined according to the actual situation, and this embodiment does not limit it.

[0044] For example, if the encoded information to be tested is numerical information, then the encoded information to be tested can specifically be numerical information in matrix form.

[0045] In this embodiment of the application, the number of objects to be tested can be multiple; the specific number of objects to be tested can be determined according to the actual situation, and this embodiment of the application does not limit this.

[0046] In this embodiment of the application, the number of business information to be tested is multiple; the specific number of business information to be tested can be determined according to the actual situation, and this embodiment of the application does not limit this.

[0047] It should be noted that the number of business information to be tested and the number of objects to be tested are in one-to-one correspondence, that is, one object to be tested corresponds to one piece of business information to be tested.

[0048] It should also be noted that a set of business information to be tested includes a set of business scope information and a set of business product information.

[0049] In this embodiment of the application, the process of the information processing device inputting the business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested includes: the information processing device preprocessing the business information to be tested to obtain the preprocessed business information to be tested; and the information processing device inputting the preprocessed business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested.

[0050] It should be noted that preprocessing can be noise reduction, i.e., removing punctuation marks or special characters from the business information to be tested, or removing characters that do not contain information from the business information to be tested; preprocessing can be other processing methods; the specific preprocessing method can be determined according to the actual situation, and this application embodiment does not limit it.

[0051] In this embodiment of the application, the process of the information processing device preprocessing the business information to be tested to obtain the preprocessed business information to be tested includes: the information processing device determining the words to be removed from the business information to be tested that match a preset word list; and the information processing device removing the words to be removed from the business information to be tested to obtain the preprocessed business information to be tested.

[0052] In this embodiment, the preset vocabulary can be a vocabulary configured in the information processing device; the preset vocabulary can also be a vocabulary transmitted to the information processing device from other devices; the preset vocabulary can also be a vocabulary obtained by the information processing device in other ways; the specific way in which the information processing device obtains the preset vocabulary can be determined according to the actual situation, and this embodiment does not limit it.

[0053] In this embodiment, if the business information to be tested includes the business scope information of the object to be tested (business scope information to be tested) and the business product information of the object to be tested (business product information to be tested), then the preset term list also includes a preset business scope term list and a preset business product term list. The information processing device can use the preset business scope term list to determine business scope removal terms from the business scope information to be tested, and use the preset business product term list to determine business product removal terms from the business product information to be tested, and use the business scope removal terms and business product removal terms as terms to be removed.

[0054] In this embodiment, the information processing device removes "product removal words" from the product information to be tested to obtain preprocessed product information to be tested; the information processing device removes "scope removal words" from the business scope information to be tested to obtain preprocessed business scope information to be tested; and the preprocessed business scope information and the preprocessed product information to be tested are used as preprocessed business information to be tested.

[0055] For example, there are 5 test targets (enterprises). Information on their business scope (the business scope information to be tested) and main products (the products to be tested) is collected according to their enterprise ID and enterprise name. Figure 2The diagram shows that the business scope of Company A Limited, with Company ID 1, includes: investment. (II) Establishing research and development centers or departments in China to engage in the research and development of new products and high technologies, transferring its research and development results, and providing corresponding technical services. (III) Authorized in writing by its invested companies (with unanimous approval from the board of directors…); Main product information includes: with optical technology as the core, covering a wide range of fields such as imaging system products, office products and industrial equipment; The business scope of Company ID 2 and Company Name B Integrated Circuit Co., Ltd. includes: production of new electronic components (power electronic devices); research and development of new electronic components (power electronic devices), power electronic devices and integrated circuit products and systems; sales of self-produced products; providing technical consultation, technical services, and technology transfer for the above products…; Main product information includes: committed to providing advanced semiconductor products and system solutions for the three major technological challenges of modern society - high energy efficiency, mobility and security; The business scope of Company ID 3 and Company Name C Semiconductor Technology Co., Ltd. includes: general items: research, development, production and sales of high-quality semiconductor silicon wafers, silicon-based semiconductor materials, semiconductor equipment and components, technology development, technology transfer, technology consultation, and technical services related to semiconductor materials and devices, import and export of goods, and import and export of technology; Main product information includes: committed to Research and development of complete sets of industrialized mass production equipment for 300mm silicon single crystal growth, silicon wafer processing, epitaxial wafer preparation, and silicon wafer analysis and testing applicable to the 40-28nm node. Company ID 4, Company Name: D Electronic Components Co., Ltd. Business scope includes: general items: wholesale, import and export, commission agency (excluding auctions) of electronic products, electrical products and their components, and other related supporting businesses; warehousing (excluding dangerous goods) and distribution of electronic products; engaging in the technical field of electrical and electronic products; main product information includes: engaging in the technical field of electrical and electronic products. Company ID 5, Company Name: E Lithography Equipment Technology Co., Ltd. Business scope includes: research and development, maintenance, and improvement of lithography equipment systems; development of related technologies and provision of corresponding technical services; wholesale, import and export, commission agency (excluding auctions) of lithography equipment and its components, and provision of after-sales service and other related supporting services; enterprise management consulting; business information consulting (excluding financial information); main product information includes: research and development, maintenance, and improvement of lithography equipment systems; development of related technologies and provision of related technical services.

[0056] For example, there are three entities to be tested, whose business scope information is as follows: Figure 3The first test subject's business scope information is as follows: General items: software development; software sales; integrated circuit design; sales of integrated circuit chips and products; information system integration services; integrated circuit chip and product manufacturing; electronic component manufacturing; electronic component retail; technical services, technical development, technical consulting, technical exchange, technology transfer, and technology promotion. (Except for projects that require approval by law, business activities can be carried out independently in accordance with the law based on the business license.) Licensed items: import and export of goods. (For projects that require approval by law, business activities can only be carried out after approval by relevant departments, and the specific business items are subject to the approval results.) The second test subject's business scope information is as follows: integrated circuit technology development, technology transfer, technical consulting, and technical services; computer system integration; computer animation design; computer system services; application software services; wholesale of security technology prevention products, computers, software and auxiliary equipment, electronic products, and mechanical equipment; technology import and export, goods import and export, and agency import and export. (Projects requiring approval by law shall be operated in accordance with the approved content after approval by relevant departments.) The business scope information of the third test subject is as follows: General items: integrated circuit design; integrated circuit sales; integrated circuit chip design and services; integrated circuit chip and product sales; software development; software sales; electronic component manufacturing; electronic component wholesale; electronic component retail; electronic product sales; optoelectronic device manufacturing; optoelectronic device sales; communication equipment manufacturing; communication equipment sales (except for projects requiring approval by law, business activities may be carried out independently in accordance with the law based on the business license). Licensed items: import and export of goods; import and export of technology (projects requiring approval by law may be operated only after approval by relevant departments, and the specific business items shall be subject to the approval documents or licenses issued by relevant departments). The information processing device uses a preset business scope term list to remove business scope removal words from the business scope information to be tested, and obtains the preprocessed business scope information to be tested, such as... Figure 4The business scope information for the first test subject is as follows: software development, software sales, integrated circuit design, integrated circuit chip and product sales, information system integration services, integrated circuit chip and product manufacturing, electronic component manufacturing, electronic component retail, technical services, technical development, technical consulting, technical exchange, technology transfer, technology promotion, and import and export of goods. The business scope information for the second test subject is as follows: integrated circuit technology transfer, technical consulting, technical services, computer system integration, computer animation design, computer system services, application software services, wholesale, security technology and prevention products, computer software and auxiliary equipment, electronic products, machinery and equipment, technology import and export, and import and export agency. The business scope information for the third test subject is as follows: integrated circuit design, integrated circuit sales, integrated circuit chip design and services, integrated circuit chip and product sales, software development, software sales, electronic component manufacturing, electronic component wholesale, electronic component retail, electronic product sales, optoelectronic device manufacturing, optoelectronic device sales, communication equipment manufacturing, communication equipment sales, import and export of goods, and import and export of technology.

