An intelligent form filling method and device based on multi-scene semantic analysis and a medium

Abstract

The embodiment of the specification discloses an intelligent filling method and device based on multi-scene semantic analysis and medium, relates to the technical field of computers, and the method comprises the following steps: acquiring voice document information of a user's to-be-filled document, performing preliminary identification on the voice document information, and generating a voice document scene corresponding to the voice document information; according to the voice document scene, performing secondary identification on the voice document information through a preset expert library corresponding to the voice document scene, acquiring current filling data corresponding to the voice document scene; determining specified purpose information in the voice document information, performing semantic extraction on the specified purpose information to generate a current filling abstract; and performing intelligent filling on the to-be-filled document through the current filling data and the current filling abstract. The method realizes semantic scene, expert library, and peripheral information, automatically triggers intelligent filling, reduces the error rate and the rate of initial review and return, reduces the training cost of all employees, improves the accounting efficiency, and avoids complicated operation steps of the user.

Description

TECHNICAL FIELD

[0001] The present specification relates to the technical field of computer technology, and particularly relates to an intelligent filling method based on multi-scene semantic analysis, equipment and medium. BACKGROUND

[0002] With the gradual popularity of mobile applications, mobile end expense control has become an inevitable trend of development, and this method is being adopted by more and more enterprises. Cloud deployment of mobile applications also enables small and medium-sized enterprises to participate in digital expense control. The traditional expense reporting system is cumbersome to fill in, and the control of individualization results in large differences in filling. The learning cost of the filler is increased. According to the consumption or filling items, the associated information is matched and filled in to simplify the filling process, which is a compulsory course for promoting the better landing of mobile filling. When using cloud deployment of mobile applications to fill in, the filling steps are cumbersome, and the filler needs to be trained, which increases the learning cost of the filler and cannot guarantee the accuracy of the filling. SUMMARY

[0003] One or more embodiments of the present specification provide an intelligent filling method based on multi-scene semantic analysis, equipment and medium, which is used to solve the following technical problems: when using cloud deployment of mobile applications to fill in, the filling steps are cumbersome, and the filler needs to be trained, which increases the learning cost of the filler and cannot guarantee the accuracy of the filling.

[0004] One or more embodiments of the present specification adopt the following technical solutions:

[0005] One or more embodiments of the present specification provide an intelligent filling method based on multi-scene semantic analysis, which comprises: acquiring voice document information of a user to be filled in a document, performing preliminary identification on the voice document information, and generating a voice document scene corresponding to the voice document information; according to the voice document scene, performing secondary identification on the voice document information through a preset expert library corresponding to the voice document scene, acquiring current filling data corresponding to the voice document scene; determining specified purpose information in the voice document information, performing semantic extraction on the specified purpose information, and generating a current filling abstract; and performing intelligent filling on the to-be-filled document through the current filling data and the current filling abstract.

[0006] Further, according to the voice invoice scene, before the secondary identification of the voice invoice information through the preset expert library, the method further comprises: pre-acquiring a plurality of invoice data and invoice information corresponding to each invoice data, wherein the invoice information comprises invoice category information and preset itemized data; according to the invoice category information in the invoice information corresponding to each invoice data, distinguishing the scene of each invoice data to determine the initial invoice scene corresponding to each invoice data; according to the preset itemized data, further subdividing the initial invoice scene to determine the invoice application scene corresponding to each invoice data; according to the invoice application scene, performing reverse optimization of semantic association by principal component analysis to construct the preset enterprise library.

[0007] Further, according to the voice invoice scene, by means of the preset expert library, the voice invoice information is subjected to secondary identification to obtain the current filling data corresponding to the voice invoice scene, specifically comprising: identifying the voice invoice information to obtain invoice data information corresponding to the voice invoice information, wherein the invoice data information comprises invoice data time and invoice data type, and the invoice data type comprises any one or more of invoices, message records, consumption records and approval records; according to the invoice data information, through the preset expert library, a plurality of invoice source data corresponding to the invoice data information are pulled in the system corresponding to the invoice data information; according to the voice invoice scene, through the preset expert library, the plurality of invoice source data are matched to obtain the current filling data corresponding to the voice invoice scene.

