Business data accounting rule generation method, business data accounting method and business data accounting system

Abstract

The embodiment of the invention provides a business data accounting rule generation method based on a large language model, and a business data accounting method and system. When a business data accounting rule is generated, business data accounting demand information in a natural language form is obtained, and the business data accounting demand information comprises business process description information related to business data accounting and accounting rule description information; and then the business data accounting demand information is provided for an accounting rule generation model to generate a business data accounting rule in a structured data form, and the business data accounting rule is used for business data accounting.

Description

Technical Field

[0001] The embodiments of this specification generally relate to the field of artificial intelligence, and in particular to a method for generating business data accounting rules based on a large language model, a business data accounting method and a system. Background Art

[0002] In the course of business operations, it is necessary to perform business data accounting on the business data (e.g., accounting data) generated in the course of operations. Business data accounting is a data processing process that records, classifies, and calculates business data such as accounting data, and then converts it into entry data such as accounting entries. Business data accounting is usually implemented with the help of a business data accounting engine. A business data accounting engine is a translator for converting business document data into entry data. When performing business data accounting, it is necessary to convert the business data accounting semantic information sent by the user into business data accounting rules that can be understood by the business data accounting engine, and then provide the business data accounting rules to the business data accounting engine for entry data translation. Since the business data accounting process requires business understanding and rule conversion of complex business data, it is necessary to rely on more program logic and program variables to express the business data accounting rules. The expression of business data accounting rules by program logic and program variables is highly technical, so that the generation of business data accounting rules needs to be dominated by technical personnel, which makes it impossible for business personnel to independently perform business data accounting. Summary of the invention

[0003] The embodiments of this specification provide a method for generating business data accounting rules based on a large language model, a business data accounting method and a system. Using the business data accounting rule generation scheme, the human-like understanding ability of the large language model can be used to convert business data accounting semantic information in natural language form into business data accounting rules in technical language form, so as to be used by the business data accounting engine to calculate and process business data, thereby achieving de-technification in the business data accounting process, thereby reducing the cost of business data accounting and improving the independence and convenience of business personnel in business data accounting.

[0004] According to one aspect of an embodiment of the present specification, a method for generating business data accounting rules based on a large language model is provided, including: obtaining business data accounting requirement information in natural language form, the business data accounting requirement information including business process description information and accounting rule description information involved in business data accounting; and providing the business data accounting requirement information to an accounting rule generation model to generate business data accounting rules in structured data form for use by a business data accounting engine to perform business data accounting, the accounting rule generation model including a large language model.

[0005] Optionally, in an example of the above aspect, the business data accounting rule includes string data in JSON format.

[0006] Optionally, in an example of the above aspect, obtaining business data accounting requirement information in natural language form may include: obtaining business data accounting requirement information in natural language form from multimodal input data, wherein the multimodal input data includes user accounting requirement data and business data to be processed input in voice format, picture format and / or text format.

[0007] Optionally, in an example of the above aspect, the business data accounting requirement information is formed into a business data accounting requirement template, and obtaining the business data accounting requirement information in natural language form from multimodal input data includes: obtaining template element values ​​of template elements in the business data accounting requirement template from the multimodal input data, and filling the obtained template element values ​​into the business data accounting requirement template to generate the business data accounting requirement information.

[0008] Optionally, in an example of the above aspect, providing the business data accounting requirement information to an accounting rule generation model to generate business data accounting rules in the form of structured data may include: providing the business data accounting requirement information and similar business data accounting requirement information retrieved from a business data accounting requirement information library to the accounting rule generation model to generate business data accounting rules in the form of structured data.

[0009] Optionally, in an example of the above aspect, the business data accounting rule generation method may further include: engineering the generated business data accounting rules and storing them in a business data accounting rule configuration library of the business data accounting engine.

[0010] Optionally, in an example of the above aspect, the business data accounting includes financial accounting data accounting, and the business data accounting engine includes an accounting engine.