[0057] It should be noted that the preset vocabulary is generated using unsupervised word segmentation Unigram. The preset vocabulary includes words such as "general items:", "must be approved by law xxxx", and "subject to approval or licenses from relevant departments".

[0058] Understandably, by using a corpus of business scope and product information and generating a vocabulary using Unigrams to segment sentences into word blocks with the highest probability of occurrence, the product removal words and business scope removal words in the business scope information to be tested are completely removed. In other words, meaningless word groups in the business information to be tested are removed, which improves the accuracy of the preprocessed business information to be tested. The highly accurate preprocessed business information to be tested is then used to identify highly accurate relevant objects.

[0059] In this embodiment of the application, the process by which the information processing device inputs the preprocessed business information to be tested into the attenuation conversion coding model to obtain the code information to be tested includes: the information processing device using the function family in the attenuation conversion coding model to convert each character in the preprocessed business information to be tested into a numerical value, obtaining multiple sets of numerical values; the information processing device determining the attenuation coefficient corresponding to each character according to the character attenuation parameter in the attenuation conversion coding model, obtaining multiple attenuation coefficients; and the information processing device determining the code information to be tested based on the multiple attenuation coefficients and the multiple sets of numerical values.

[0060] In this embodiment of the application, the amount of information carried by the later text information in the preprocessed business information to be tested is relatively small. Without discarding information, the information weight of the earlier text information in the preprocessed business information to be tested can be increased by the attenuation conversion method of the attenuation conversion coding model, so as to improve the accuracy of determining the coding information to be tested.

[0061] In this embodiment, a preprocessed business information to be tested includes a set of preprocessed business scope information and a set of preprocessed business product information. The information processing device uses the function family in the attenuation conversion coding model to convert each character in the preprocessed business scope information into a numerical value, obtaining a set of business scope numerical values. The device also uses the function family in the attenuation conversion coding model to convert each character in the preprocessed business product information into a numerical value, obtaining a set of business product numerical values. The information processing device uses the set of business scope numerical values ​​and the set of business product numerical values ​​as a set of numerical values. When there are multiple sets of preprocessed business information to be tested, multiple sets of numerical values ​​are obtained.

[0062] It should be noted that there are multiple functions in a function family; the specific number of functions in a function family can be determined according to the actual situation, and this application does not limit this.

[0063] In this embodiment of the application, a set of preprocessed business scope information to be tested can be X string ={x i}, i∈{1,2,3…,m}, where i is the i-th character in the preprocessed business scope information to be tested; the function family is Using the function family in the attenuation conversion coding model, each character in a preprocessed set of business scope information to be tested is converted into a numerical value, resulting in a set of business scope numerical values.

[0064]

[0065] It should be noted that the vectorization expansion is as follows: By using the family of functions in formula (1), each character in the preprocessed business information to be tested is converted into a numerical value, that is, the preprocessed business information to be tested is mapped to the numerical space.

[0066] It should be noted that if there are multiple sets of preprocessed business scope information to be tested, the length of each text vector in the multiple sets of preprocessed business scope information to be tested is the same, which is m.

[0067] In this embodiment of the application, the information processing device determines the attenuation coefficient corresponding to each character based on the character attenuation parameter in the attenuation conversion coding model, and obtains multiple attenuation coefficients, as shown in formula (2):

[0068] G(i)=e -α(i+l) ,i∈{1,2,3…,m}——(2)

[0069] It should be noted that α and l represent the initial and final states of attenuation, respectively. By determining the attenuation coefficient corresponding to each character, the equal importance of each character in the processed business scope information to be tested is avoided. The information weight of the first text information in the preprocessed business information to be tested is increased to improve the accuracy of determining the encoded information to be tested. In formula (2), G(i) is the attenuation coefficient corresponding to each character, and α and l are the attenuation conversion coding model, which can be adjusted through the model training process.

[0070] In this embodiment of the application, the process by which the information processing device determines the encoded information to be tested based on multiple attenuation coefficients and multiple sets of values ​​is as shown in formula (3):

[0071]

[0072] It should be noted that, For in matrix The averaging operation of the corresponding index positions (element-wise) of matrix G, for a set of business scope values, is performed using formula (3). After the operation, a matrix with dimensions m * embed_size can be obtained. The operation yields a vector of length m; the specific details are shown in formula (4):

[0073]

[0074] It should be noted that (4) is the formula for calculating the mean. Its purpose is to take the mean of the first value of each character in a set of preprocessed business scope information to be tested, then take the mean of the second value of each character in a set of preprocessed business scope information to be tested, and so on, until the mean of the last value of each character in a set of preprocessed business scope information to be tested, thereby reducing the data dimension of the obtained code information to be tested to 1*embed_size, thus improving the calculation speed.

[0075] For example, 1 million companies correspond to 1 million sets of preprocessed business scope information to be tested. The 1 million sets of preprocessed business scope information to be tested are converted into numerical values, and the attenuation coefficient corresponding to each character is determined (embed_size = 768, each character has 768 attenuation coefficients). The data of the business scope to be tested encoded based on the attenuation coefficients and numerical values ​​is 1 million numerical vectors of length 768, that is, a matrix of size 1 million * 768.

[0076] It should be noted that the value of embed_size can be 768, 512, or 256, and the specific value can be determined according to the actual situation. This application embodiment does not limit this.

[0077] For example, the preprocessed business information of Company H to be tested is: X string =“Microelectronics semiconductor products semi-finished products materials equipment R&D manufacturing R&D installation testing...trade agency commission agency enterprise marketing planning business consulting enterprise management consulting financial consulting” totaling 256 characters. Through the numerical conversion part in the attenuation conversion coding model, i.e. formula (1), a vector of a set of business scope values ​​is obtained as follows Each column is 768 characters long, representing a single character; there are a total of 256 columns for 'v'. As you can see, X... string In this text, all the important information is concentrated at the beginning, while the end, such as "enterprise marketing planning, business consulting, enterprise management consulting, and financial consulting," is quite different from the company's main information. Using formula (2), where α = 0.08 and l = 0.0, the attenuation weight (attenuation coefficient) of each word is obtained as G[1.0 0.92268 0.85133…1.8418e-14 1.6994e-14]. Each value in G is multiplied by each column in the vector v of a set of business scope values ​​to obtain the business scope code to be tested. From formula (4), we can calculate...

[0078] For example, such as Figure 5 As shown: The process by which the information processing device inputs the preprocessed business information to be tested into the attenuation conversion coding model (Emie-decay architecture) to obtain the coded information to be tested includes: the information processing device using the function family in the attenuation conversion coding model The preprocessed business information to be tested (input text X) will be processed separately. string Each character in the vector is converted into a numerical value, resulting in multiple sets of numerical values ​​(intermediate vectors). The information processing device determines the attenuation coefficient corresponding to each character based on the character attenuation parameters in the attenuation conversion coding model, and obtains multiple attenuation coefficients (G(.)). The information processing device determines the coding information to be tested (output vector) based on the multiple attenuation coefficients and multiple sets of values.