[0008] Further, the voice invoice information is subjected to preliminary identification to generate a voice invoice scene corresponding to the voice invoice information, specifically comprising: extracting key semantic information of the voice invoice information, wherein the key semantic information is used to represent the application scene of the invoice; identifying the key semantic information to generate scene semantics; according to the scene semantics and a preset first mapping relationship, determining a voice invoice scene corresponding to the scene semantics, wherein the first mapping relationship comprises a plurality of scene semantics and a voice invoice scene corresponding to each scene semantic.

[0009] Further, the specified purpose information in the voice invoice information is determined, and the semantic extraction of the specified purpose information is performed to generate a current filling summary, specifically comprising: presetting a plurality of specified semantic fields, wherein the specified semantic fields are used to represent the purpose; splitting the voice invoice information to obtain a plurality of voice information; calculating the similarity between the voice fields corresponding to the voice information and the specified semantic fields; in a plurality of voice fields, determining a specified voice field with a similarity greater than a preset similarity threshold, and taking the specified voice field as the specified purpose information; extracting the semantics in the specified purpose information to generate a current filling summary.

[0010] Further, the method further comprises: acquiring the current filling data and the current filling summary; and intelligently filling the to-be-filled document according to the current filling data and the current filling summary.

[0011] Further, after the to-be-filled document is intelligently filled according to the current filling data and the current filling summary, the method further comprises: acquiring business association data in the current filling data; and perfecting associated information of the to-be-filled document according to the business association data, wherein the perfecting the associated information of the to-be-filled document specifically comprises: triggering an expression, perfecting a fee item, and perfecting apportionment information.

[0012] Further, the preset enterprise library is constructed according to the document application scenario by reverse optimization of semantic association through principal component analysis, specifically comprising: according to the document application scenario, the reverse optimization of semantic association through principal component analysis is performed to generate an optimized semantic recognition result corresponding to the document application scenario and a current scenario recognition model corresponding to the document application scenario; and the preset enterprise library is determined according to the optimized semantic recognition result and the current scenario recognition model.

[0013] One or more embodiments of the present specification provide an intelligent filling device based on multi-scene semantic analysis, comprising:

[0014] at least one processor; and

[0015] a memory in communication connection with the at least one processor; wherein

[0016] the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0017] acquire voice document information of a to-be-filled document of a user, perform preliminary identification on the voice document information, and generate a voice document scene corresponding to the voice document information; perform secondary identification on the voice document information through a preset expert library corresponding to the voice document scene according to the voice document scene, acquire current filling data corresponding to the voice document scene, determine specified purpose information in the voice document information, perform semantic extraction on the specified purpose information, and generate a current filling summary; and intelligently fill the to-be-filled document according to the current filling data and the current filling summary.

[0018] The one or more embodiments of the specification provide a non-volatile computer storage medium storing computer executable instructions configured to:

[0019] The voice document information of a user to-be-filled document is acquired, the voice document information is preliminarily identified to generate a voice document scene corresponding to the voice document information, the voice document information is secondarily identified through a preset expert library corresponding to the voice document scene according to the voice document scene to acquire current filling data corresponding to the voice document scene, specified purpose information in the voice document information is determined, semantic extraction is performed on the specified purpose information to generate a current filling abstract, and the to-be-filled document is intelligently filled through the current filling data and the current filling abstract.

[0020] The above at least one technical solution adopted by the embodiments of the specification can achieve the following beneficial effects: through the above technical solution, the voice document information input by the user voice is used to determine the voice document scene, the document data is identified according to the voice document scene through the expert library, the current filling data corresponding to the scene is obtained, the semantic scene is realized, the scene has the surrounding information brought out by the expert library, the intelligent filling method is automatically triggered at last, the error rate and the initial review rejection rate are reduced, the full-time expense report training cost is reduced, the expense report efficiency is improved, the user is avoided from complicated operation steps, and the user demand is met. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the specification or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the specification, and other drawings can be obtained by those skilled in the art without creative labor. In the drawings:

[0022] Figure 1 A flowchart of an intelligent filling method based on multi-scene semantic analysis provided by the embodiments of the specification is shown in the figure.

[0023] Figure 2 A structural schematic diagram of an intelligent filling device based on multi-scene semantic analysis provided by the embodiments of the specification is shown in the figure. DETAILED DESCRIPTION

[0024] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in combination with the drawings in the specification. Obviously, the described embodiments are only a part of the embodiments of the specification, not all the embodiments. Based on the embodiments of the specification, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the protection scope of the specification.

[0025] With the gradual popularity of mobile applications, mobile expense control has become an inevitable trend of development, and this method is being adopted by more and more enterprises. Cloud-deployed mobile applications also enable small and medium-sized enterprises to participate in digital expense control. The traditional expense reporting system is cumbersome, and the filling difference is large in the case of personalized control. The learning cost of the person filling in the form is increased. According to the consumption or filling matter, the associated information is matched and filled in to simplify the filling process, which is a compulsory course for promoting the better landing of mobile filling. When using a cloud-deployed mobile application to fill in the form, the filling step is cumbersome, the filling user needs to be trained, the learning cost of the filling user is increased, and the accuracy of the filling cannot be guaranteed.

[0026] The embodiments of the specification provide an intelligent filling method based on multi-scene semantic analysis. It should be noted that the execution subject in the embodiments of the specification can be a server or any device with data processing capability. Here, filling in the form refers to filling in the form. The embodiments of the specification are mainly used for the filling link of the ERP expense control cloud product, which belongs to distinguishing the filling scene (not the traditional voice recognition environment scene but the form filling scene), matching the scene to the preliminary result of voice recognition, and correcting errors; according to the recognition and correction result, the semantic association matching is performed, the related form content is brought out according to the semantic association expert library, and the filling of the form is realized through the intelligent filling method. Figure 1 The flowchart of the intelligent filling method based on multi-scene semantic analysis provided by the embodiments of the specification is shown in FIG. 1. Figure 1 As shown in FIG. 1, the method mainly includes the following steps:

[0027] Step S101, obtaining voice form information of a form to be filled by a user, performing preliminary recognition on the voice form information, and generating a voice form scene corresponding to the voice form information.

[0028] In an embodiment of the specification, when the user needs to fill in the form, the voice form information is input, for example, "I want to fill in the November business trip form", the voice form information of the form to be filled by the user is obtained, the voice form information is preliminarily recognized, and the voice form scene corresponding to the voice form information is determined, for example, a business trip form filling scene, a material form filling scene, a project expense form filling scene, or a salary expense form filling scene.

[0029] The voice invoice information is preliminarily identified, and a voice invoice scene corresponding to the voice invoice information is generated. Specifically, key semantic information of the voice invoice information is extracted, where the key semantic information is used to represent an application scene of an invoice. The key semantic information is identified to generate a scene semantic. According to the scene semantic and a preset first mapping relationship, a voice invoice scene corresponding to the scene semantic is determined, where the first mapping relationship includes a plurality of scene semantics and a voice invoice scene corresponding to each scene semantic.

[0030] In an embodiment of the present specification, key semantic information used to represent an application scene of an invoice is extracted from voice invoice information. Taking a business trip invoice filling scene as an example, the key semantic information used to represent the application scene of the business trip invoice can be "November business trip invoice", "business trip invoice", "departure to a certain place", "go to a certain place", and the like. The key semantic information is identified to obtain a scene semantic, such as "business trip", "business trip", "departure", "go", and the like. According to the obtained scene semantic, a voice invoice scene corresponding to the scene semantic is determined in a preset first mapping relationship, that is, according to the obtained scene semantic "departure", a voice invoice scene "business trip invoice filling scene" corresponding to the invoice is obtained. In addition, it should be noted that the first mapping relationship is a pre-constructed corresponding relationship between a scene semantic and a voice invoice scene corresponding to each scene semantic. A plurality of scene semantics under each voice invoice scene can be listed in advance, and one or more scene semantics corresponding to each voice invoice scene are obtained according to the required voice invoice scene of the enterprise in the historical filling process, and a corresponding relationship between the voice invoice scene and the scene semantic is established. The corresponding relationship here can be one-to-many or one-to-one. In addition, in addition to the type of the invoice, the voice invoice scene can also be distinguished according to the accounting unit.

[0031] In an embodiment of the present specification, after obtaining the voice invoice scene, the voice invoice scene corresponding filling type can be called according to the voice invoice scene, for example, after obtaining the business trip invoice filling scene, the type of the business trip invoice is automatically selected, which can be understood as a template for filling the business trip invoice. It should be noted that the selection of the business trip invoice filling template here can also be selected by the user.