[0079] S102. Obtain the diffusion encoding information corresponding to the object to be diffused.

[0080] In this embodiment, the information processing device inputs the business information to be tested into the attenuation conversion coding model to obtain the code information to be tested. After obtaining the code information to be tested, the information processing device can then acquire the code information to be diffused corresponding to the object to be diffused. Alternatively, the information processing device can first acquire the code information to be diffused corresponding to the object to be diffused, and then input the business information to be tested into the attenuation conversion coding model to obtain the code information to be tested. The information processing device can also acquire the code information to be diffused corresponding to the object to be diffused while inputting the business information to be tested into the attenuation conversion coding model to obtain the code information to be tested. The specific method can be determined according to the actual situation, and this embodiment does not limit it.

[0081] In this embodiment of the application, the number of objects to be diffused can be multiple; the specific number of objects to be diffused can be determined according to the actual situation, and this embodiment of the application does not limit this. For example, the number of objects to be diffused can be 1 million.

[0082] In this embodiment of the application, the number of encoded information to be diffused is the same as the number of objects to be diffused.

[0083] In this embodiment of the application, the information to be diffused is also the information obtained by inputting the information to be diffused of the target object into the attenuation conversion coding model.

[0084] In this embodiment of the application, the object to be disseminated can be an enterprise, a commodity, or other items. The specific object to be disseminated can be determined according to the actual situation, and this embodiment of the application does not limit this.

[0085] It should be noted that if the target of dissemination is a company, then the target can be all existing companies collected. The business information to be disseminated includes the target company's business scope information and its product information. Specifically, the product information can be the target company's main product information.

[0086] S103. Using a diffusion model, diffuse the target information corresponding to the coding information to be tested from the coding information to be diffused.

[0087] In this embodiment of the application, after the information processing device obtains the coding information to be diffused corresponding to the object to be diffused, the information processing device can use the diffusion model to diffuse the target information corresponding to the coding information to be tested from the coding information to be diffused.

[0088] In this embodiment, the diffusion model can be a model configured in the information processing device; the diffusion model can also be a model transmitted to the information processing device by other devices; the diffusion model can also be a model obtained by the information processing device in other ways; the specific way in which the information processing device obtains the diffusion model can be determined according to the actual situation, and this embodiment does not limit it.

[0089] In this embodiment, the coding information to be diffused includes the business scope code to be diffused and the business product code to be diffused; the coding information to be tested includes the business scope code to be tested and the business product code to be tested; the process by which the information processing device diffuses the target information corresponding to the coding information to be tested from the coding information to be diffused using a diffusion model includes: the information processing device inputs the business scope code to be tested and the business product code to be tested into a fusion model to obtain first fusion information; the information processing device inputs the business scope code to be diffused and the business product code to be diffused into the fusion model to obtain second fusion information; and the information processing device uses the diffusion model to diffuse the target information matching the first fusion information from the second fusion information.

[0090] In the embodiments of this application, the fusion model can be a model configured in the information processing device; the fusion model can also be a model transmitted to the information processing device by other devices; the fusion model can also be a model obtained by the information processing device in other ways; the specific way in which the information processing device obtains the fusion model can be determined according to the actual situation, and the embodiments of this application do not limit this.

[0091] In this embodiment of the application, the business scope code to be expanded is obtained by inputting the business scope information to be expanded (the business scope information of the object to be expanded) into the attenuation conversion coding model; the business product code to be expanded is obtained by inputting the business product information to be expanded (the business product information of the object to be expanded) into the attenuation conversion coding model.

[0092] In this embodiment of the application, the business scope code to be tested is the information obtained by inputting the business scope information to be tested (the business scope information of the object to be tested) into the attenuation conversion coding model; the business product code to be tested is the information obtained by inputting the business product information to be tested (the business product information of the object to be tested) into the attenuation conversion coding model.

[0093] In this embodiment of the application, the process by which the information processing device inputs the business scope code and the business product code to be tested into the fusion model to obtain the first fusion information can be that the information processing device uses the fusion coefficient in the fusion model to fuse the business scope code and the business product code to be tested to obtain the first fusion information.

[0094] It should be noted that there can be multiple first fusion information; the specific number of first fusion information can be determined according to the actual situation, and this application embodiment does not limit this.

[0095] In this embodiment of the application, the process of the information processing device inputting the business scope code and the business product code to be diversified into the fusion model to obtain the second fusion information includes the information processing device using the fusion coefficient in the fusion model to fuse the business scope code and the business product code to be diversified to obtain the second fusion information.

[0096] It should be noted that there can be multiple second fusion information; the specific number of second fusion information can be determined according to the actual situation, and this application embodiment does not limit this.

[0097] In this embodiment of the application, there are multiple first fusion information and multiple second fusion information. The process by which the information processing device diffuses target information matching the first fusion information from the second fusion information using a diffusion model includes: the information processing device filtering out objects of the same category from the multiple first fusion information to obtain multiple first test sub-objects and multiple first sub-fusion information corresponding to the multiple first test sub-objects; the information processing device determining the object range of the multiple first test sub-objects based on the multiple first sub-fusion information; and the information processing device determining the fusion information within the object range from the multiple second fusion information as target information.

[0098] In this embodiment of the application, the process by which the information processing device filters out objects of the same category from multiple first fused information to obtain multiple first test sub-objects can be that the information processing device filters out objects with the same category of business product information and the same category of business scope information from multiple first fused information to obtain multiple first test sub-objects.

[0099] In this embodiment of the application, the number of objects to be tested is multiple; the process by which the information processing device filters out objects of the same category from multiple first fusion information to obtain multiple first sub-objects to be tested includes: the information processing device determines the distance between multiple objects to be tested based on multiple first fusion information to obtain multiple first distances; the information processing device filters out objects of the same category from multiple objects to be tested based on multiple first distances to obtain multiple first sub-objects to be tested.

[0100] In this embodiment of the application, the process by which the information processing device determines the distance between multiple test objects based on multiple first fusion information and obtains multiple first distances can be that the information processing device determines the distance between any two test objects among the multiple test objects based on multiple first fusion information, thereby obtaining multiple first distances.

[0101] For example, if the number of objects to be tested is 3: A, B, and C, the information processing device can determine the first distance between A and B, the first distance between A and C, and the first distance between B and C based on multiple first fusion information, thereby obtaining multiple first distances.

[0102] For example, as shown in formula (6): Given k test objects (enterprises), calculate the distance between each enterprise and other enterprises, i.e., the pairwise distance between enterprises, to obtain a matrix M with dimension k*k, where each column M... j Let be the distances from the j-th enterprise to all other enterprises. This matrix is ​​a symmetric matrix.

[0103]

[0104] In this embodiment, the process by which the information processing device filters out objects of the same category from multiple test objects based on multiple first distances to obtain multiple first test sub-objects can be described as the information processing device sorting each row of matrix M in ascending order to obtain M sorted As shown in formula (7), by drawing on the idea of ​​nearest neighbors, outliers should be far away from most points. Thus, the first ones in each row are the companies that are most similar to the corresponding companies in terms of value.

[0105]

[0106] Information processing device for M sorted The first t values ​​in each row are averaged and then sorted in ascending order to obtain M′. mean_sorted =(d1…d t ).