[0032] In step S102, according to the voice invoice scene, the voice invoice information is identified again through a preset expert library corresponding to the voice invoice scene, and current filling data corresponding to the voice invoice scene is obtained.

[0033] Based on the voice document scenario, before performing secondary recognition of the voice document information using a pre-set expert database, the method further includes: pre-acquiring multiple document data and document information corresponding to each document data, wherein the document information includes document category information and pre-set project-based data; classifying each document data into scenarios based on the document category information in the document information corresponding to each document data, and determining the initial document scenario corresponding to each document data; further subdividing the initial document scenario based on the pre-set project-based data, and determining the document application scenario corresponding to each document data; and constructing the pre-set enterprise database by performing reverse optimization of semantic association using principal component analysis based on the document application scenario.

[0034] Based on the application scenario of the document, the preset enterprise database is constructed by performing reverse optimization of semantic association through principal component analysis. Specifically, this includes: performing reverse optimization of semantic association through principal component analysis based on the application scenario of the document to generate the optimized semantic recognition result and the current scenario recognition model corresponding to the application scenario of the document; and determining the preset enterprise database based on the optimized semantic recognition result and the current scenario recognition model.

[0035] In real-world applications, when filling out forms, such as travel documents for a specific time period, multiple related data sets are usually generated, such as numerous invoices. However, if the user is only filling out travel documents and only needs travel-related documents, it will increase the user's workload in filtering. If the generated invoice data is directly used for intelligent filling, it will lead to inaccurate form filling.

[0036] In an embodiment of the present specification, before secondary identification is performed according to the voice invoice scenario, it is necessary to establish an expert library in advance. The expert library here can be understood as a rule library. The establishment of the expert library mainly relies on principal component analysis for induction and summary. A plurality of invoice data and invoice information corresponding to each invoice data are obtained in advance. The invoice information includes invoice category information and pre-set itemized data. The itemized data is pre-set data for enterprise users. For example, a certain company requires invoice data, but other companies require consumption records. The initiator of the invoice initiates different invoices. The basic logic corresponding to the invoice includes invoice category information and invoice scenario information such as invoice type, billing unit, etc. Different invoice business scenarios are distinguished and defined. The business scenario mainly distinguishes information including time, invoice type, personal type use processing business preference, etc. The semantic analysis efficiency after voice recognition and the accuracy of semantic results in the process of establishing the expert library are improved, and the expert library corresponding to the real use of the user is correctly matched as much as possible. After the scene is distinguished according to the data bias category and the itemized category of human intervention, the principal component data analysis under the scene is performed according to the scene, so as to associate and find the correct expert library in the actual invoice filling process. The expert library can analyze and preform the scene data according to the corresponding scene, such as the travel expense reimbursement form which can bring out the invoice and message record in the corresponding time period according to the time and place semantics.

[0037] According to the invoice category information in the invoice information corresponding to the invoice data, the scene of each invoice data is distinguished, and the initial scene of each invoice data corresponding to the invoice is determined. According to the pre-set itemized data, the initial scene of the invoice is further subdivided, and the application scene of each invoice data corresponding to the invoice is determined. After the scene is subdivided, the reverse optimization of semantic association is performed according to the principal component analysis method, and the corresponding model of the semantic recognition result and the expert library is confirmed, so as to subsequently identify the voice invoice information input by the user according to the voice invoice scenario through the corresponding model of the expert library, and analyze and preform the scene data according to the corresponding scene.

[0038] According to the voice invoice scenario, the voice invoice information is identified again through the preset expert library, and the current filling data corresponding to the voice invoice scenario is obtained. Specifically, the voice invoice information is identified to obtain invoice data information corresponding to the voice invoice information, wherein the invoice data information includes invoice data time and invoice data type, and the invoice data type includes any one or more of invoice, message record, consumption record and approval record; through the preset expert library, a plurality of invoice source data corresponding to the invoice data information are pulled in a system corresponding to the invoice data information according to the invoice data information; and through the preset expert library, the plurality of invoice source data are matched according to the voice invoice scenario, and the current filling data corresponding to the voice invoice scenario is obtained.