[0107] It should be noted that t can be determined based on M′ mean_sorted Adjustments are made based on the fluctuations in the value, if it is d j If the value fluctuates greatly, then t should be smaller; otherwise, it should be larger.

[0108] It should also be noted that the magnitude of fluctuation can be determined by detecting the difference between each row and the mean, or by considering the variance.

[0109] In this embodiment, since some enterprises (test targets) may have significantly different business scopes from most other enterprises, without filtering by t, the diffusion process might include enterprises of different types. Therefore, the goal is to identify "homogeneous / completely similar" enterprises from the k test targets (enterprises), where t can be equal to k. This indicates that the provided test targets are highly similar. In other words, multiple first test sub-objects constitute multiple test targets; when t is less than k, multiple first test sub-objects can be identified based on t.

[0110] In this embodiment of the application, the process by which the information processing device determines the object range of multiple first sub-objects to be tested based on multiple first sub-fusion information includes: the information processing device determining the mean of multiple first sub-fusion information and using the mean as a first center point; the information processing device determining multiple second distances between the multiple first sub-fusion information and the first center point; and selecting the target distance with the largest distance value from the multiple second distances; and the information processing device determining the object range based on the first center point and the target distance.

[0111] In this embodiment of the application, the process by which the information processing device determines the object range based on the first center point and the target distance can be that the information processing device maps multiple first sub-fusion information to a numerical space, draws a circle with the first center point as the center and the target distance as the radius, and thus obtains the object range.

[0112] In this embodiment of the application, the information processing device also maps multiple second fusion information into the numerical space, so as to determine the fusion information within the object range among the multiple second fusion information as the target information.

[0113] For example, the information processing device for M′ mean_sorted The original codes of the first t enterprises (multiple first sub-fusion information) are averaged at their corresponding positions to obtain the diffusion center point (first center point) O. The distances between O and each of the t enterprises are then calculated, and the maximum distance is the maximum diffusion radius (target distance) r. If the original code matrix of the t enterprises is... Taking the arithmetic mean at the corresponding position involves averaging each column of matrix A, ultimately yielding the center point. Then, calculating the distance between O and each of the t companies, we obtain the following: For each row of matrix A, calculate its Euclidean distance to O. The maximum diffusion radius r is the maximum value in D, i.e., r = max(D).

[0114] For example, such as Figure 6As shown, the information processing device maps multiple first sub-fusion information and multiple second fusion information into a numerical space, and draws a circle with the first center point as the center and the target distance as the radius (r), thus obtaining the object range (circles in the figure). The information processing device determines the fusion information within the object range from the multiple second fusion information as target information, and the corresponding target object is taken as the related object related to the object to be tested. The object to be tested within the circle consists of multiple first sub-objects to be tested, and the object to be diffused within the circle is the related object.

[0115] It should be noted that each point uniquely corresponds to one enterprise (the object to be tested or the object to be disseminated).

[0116] In this embodiment of the application, the process by which the information processing device inputs the business scope code and the business product code to be tested into the fusion model to obtain the first fusion information includes: the information processing device determining the fusion coefficient in the fusion model; and the information processing device using the fusion model and the fusion coefficient to fuse the business scope code and the business product code to be tested to obtain the first fusion information.

[0117] In the embodiments of this application, the fusion coefficient can be a value in the range of 0-1.

[0118] In this embodiment of the application, the information processing device can determine the difference coefficient between 1 and the fusion coefficient; then the information processing device determines the first product between the fusion coefficient and the business scope code to be tested, and the second product between the difference coefficient and the business product code to be tested; finally, the information processing device uses the sum of the first product and the second product as the first fusion information.

[0119] For example, the information processing device uses a fusion model and a fusion coefficient to fuse the business scope code and the business product code to be tested to obtain the first fused information in the manner shown in formula (5):

[0120] ν i =β*ν i,待测试经营范围编码 +(1β)*ν i,待测试经营产品编码 —(5)

[0121] It should be noted that β is the fusion coefficient, (1β) is the difference coefficient, and ν i,待测试经营范围编码 Encode the business scope to be tested for the i-th test object; ν i,待测试经营产品编码 The product code to be tested for the i-th test object is defined. Here, β can be 0.6; β can also be a numerical value; the specific value of the fusion coefficient can be determined according to the actual situation, and this application embodiment does not limit this.

[0122] In this embodiment, the information processing device inputs the business information to be tested into the attenuation conversion coding model. Before obtaining the coding information to be tested, the information processing device also obtains the sample business information corresponding to the sample object; and inputs the sample business information into the initial attenuation conversion coding model to obtain sample coding information; the information processing device uses the initial diffusion model to diffuse the target sample coding information corresponding to the first sample coding information from the second sample coding information; the information processing device inputs the target sample coding information and the first sample coding information into the training evaluation model to obtain the training evaluation result; if the training evaluation result does not meet the preset training evaluation parameter range, the information processing device continues to train the initial attenuation conversion coding model and the initial diffusion model using the sample business information until the training evaluation result after continued training meets the preset training evaluation parameter range, thus obtaining the attenuation conversion coding model and the diffusion model.

[0123] It should be noted that the sample coding information includes first sample coding information and second sample coding information. The sample objects include the sample objects to be tested and the sample objects to be diffused; the first sample coding information is the coding information corresponding to the sample objects to be tested; the second sample coding information is the coding information corresponding to the sample objects to be diffused.

[0124] It should be noted that the target sample encoding information can be the first sample encoding information; the target sample encoding information can also be a part of the first sample encoding information; the specific information can be determined according to the actual situation, and this application embodiment does not limit it.

[0125] In this embodiment, the process by which the information processing device inputs sample operating information into the initial attenuation conversion coding model to obtain sample coding information is the same as the process by which the information processing device inputs the operating information to be tested into the attenuation conversion coding model to obtain the coding information to be tested. Specifically, the process by which the information processing device inputs sample operating information into the initial attenuation conversion coding model to obtain sample coding information can be referred to the process by which the information processing device inputs the operating information to be tested into the attenuation conversion coding model to obtain the coding information to be tested.

[0126] In this embodiment of the application, the process of the information processing device inputting sample operation information into the initial attenuation conversion coding model to obtain sample coding information includes the information processing device preprocessing the sample operation information to obtain preprocessed sample operation information; and the information processing device inputting the preprocessed sample operation information into the initial attenuation conversion coding model to obtain sample coding information.

[0127] It should be noted that the process by which the information processing device preprocesses the sample business information to obtain the preprocessed sample business information is the same as the process by which the information processing device preprocesses the business information to be tested to obtain the preprocessed business information to be tested. The specific process by which the information processing device preprocesses the sample business information to obtain the preprocessed sample business information can be referred to the process by which the information processing device preprocesses the business information to be tested to obtain the preprocessed business information to be tested.

[0128] S104. Determine the target object corresponding to the target information from the objects to be diffused, and treat the target object as a related object related to the object to be tested.

[0129] In this embodiment of the application, after the information processing device uses a diffusion model to diffuse target information corresponding to the code information to be tested from the code information to be diffused, the information processing device determines the target object corresponding to the target information from the object to be diffused, and regards the target object as a related object related to the object to be tested.