[0039] In an embodiment of the present specification, the voice invoice information is recognized to obtain invoice data information corresponding to the voice invoice information, the invoice data information includes invoice data time and invoice data type, the invoice data type includes any one or more of invoices, message records, consumption records and approval records, for example, the obtained invoice data information is the invoice of September, through the preset expert library, according to the invoice data information, the invoice data information corresponding to the invoice data information is pulled in the system; That is, when the invoice data information obtained is the invoice of September, all invoices in the invoice folder in September are pulled through the expert library, that is, multiple invoice source data. According to the voice invoice scene, the multiple invoice source data are matched through the preset expert library to obtain the current filling data corresponding to the voice invoice scene. For example, after obtaining all invoices in September, since the voice invoice scene of the user is a business trip filling scene, the travel class invoice is obtained from all invoices in September according to the business trip filling scene, and the travel class invoice here is the current filling data.

[0040] In an embodiment of the present specification, the invoice scene classification is identified according to the invoice category information and the invoice scene information, such as the invoice type, the accounting unit, etc., to confirm the voice recognition scene. The components of the voice recognition method are mainly realized by relying on mature third-party recognition library. The semantic recognition training optimizes the semantic recognition training relying on the recognition scene to obtain more suitable semantic recognition results. The expert library matching of the semantic recognition result confirms the corresponding expert library according to the association matching rule.

[0041] Step S103, determining the specified purpose information in the voice invoice information, performing semantic extraction on the specified purpose information to generate a current filling summary.

[0042] Determining the specified purpose information in the voice invoice information, performing semantic extraction on the specified purpose information to generate a current filling summary, specifically including: presetting a plurality of specified semantic fields, wherein the specified semantic field is used to represent the purpose; splitting the voice invoice information to obtain a plurality of voice information; calculating the similarity between the voice field corresponding to the voice information and the specified semantic field; in the plurality of voice fields, determining the specified voice field with a similarity greater than a preset similarity threshold, and taking the specified voice field as the specified purpose information; extracting the semantics in the specified purpose information to generate a current filling summary.

[0043] In the actual filling scene, in addition to the filling data corresponding to the invoice, such as invoice, consumption record and other voucher information, the current filling setting filling summary is also needed.

[0044] In an embodiment of the present specification, a plurality of designated semantic fields for indicating purposes are preset, such as "for" and "because" fields. The voice document information is split, and the splitting rule here can be set according to the needs, aiming to split the voice document information into a plurality of fields to compare the designated semantic fields respectively, calculate the similarity between the voice fields corresponding to the split voice information and the designated semantic fields, and select the designated voice field with a similarity greater than a preset similarity threshold. The similarity threshold here is determined according to the calculation method of the similarity, and the calculation method of the similarity can be set according to the user's needs. The obtained designated voice field is taken as the designated purpose information, the semantics in the designated purpose information is extracted, and a document filling summary is generated.

[0045] In step S104, the current document filling data and the current document filling summary are used to intelligently fill the to-be-filled document.

[0046] The current document filling data and the current document filling summary are used to intelligently fill the to-be-filled document, specifically including: according to the current document filling summary, the application matters corresponding to the to-be-filled document are filled in; and according to the current document filling data, the document source data corresponding to the to-be-filled document is perfected, so as to realize intelligent filling of the to-be-filled document.

[0047] In an embodiment of the present specification, the summary is filled in according to the semantic recognition summary or the semantic summary, and at the same time, the intelligent document filling method is triggered to perfect the document information. The current document filling data, that is, the invoice and consumption record information, is perfected into the document source data, and the current document filling summary is perfected into the application matters.

[0048] After the current document filling data and the current document filling summary are used to intelligently fill the to-be-filled document, the method further includes: obtaining the business association data in the current document filling data; and according to the business association data, the associated information of the to-be-filled document is perfected, wherein the perfection of the associated information of the to-be-filled document specifically includes: triggering an expression, perfecting a cost item, and perfecting split information.

[0049] In an embodiment of the present specification, since the application scenario of document filling is enterprise daily management, when filling a document, the filling content of this document may be an expenditure item of a certain project or have other business needs. In order to meet such situations, the business association data in the current document filling data needs to be obtained. The business association data here can be business information or a project name. According to the business association data, the associated information of the to-be-filled document is perfected, such as triggering an expression, perfecting a cost item, and perfecting split information.