[0130] In this embodiment, the target sample coding information includes the target sample business scope code and the target sample business product code; the first sample coding information includes the first sample business scope code and the first sample business product code; the process by which the information processing device inputs the target sample coding information and the first sample coding information into the training evaluation model to obtain the training evaluation result includes: the information processing device inputs the target sample business scope code and the target sample business product code into the initial fusion model to obtain target sample fusion information; inputs the first sample business scope code and the first sample business product code into the initial fusion model to obtain first sample fusion information; the information processing device obtains the target sample national standard information corresponding to the target sample fusion information and the first sample national standard information corresponding to the first sample fusion information; the information processing device classifies the first sample fusion information and the target sample fusion information according to the national standard classification level of the target sample national standard information and the first sample national standard information to obtain multi-class sample fusion information; the information processing device determines the ratio between adjacent national standard classification levels according to the multi-class sample fusion information; and uses the ratio as the training evaluation result.

[0131] In this embodiment of the application, if the training evaluation result does not meet the preset training evaluation parameter range, the initial decay conversion coding model, the initial diffusion model, and the initial fusion model are continued to be trained using sample business information until the training evaluation result after continued training meets the preset training evaluation parameter range, and then the decay conversion coding model, the diffusion model, and the fusion model are obtained.

[0132] In this embodiment of the application, the process by which the information processing device continues to train the initial decay conversion coding model and the initial diffusion model using sample business information when the training evaluation result does not meet the preset training evaluation parameter range includes: the information processing device determining that the training evaluation result does not meet the preset training evaluation parameter range when the trend of the ratio between adjacent national standard classification levels is different; and continuing to train the initial decay conversion coding model and the initial diffusion model using sample business information.

[0133] In this embodiment of the application, the process by which the information processing device determines the ratio between adjacent national standard classification levels based on multi-class sample fusion information includes: the information processing device determining the center point of each class of sample fusion information in the multi-class sample fusion information to obtain multiple second center points; the information processing device determining the distance between the multi-class sample fusion information and the multiple second center points to obtain multiple sets of third distances; the information processing device determining the mean of the multiple sets of third distances to obtain multiple mean values; and the information processing device determining the ratio between adjacent national standard classification levels based on the multiple mean values.

[0134] It should be noted that the multi-class sample fusion information corresponds one-to-one with multiple sets of third distances.

[0135] In this embodiment, using the national standard classification, the average distance ratio of enterprises in the large and small categories at the industry level can be gradually reduced. The average distance ratio of enterprises under different classification levels is used to measure the model effect. The specific implementation method can be: calculate the distance from each point in each category to the center point of that category in the national standard classification in turn, and then take the average of all categories for each level. The average distance of each level is shown in formula (8):

[0136]

[0137] Where j = {1, 2, 3, 4}, is the corresponding national standard classification, namely the national standard first-level classification, second-level classification, etc.;

[0138] It is the number of categories under the j-th level of the national standard classification.

[0139] O j,k : The center point of the k-th category under the j-th level national standard classification,

[0140] n j,k The number of enterprises in the k-th category under the j-th national standard classification.

[0141] x j,k,z : The z-th point of the k-th category under the j-th level national standard classification.

[0142] The final statistical indicator is the ratio of average distances between adjacent levels:

[0143]

[0144] Where i = {2, 3, 4}.

[0145] For example, in the scenario described, a sample of 1 million enterprises belonging to the primary industry of manufacturing was taken. There are 31 secondary categories, and partial information of the tertiary categories under each secondary category is as follows: Figure 7 As shown: If the secondary national standard classification category is Computer, Communication and Other Electronic Equipment Manufacturing, then the corresponding number of tertiary national standard classification categories is 9; if the secondary national standard classification category is Communication Equipment Manufacturing, then the corresponding number of tertiary national standard classification categories is 9; if the secondary national standard classification category is Special Equipment Manufacturing, then the corresponding number of tertiary national standard classification categories is 9; if the secondary national standard classification category is Metal Products Manufacturing, then the corresponding number of tertiary national standard classification categories is 9; if the secondary national standard classification category is Non-metallic Mineral Products Manufacturing, then the corresponding number of tertiary national standard classification categories is 9; If the national standard classification category is railway, shipbuilding, aerospace and other transportation equipment manufacturing, then the corresponding number of tertiary national standard classification categories is 9; if the secondary national standard classification category is pharmaceutical manufacturing, then the corresponding number of tertiary national standard classification categories is 8; if the secondary national standard classification category is electrical machinery and equipment manufacturing, then the corresponding number of tertiary national standard classification categories is 8; if the secondary national standard classification category is textile industry, then the corresponding number of tertiary national standard classification categories is 8; if the secondary national standard classification category is chemical raw materials and chemical products manufacturing, then the corresponding number of tertiary national standard classification categories is 8. If the first-level national standard classification category is agricultural and sideline food processing, then the corresponding number of third-level national standard classification categories is 8; if the second-level national standard classification category is metal products, machinery and equipment repair, then the corresponding number of third-level national standard classification categories is 7; if the second-level national standard classification category is food manufacturing, then the corresponding number of third-level national standard classification categories is 7; if the second-level national standard classification category is automobile manufacturing, then the corresponding number of third-level national standard classification categories is 7; if the second-level national standard classification category is instrument and meter manufacturing, then the corresponding number of third-level national standard classification categories is 6; if the second-level national standard classification category is education and culture... The manufacturing of arts and crafts, sports and entertainment products corresponds to 6 categories in the third-level national standard classification; the leather, fur, feather and related products and footwear manufacturing industries correspond to 5 categories in the second-level national standard classification; the furniture manufacturing industry corresponds to 5 categories in the second-level national standard classification; the non-ferrous metal smelting and rolling processing industry corresponds to 5 categories in the second-level national standard classification; and the ferrous metal smelting and rolling processing industry corresponds to 4 categories in the second-level national standard classification. Partial information on the fourth-level categories under each third-level classification is as follows: Figure 8As shown: After encoding the enterprises in each of the 31 secondary classification categories according to the above steps, calculate the center point O of each category, and then calculate the Euclidean distance d from the center point to all other points in that category. i The average distance for all distances belonging to the same national standard level is calculated by averaging all distances. Thus, the average distances for Level 1, Level 2, Level 3, and Level 4 are 0.1126, 0.0985, 0.0832, and 0.0714, respectively, according to formula (8). Introducing the average distance ratio between adjacent categories, using formula (9), we obtain Level 2 / Level 1 as 0.8748, Level 3 / Level 2 as 0.8447, and Level 4 / Level 3 as 0.8582, respectively. Figure 9 As shown.

[0146] It should be noted that if the national standard level 2 includes general equipment manufacturing, special equipment manufacturing, metal products manufacturing, etc., the average distance of the center point and other points under each category to the center point should be calculated for each enterprise belonging to these subcategories. Then, the average distance of these subcategories should be calculated to obtain the average distance of the entire level 2 category.

[0147] In this embodiment, the information processing device inputs the target sample business scope code and the target sample business product code into the initial fusion model to obtain target sample fusion information; after inputting the first sample business scope code and the first sample business product code into the initial fusion model to obtain first sample fusion information, the information processing device draws an aggregation image based on the target sample fusion information and the first sample fusion information; the information processing device determines the aggregation of the target sample fusion information and the first sample fusion information based on the aggregation image; and uses the aggregation as the training evaluation result; correspondingly, when the training evaluation result does not meet the preset training evaluation parameter range, the information processing device continues to train the initial decay conversion coding model and the initial diffusion model using the sample business information, including: when the target sample fusion information and the first sample fusion information do not aggregate, the information processing device determines that the training evaluation result does not meet the preset training evaluation parameter range; the information processing device continues to train the initial decay conversion coding model and the initial diffusion model using the sample business information.

[0148] In the embodiments of this application, the aggregated image can be a two-dimensional (2D) image; the aggregated image can also be a three-dimensional image; the specific aggregated image can be determined according to the actual situation, and the embodiments of this application do not limit it.

[0149] In this embodiment of the application, the process of the information processing device drawing an aggregated image based on the target sample fusion information and the first sample fusion information includes: the information processing device performing dimensionality reduction on the target sample fusion information and the first sample fusion information to obtain dimensionality-reduced target sample fusion information and dimensionality-reduced first sample fusion information; and the information processing device drawing an aggregated image based on the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information.

[0150] In this embodiment of the application, the target sample fusion information and the first sample fusion information are 768-dimensional information; the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information can be 2-dimensional information.

[0151] For example, following the method proposed in this application embodiment of "inputting the target sample encoding information and the first sample encoding information into the training evaluation model to obtain the training evaluation result", the average distance ratio (i.e., the ratio between adjacent national standard classification levels) of 1 million manufacturing enterprises is calculated. Then, based on the four versions generated during the model iteration process, i.e., four models, the average distance between adjacent categories under the four levels of national standard classification is calculated for each. Figure 10 As shown.

[0152] In this embodiment of the application, the obtained aggregated image is as follows: Figure 11 As shown: The three colors of dots in the figure represent three categories. Ideally, dots of the same color should cluster together.

[0153] For example, such as Figure 12 As shown:

[0154] S1. The information processing device acquires the sample business information corresponding to the sample object; and preprocesses the sample business information to obtain the preprocessed sample business information.

[0155] S2. The information processing device inputs the preprocessed sample operation information into the initial attenuation conversion coding model to obtain sample coding information.

[0156] It should be noted that the sample coding information includes the first sample coding information and the second sample coding information; the first sample coding information includes the first sample business scope code and the first sample business product code.

[0157] S3. The information processing device uses the initial diffusion model to diffuse the target sample coding information corresponding to the first sample coding information from the second sample coding information.

[0158] It should be noted that the target sample coding information includes the target sample business scope code and the target sample business product code.

[0159] S4. The information processing device inputs the target sample's business scope code and target sample's business product code into the initial fusion model to obtain the target sample fusion information; it also inputs the first sample's business scope code and first sample's business product code into the initial fusion model to obtain the first sample fusion information.

[0160] S5. The information processing device acquires the target sample national standard information corresponding to the target sample fusion information and the first sample national standard information corresponding to the first sample fusion information.

[0161] S6. The information processing device classifies the first sample fusion information and the target sample fusion information according to the national standard classification level of the target sample national standard information and the first sample national standard information, and obtains multi-class sample fusion information.

[0162] S7. The information processing device determines the ratio between adjacent national standard classification levels based on the fusion information of multiple samples; and uses the ratio as the training evaluation result.

[0163] S8. If the training evaluation results do not meet the preset training evaluation parameter range, the information processing device continues to train the initial decay conversion coding model, the initial diffusion model, and the initial fusion model using sample business information until the training evaluation results after continued training meet the preset training evaluation parameter range, and then obtains the decay conversion coding model, the diffusion model, and the fusion model.

[0164] Understandably, the information processing device uses an attenuation conversion coding model to attenuate and convert the business information of the test object to obtain the test coding information. Then, using a diffusion model, it diffuses the target information corresponding to the test coding information from the diffusion coding information to determine the target object corresponding to the target information from the diffusion object. That is, it obtains the relevant objects related to the test object. Even when the business information of the test object changes over time, the attenuation conversion coding model and the diffusion model can be used to directly determine the relevant objects related to the test object based on the changed business information of the test object, thus improving the accuracy of determining the relevant objects.

[0165] Based on the same inventive concept as the above-mentioned information processing method, this application provides an information processing device 1, corresponding to an information processing method; Figure 13 A schematic diagram of the composition structure of an information processing device provided in this application embodiment. Figure 1 The information processing device 1 may include:

[0166] Input unit 11 is used to input the business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested when the business information to be tested of the object to be tested is obtained.

[0167] Acquisition unit 12 is used to acquire the diffusion encoding information corresponding to the object to be diffused;

[0168] Diffusion unit 13 is used to diffuse target information corresponding to the coding information to be tested from the coding information to be diffused using a diffusion model;

[0169] The determining unit 14 is used to determine the target object corresponding to the target information from the object to be diffused, and to regard the target object as a related object related to the object to be tested.

[0170] In some embodiments of this application, the coding information to be disseminated includes the business scope code to be disseminated and the business product code to be disseminated; the coding information to be tested includes the business scope code to be tested and the business product code to be tested.

[0171] The input unit 11 is used to input the business scope code to be tested and the business product code to be tested into the fusion model to obtain first fusion information; and to input the business scope code to be diffused and the business product code to be diffused into the fusion model to obtain second fusion information.

[0172] The diffusion unit 13 is used to diffuse target information that matches the first fusion information from the second fusion information using a diffusion model.

[0173] In some embodiments of this application, the number of the first fused information is multiple, the number of the second fused information is multiple, and the device further includes a filtering unit;

[0174] The filtering unit is used to filter out objects of the same category from multiple first fusion information to obtain multiple first test sub-objects and multiple first sub-fusion information corresponding to the multiple first test sub-objects;

[0175] The determining unit 14 is used to determine the object range of the plurality of first sub-sub-objects to be tested based on the plurality of first sub-fusion information; and to determine the fusion information that is within the object range from the plurality of second fusion information as the target information.

[0176] In some embodiments of this application, the number of objects to be tested is multiple;

[0177] The determining unit 14 is used to determine the distance between multiple test objects based on the multiple first fusion information, and obtain multiple first distances;

[0178] The filtering unit is used to filter out objects of the same category from the plurality of test objects based on the plurality of first distances, thereby obtaining the plurality of first test sub-objects.

[0179] In some embodiments of this application, the determining unit 14 is configured to determine the mean of the plurality of first sub-fusion information and use the mean as a first center point; determine a plurality of second distances between the plurality of first sub-fusion information and the first center point; and determine the object range based on the first center point and the target distance.

[0180] The filtering unit is used to filter out the target distance with the largest distance value from the plurality of second distances.

[0181] In some embodiments of this application, the apparatus further includes a fusion unit;

[0182] The determining unit 14 is used to determine the fusion coefficients in the fusion model;

[0183] The fusion unit is used to fuse the business scope code to be tested and the business product code to be tested using the fusion model and according to the fusion coefficient to obtain the first fusion information.

[0184] In some embodiments of this application, the apparatus further includes a processing unit;

[0185] The processing unit is used to preprocess the business information to be tested to obtain preprocessed business information to be tested;

[0186] The input unit 11 is used to input the preprocessed business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested.

[0187] In some embodiments of this application, the apparatus further includes a removal unit;

[0188] The determining unit 14 is used to determine the words to be cleared that match the preset word list from the business information to be tested;

[0189] The removal unit is used to remove the words to be removed from the business information to be tested, so as to obtain the preprocessed business information to be tested.

[0190] In some embodiments of this application, the apparatus further includes a conversion unit;

[0191] The conversion unit is used to convert each character in the preprocessed business information to be tested into a numerical value using the family of functions in the attenuation conversion coding model, thereby obtaining multiple sets of numerical values.