[0050] The embodiment of the present specification also provides a specific implementation method corresponding to the above-mentioned embodiment. First, to implement voice recognition and semantic association scenarios, relevant premise parameters need to be set. Define relevant variable information: a bill class identifier BillClass, define a bill type unique identifier BillType, define a bill maker unique identifier PeopleID, a billing unit DEPT, confirm the input parameters of the matching expert library as the formatted output parameters of the recognition result. Moreover, a third-party voice recognition library is used as the running framework of voice recognition. The scenario is used as the distinguishing library of the third-party semantic analysis training library for training in different scenarios.

[0051] Secondly, to implement the expert library and the semantic corresponding relationship of the expert library, an N-dimensional vector needs to be defined, the main table data is directly defined, the sub-table needs to be increased in number of rows to mark the definition, and the project structure data is used as a fixed component; the main component occupies the minimum proportion Components = 0.95. Then, the running framework of principal component analysis is established. Standardize all extracted data to ensure the standard availability of the data; centralize each variable by subtracting the average value of each variable; call a function to calculate the covariance matrix and its eigenvalues and eigenvectors; call an interface to restore the original data set. According to the extracted principal components, new expert library sub-items are obtained. Finally, the summary and the triggered intelligent filling method are implemented, the semantic analysis result extracts sentences containing summaries or inductive properties. According to the voice recognition scenario, the data main input parameter is distinguished as the data main input parameter of the intelligent method, and other data are used as auxiliary variables.

[0052] Through the above technical solution, the voice bill information input by the user's voice is used to determine the voice bill scenario, the expert library is used to recognize the bill data according to the voice bill scenario, the current filling data corresponding to the scenario is obtained, the semantic is implemented, that is, the scenario has the surrounding information brought out by the expert library, finally the intelligent filling method is automatically triggered, the filling error rate and the initial review return rate are reduced, the full-time billing training cost is reduced, the billing efficiency is improved, the user's cumbersome operation steps are avoided, and the user's demand is met.

[0053] The embodiment of the present specification also provides an intelligent filling device based on multi-scene semantic analysis, as shown in Figure 2 The device includes at least one processor and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0054] The voice document information of the user to-be-filled document is acquired, the voice document information is preliminarily identified, and the voice document scene corresponding to the voice document information is generated; the voice document information is identified again according to the voice document scene through a preset expert library corresponding to the voice document scene, and the current filling data corresponding to the voice document scene is acquired; the specified purpose information in the voice document information is determined, the semantic extraction of the specified purpose information is performed, and the current filling summary is generated; and the to-be-filled document is intelligently filled through the current filling data and the current filling summary.

[0055] The embodiments of the present specification also provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions are configured to:

[0056] The voice document information of the user to-be-filled document is acquired, the voice document information is preliminarily identified, and the voice document scene corresponding to the voice document information is generated; the voice document information is identified again according to the voice document scene through a preset expert library corresponding to the voice document scene, and the current filling data corresponding to the voice document scene is acquired; the specified purpose information in the voice document information is determined, the semantic extraction of the specified purpose information is performed, and the current filling summary is generated; and the to-be-filled document is intelligently filled through the current filling data and the current filling summary.

[0057] Each of the embodiments in the present specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0058] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0059] The device and medium provided by the embodiments of the present specification are one-to-one corresponding to the method, and therefore, the device and medium also have similar beneficial technical effects to the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here.

[0060] Those skilled in the art will appreciate that embodiments of the present description can be readily used as a method, a system or a computer program product. Accordingly, the present description can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0061] The present description is described in reference to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present description. It will be understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams 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 processing device 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, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks.

[0062] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks.

[0063] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks.

[0064] In one typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0065] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory. The memory can also include non-volatile memory, such as read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable read-only memory (EEPROM), flash memory, or a combination of non-volatile memories in different forms. The memory is an example of computer readable storage media.

[0066] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology for storing information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0067] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or other elements inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0068] The above description is only one or more embodiments of the specification and is not intended to limit the specification. One or more embodiments of the specification can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of one or more embodiments of the specification should be included in the scope of the claims of the specification.