[0192] The determining unit 14 is used to determine the attenuation coefficient corresponding to each character according to the character attenuation parameter in the attenuation conversion coding model, thereby obtaining multiple attenuation coefficients; and to determine the coding information to be tested according to the multiple attenuation coefficients and the multiple sets of values.

[0193] In some embodiments of this application, the apparatus further includes a training unit;

[0194] The acquisition unit 12 is used to acquire sample operation information corresponding to the sample object;

[0195] The input unit 11 is used to input the sample management information into the initial attenuation conversion coding model to obtain sample coding information; the sample coding information includes first sample coding information and second sample coding information; the target sample coding information and the first sample coding information are input into the training evaluation model to obtain training evaluation results;

[0196] The diffusion unit 13 is used to use an initial diffusion model to diffuse out target sample coding information corresponding to the first sample coding information from the second sample coding information.

[0197] The training unit is used to continue training the initial decay conversion coding model and the initial diffusion model using the sample business information when the training evaluation result does not meet the preset training evaluation parameter range, until the decay conversion coding model and the diffusion model are obtained when the training evaluation result after continued training meets the preset training evaluation parameter range.

[0198] In some embodiments of this application, the target sample coding information includes a target sample business scope code and a target sample business product code; the first sample coding information includes a first sample business scope code and a first sample business product code; the device further includes a classification unit;

[0199] The input unit 11 is used to input the target sample business scope code and the target sample business product code into the initial fusion model to obtain target sample fusion information; and to input the first sample business scope code and the first sample business product code into the initial fusion model to obtain first sample fusion information.

[0200] The acquisition unit 12 is used to acquire the target sample national standard information corresponding to the target sample fusion information, and the first sample national standard information corresponding to the first sample fusion information;

[0201] The classification unit is used to classify the first sample fusion information and the target sample fusion information according to the national standard classification level of the target sample national standard information and the first sample national standard information, so as to obtain multi-class sample fusion information;

[0202] The determining unit 14 is used to determine the ratio between adjacent national standard classification levels based on the multi-class sample fusion information, and to use the ratio as the training evaluation result.

[0203] In some embodiments of this application, the determining unit 14 is used to determine that the training evaluation result does not meet the preset training evaluation parameter range when the trends of the ratios between adjacent national standard classification levels are different.

[0204] The training unit is used to continue training the initial decay conversion coding model and the initial diffusion model using the sample management information.

[0205] In some embodiments of this application, the determining unit 13 is used to determine the center point of each type of sample fusion information in the multi-type sample fusion information to obtain multiple second center points; determine the distance between the multi-type sample fusion information and the multiple second center points to obtain multiple sets of third distances; the multi-type sample fusion information and the multiple sets of third distances correspond one-to-one; determine the mean of the multiple sets of third distances to obtain multiple mean values; and determine the ratio between adjacent national standard classification levels based on the multiple mean values.

[0206] In some embodiments of this application, the apparatus further includes a drawing unit;

[0207] The drawing unit is used to draw an aggregated image based on the target sample fusion information and the first sample fusion information;

[0208] The determining unit 14 is configured to determine the clustering of the target sample fusion information and the first sample fusion information based on the clustered image, and use the clustering as the training evaluation result.

[0209] Correspondingly, the determining unit 14 is used to determine that the training evaluation result does not meet the preset training evaluation parameter range when the target sample fusion information and the first sample fusion information do not converge.

[0210] The training unit is used to continue training the initial decay conversion coding model and the initial diffusion model using the sample management information.

[0211] In some embodiments of this application, the device further includes a dimensionality reduction unit;

[0212] The dimensionality reduction unit is used to reduce the dimensionality of the target sample fusion information and the first sample fusion information to obtain the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information.

[0213] The drawing unit is used to draw the clustered image based on the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information.

[0214] It should be noted that, in practical applications, the aforementioned input unit 11, acquisition unit 12, diffusion unit 13, and determination unit 14 can be implemented by the processor 15 on the information processing device 1, specifically by a CPU (Central Processing Unit), MPU (Microprocessor Unit), DSP (Digital Signal Processor), or Field Programmable Gate Array (FPGA), etc.; the aforementioned data storage can be implemented by the memory 16 on the information processing device 1.

[0215] This application embodiment also provides an information processing device 1, such as... Figure 14 As shown, the information processing device 1 includes a processor 15, a memory 16, and a communication bus 17. The memory 16 communicates with the processor 15 through the communication bus 17. The memory 16 stores programs executable by the processor 15. When the program is executed, the processor 15 performs the information processing method as described above.

[0216] In practical applications, the aforementioned memory 16 can be volatile memory, such as random-access memory (RAM); or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 15.

[0217] This application provides a computer-readable storage medium having a computer program thereon, which, when executed by a processor 15, implements the information processing method as described above.

[0218] Understandably, the information processing device uses an attenuation conversion coding model to attenuate and convert the business information of the test object to obtain the test coding information. Then, using a diffusion model, it diffuses the target information corresponding to the test coding information from the diffusion coding information to determine the target object corresponding to the target information from the diffusion object. That is, it obtains the relevant objects related to the test object. Even when the business information of the test object changes over time, the attenuation conversion coding model and the diffusion model can be used to directly determine the relevant objects related to the test object based on the changed business information of the test object, thus improving the accuracy of determining the relevant objects.

[0219] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0220] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0221] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0222] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0223] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. An information processing method, characterized in that, The method includes: Once the test operation information of the test object is obtained, the test operation information is input into the attenuation conversion coding model to obtain the test coding information; The attenuation conversion coding model is a model that increases the information weight of the first text information in the preprocessed business information to be tested. Obtain the diffusion encoding information corresponding to the object to be diffused; Using a diffusion model, target information corresponding to the coding information to be tested is diffused from the coding information to be diffused. The target object corresponding to the target information is determined from the objects to be diffused, and the target object is used as a related object associated with the object to be tested; the coding information to be diffused includes the business scope code to be diffused and the business product code to be diffused; the coding information to be tested includes the business scope code to be tested and the business product code to be tested; the step of using a diffusion model to diffuse the target information corresponding to the coding information to be tested from the coding information to be diffused includes: The business scope code and the business product code to be tested are input into the fusion model to obtain the first fusion information; The code for the business scope to be expanded and the code for the business products to be expanded are input into the fusion model to obtain the second fusion information; A diffusion model is used to diffuse target information matching the first fusion information from the second fusion information; the number of first fusion information pieces is multiple, the number of second fusion information pieces is multiple, and the step of using a diffusion model to diffuse target information matching the first fusion information from the second fusion information includes: From multiple first fusion information, objects of the same category are selected to obtain multiple first test sub-objects and multiple first sub-fusion information corresponding to the multiple first test sub-objects; Based on the plurality of first sub-fusion information, the object range of the plurality of first sub-objects to be tested is determined; The fused information that falls within the scope of the object from among multiple second fused information is determined as the target information. The scope of the object is the range within which multiple first sub-fusion information are mapped into the numerical space, with the first center point as the center and the target distance as the radius.

2. The method according to claim 1, characterized in that, The number of objects to be tested is multiple; the step of filtering out objects of the same category from multiple first fusion information to obtain multiple first sub-objects to be tested includes: Based on the multiple first fusion information, the distance between multiple test objects is determined, and multiple first distances are obtained; Based on the multiple first distances, objects of the same category are selected from the multiple test objects to obtain the multiple first test sub-objects.

3. The method according to claim 1, characterized in that, The step of determining the object range of the plurality of first sub-objects to be tested based on the plurality of first sub-fusion information includes: Determine the mean of the plurality of first sub-fusion information, and use the mean as the first center point; Determine multiple second distances between the plurality of first sub-fusion information and the first center point; and filter out the target distance with the largest distance value from the plurality of second distances; The range of the object is determined based on the first center point and the target distance.

4. The method according to claim 1, characterized in that, The step of inputting the business scope code and the business product code to be tested into the fusion model to obtain the first fusion information includes: Determine the fusion coefficients in the fusion model; The first fusion information is obtained by fusing the business scope code and the business product code to be tested using the fusion model and according to the fusion coefficient.

5. The method according to claim 1, characterized in that, The step of inputting the business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested includes: The business information to be tested is preprocessed to obtain preprocessed business information to be tested; The preprocessed business information to be tested is input into the attenuation conversion coding model to obtain the coding information to be tested.

6. The method according to claim 5, characterized in that, The preprocessing of the business information to be tested to obtain preprocessed business information to be tested includes: Determine the words to be removed from the business information to be tested that match the preset word list; The words to be removed are removed from the business information to be tested to obtain the preprocessed business information to be tested.

7. The method according to claim 5, characterized in that, The step of inputting the preprocessed operational information to be tested into the attenuation conversion coding model to obtain the coding information to be tested includes: The function family in the attenuation conversion coding model is used to convert each character in the preprocessed business information to be tested into a numerical value, resulting in multiple sets of numerical values. Based on the character attenuation parameters in the attenuation conversion coding model, the attenuation coefficient corresponding to each character is determined, resulting in multiple attenuation coefficients; The coding information to be tested is determined based on the multiple attenuation coefficients and the multiple sets of values.

8. The method according to claim 1, characterized in that, Before inputting the business information to be tested into the attenuation conversion coding model to obtain the coding information to be tested, the method further includes: Obtain sample operation information corresponding to the sample object; and input the sample operation information into the initial attenuation conversion coding model to obtain sample coding information; the sample coding information includes first sample coding information and second sample coding information; Using an initial diffusion model, target sample coding information corresponding to the first sample coding information is diffused from the second sample coding information; The target sample encoding information and the first sample encoding information are input into the training evaluation model to obtain the training evaluation result; If the training evaluation result does not meet the preset training evaluation parameter range, the initial decay conversion coding model and the initial diffusion model are trained again using the sample business information until the training evaluation result after continued training meets the preset training evaluation parameter range, and the decay conversion coding model and the diffusion model are obtained.

9. The method according to claim 8, characterized in that, The target sample coding information includes the target sample business scope code and the target sample business product code; the first sample coding information includes the first sample business scope code and the first sample business product code; The step of inputting the target sample encoding information and the first sample encoding information into the training evaluation model to obtain the training evaluation result includes: The target sample's business scope code and the target sample's business product code are input into the initial fusion model to obtain target sample fusion information; the first sample's business scope code and the first sample's business product code are input into the initial fusion model to obtain first sample fusion information. Obtain the target sample national standard information corresponding to the target sample fusion information, and the first sample national standard information corresponding to the first sample fusion information; Based on the national standard classification levels of the target sample national standard information and the first sample national standard information, the first sample fusion information and the target sample fusion information are classified to obtain multiple types of sample fusion information; Based on the multi-class sample fusion information, the ratio between adjacent national standard classification levels is determined; and the ratio is used as the training evaluation result.

10. The method according to claim 9, characterized in that, When the training evaluation result does not meet the preset training evaluation parameter range, the method of continuing to train the initial decay conversion coding model and the initial diffusion model using the sample management information includes: When the trends of the ratios between adjacent national standard classification levels are different, it is determined that the training evaluation results do not meet the preset training evaluation parameter range; and the initial decay conversion coding model and the initial diffusion model are continued to be trained using the sample business information.

11. The method according to claim 9, characterized in that, The step of determining the ratio between adjacent national standard classification levels based on the multi-class sample fusion information includes: Determine the center point of each type of sample fusion information in the multi-type sample fusion information to obtain multiple second center points; The distances between the multi-class sample fusion information and the multiple second center points are determined respectively to obtain multiple sets of third distances; the multi-class sample fusion information and the multiple sets of third distances correspond one-to-one. The mean of the third distance is determined for multiple groups, resulting in multiple means; Based on the multiple mean values, determine the ratio between adjacent national standard classification levels.

12. The method according to claim 9, characterized in that, After inputting the target sample's business scope code and the target sample's business product code into the initial fusion model to obtain target sample fusion information; and inputting the first sample's business scope code and the first sample's business product code into the initial fusion model to obtain first sample fusion information, the method further includes: Based on the target sample fusion information and the first sample fusion information, a clustered image is drawn; The clustering of the target sample fusion information and the first sample fusion information is determined based on the clustered image; and the clustering is used as the training evaluation result. Accordingly, when the training evaluation result does not meet the preset training evaluation parameter range, the step of continuing to train the initial decay conversion coding model and the initial diffusion model using the sample management information includes: If the target sample fusion information and the first sample fusion information do not converge, it is determined that the training evaluation result does not meet the preset training evaluation parameter range. The initial attenuation conversion coding model and the initial diffusion model are then trained using the sample management information.

13. The method according to claim 12, characterized in that, The step of drawing an aggregated image based on the target sample fusion information and the first sample fusion information includes: The target sample fusion information and the first sample fusion information are dimensionality reduced to obtain the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information. The clustered image is drawn based on the dimensionality-reduced target sample fusion information and the dimensionality-reduced first sample fusion information.

14. An information processing device, characterized in that, The device includes: The input unit is used to input the business information to be tested into the attenuation conversion coding model when the business information to be tested is obtained, so as to obtain the code information to be tested; the attenuation conversion coding model is a model that increases the information weight of the first text information in the preprocessed business information to be tested. The acquisition unit is used to acquire the diffusion encoding information corresponding to the object to be diffused; A diffusion unit is used to diffuse target information corresponding to the coding information to be tested from the coding information to be diffused using a diffusion model. A determining unit is configured to determine the target object corresponding to the target information from the object to be diffused, and to regard the target object as a related object associated with the object to be tested; The input unit is used to input the business scope code to be tested and the business product code to be tested into the fusion model to obtain first fusion information; and to input the business scope code to be diffused and the business product code to be diffused into the fusion model to obtain second fusion information. The diffusion unit is used to diffuse target information that matches the first fusion information from the second fusion information using a diffusion model; The number of the first fused information is multiple, the number of the second fused information is multiple, and the device further includes a filtering unit; The filtering unit is used to filter out objects of the same category from multiple first fusion information to obtain multiple first test sub-objects and multiple first sub-fusion information corresponding to the multiple first test sub-objects; The determining unit is configured to determine the object range of the plurality of first sub-sub-objects to be tested based on the plurality of first sub-fusion information; and to determine the fusion information that falls within the object range from the plurality of second fusion information as the target information; The scope of the object is the range within which multiple first sub-fusion information are mapped into the numerical space, with the first center point as the center and the target distance as the radius.

15. An information processing device, characterized in that, The device includes: The system includes a memory, a processor, and a communication bus, wherein the memory communicates with the processor via the communication bus, and the memory stores a program for information processing that can be executed by the processor. When the program for information processing is executed, the processor performs the method as described in any one of claims 1 to 13.

16. A storage medium having a computer program stored thereon, used in an information processing device, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 13.

Images