Claims

1. An intelligent form-filling method based on multi-scenario semantic analysis, characterized in that, The method includes: Obtain the voice form information of the user's form to be filled in, perform preliminary recognition on the voice form information, and generate the voice form scene corresponding to the voice form information; Based on the voice document scenario, the voice document information is re-identified through a preset expert database corresponding to the voice document scenario to obtain the current filling data corresponding to the voice document scenario; Determine the specified destination information in the voice form information, perform semantic extraction on the specified destination information, and generate the current form filling summary; The method further includes: intelligently filling out the form to be filled using the current form data and the current form summary; and performing secondary recognition of the voice form information using a preset expert database based on the voice form scenario. Multiple document data and document information corresponding to each document data are acquired in advance, wherein the document information includes document category information and pre-set project data; Based on the document category information in the document information corresponding to each document data, the scenario of each document data is distinguished to determine the initial scenario of each document data. Based on the preset project data, the initial scenario of the document is further subdivided to determine the document application scenario corresponding to each document data. Based on the application scenario of the document, the preset expert database is constructed by reverse optimization of semantic association through principal component analysis. Based on the aforementioned voice document scenario, a secondary recognition is performed on the voice document information using a preset expert database to obtain the current form-filling data corresponding to the voice document scenario, specifically including: The voice receipt information is identified to obtain the receipt data information corresponding to the voice receipt information. The receipt data information includes receipt data time and receipt data type. The receipt data type includes any one or more of invoices, message records, consumption records and approval records. Using the preset expert database, and based on the document data information, multiple document source data corresponding to the document data information are retrieved from the system corresponding to the document data information. Using the preset expert database, the multiple document source data are matched according to the voice document scenario to obtain the current form filling data corresponding to the voice document scenario.

2. The intelligent form-filling method based on multi-scenario semantic analysis according to claim 1, characterized in that, The voice document information is initially identified, and a corresponding voice document scenario is generated, specifically including: Extract key semantic information from the voice document information, wherein the key semantic information is used to represent the application scenario of the document; The key semantic information is identified to generate scene semantics; Based on the scene semantics and the preset first mapping relationship, the voice document scene corresponding to the scene semantics is determined, wherein the first mapping relationship includes multiple scene semantics and the voice document scene corresponding to each scene semantics.

3. The intelligent form-filling method based on multi-scenario semantic analysis according to claim 1, characterized in that, Determine the specified destination information in the voice form information, perform semantic extraction on the specified destination information, and generate a current form summary, specifically including: Multiple specified semantic fields are preset, wherein the specified semantic fields are used to indicate the purpose; The voice document information is split into multiple voice information segments; Calculate the similarity between the speech field corresponding to the speech information and the specified semantic field; Among multiple speech fields, a specific speech field with a similarity greater than a preset similarity threshold is identified, and the specified speech field is used as the specified target information; Extract the semantics from the specified target information to generate the current form summary.

4. The intelligent form-filling method based on multi-scenario semantic analysis according to claim 1, characterized in that, Intelligent form filling is performed on the form to be filled using the current form data and the current form summary, specifically including: Based on the current form summary, fill in the application items corresponding to the form to be filled; Based on the current form data, the source data of the form to be filled is improved to achieve intelligent form filling of the form to be filled.

5. The intelligent form-filling method based on multi-scenario semantic analysis according to claim 1, characterized in that, After intelligently filling out the form to be filled using the current form data and the current form summary, the method further includes: Retrieve business-related data from the current form data; Based on the business-related data, the association information of the form to be filled is improved. Specifically, improving the association information of the form to be filled includes: triggering expressions, improving expense items and allocation information.

6. The intelligent form-filling method based on multi-scenario semantic analysis according to claim 1, characterized in that, Based on the document application scenario, the preset expert database is constructed by performing reverse optimization of semantic association using principal component analysis, specifically including: Based on the document application scenario, reverse optimization of semantic association is performed using principal component analysis to generate optimized semantic recognition results and current scenario recognition models corresponding to the document application scenario. Based on the optimized semantic recognition results and the current scene recognition model, the preset expert database is determined.

7. An intelligent form-filling device based on multi-scenario semantic analysis, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-6.

8. A non-volatile computer storage medium storing computer-executable instructions, the computer-executable instructions being configured to perform the method as described in any one of claims 1-6.

Images