[0011] Optionally, in an example of the above aspect, the accounting rule generation model includes a business scenario recognition model, an accounting event recognition model, and a segment value rule recognition model. Providing the business data accounting requirement information to the accounting rule generation model to generate the business data accounting rules in the form of structured data may include: providing the business data accounting requirement information to the business scenario recognition model to identify the transaction relationship composition and the settlement feature information under each group of transaction relationships; providing the transaction relationship composition and the settlement feature information under each group of transaction relationships to the accounting event recognition model to identify the accounting event description of the accounting events involved in the business data accounting; and providing the accounting event description information to the segment value rule recognition model to identify the segment value rules of each accounting event.

[0012] According to another aspect of an embodiment of the present specification, a business data accounting method based on a large language model is provided, including: obtaining business data to be accounted for; and performing business data accounting on the business data to be accounted for using business data accounting rules via a business data accounting engine, wherein the business data accounting rules are generated according to the business data accounting rule generation method as described above.

[0013] According to another aspect of an embodiment of the present specification, there is provided a device for generating business data accounting rules based on a large language model, including: an accounting requirement information acquisition unit, which acquires business data accounting requirement information in a natural language form, wherein the business data accounting requirement information includes business process description information and accounting rule description information involved in business data accounting; and an accounting rule generation unit, which provides the business data accounting requirement information to an accounting rule generation model to generate business data accounting rules in a structured data form, for use by a business data accounting engine to perform business data accounting, wherein the accounting rule generation model includes a large language model.

[0014] According to another aspect of an embodiment of the present specification, there is provided a business data accounting system based on a large language model, comprising: a business data acquisition device for acquiring business data to be accounted for; a business data accounting rule generation device as described above; and a business data accounting device for performing business data accounting on the business data to be accounted for using the business data accounting rules generated by the business data accounting rule generation device.

[0015] According to another aspect of an embodiment of the present specification, there is provided a device for generating business data accounting rules based on a large language model, comprising: at least one processor; a memory coupled to the at least one processor; and a computer program stored in the memory, wherein the at least one processor executes the computer program to implement the business data accounting rule generation method as described above.

[0016] According to another aspect of an embodiment of the present specification, there is provided a business data accounting system based on a large language model, comprising: at least one processor; a memory coupled to the at least one processor; and a computer program stored in the memory, wherein the at least one processor executes the computer program to implement the business data accounting method as described above.

[0017] According to another aspect of an embodiment of the present specification, a computer-readable storage medium is provided, which stores executable instructions, which, when executed, enable a processor to execute the business data accounting rule generation method described above or execute the business data accounting method described above.

[0018] According to another aspect of an embodiment of the present specification, a computer program product is provided, including a computer program, wherein the computer program is executed by a processor to implement the business data accounting rule generation method as described above or to implement the business data accounting method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] A further understanding of the nature and advantages of the content of this specification may be achieved by referring to the following drawings. In the drawings, similar components or features may have the same reference numerals.

[0020] Figure 1 An example architecture diagram of a business data accounting system according to an embodiment of this specification is shown.

[0021] Figure 2 An exemplary schematic diagram showing an implementation scheme of a financial accounting process is shown.

[0022] Figure 3 An exemplary flow chart of a financial accounting method according to an embodiment of the present specification is shown.

[0023] Figure 4 An example schematic diagram of a large language model based on the Transformer architecture according to an embodiment of the present specification is shown.

[0024] Figure 5 An example flowchart of a process for generating business data accounting rules according to an embodiment of this specification is shown.

[0025] Figure 6 An example schematic diagram of a process for generating business data accounting requirement information according to an embodiment of this specification is shown.

[0026] Figure 7 An example schematic diagram of a business data accounting rule according to an embodiment of this specification is shown.

[0027] Figure 8 An example flowchart of a process for generating business data accounting rules in a financial accounting scenario according to an embodiment of this specification is shown.

[0028] Fig. 9 An example schematic diagram of a business scenario identification process according to an embodiment of this specification is shown.

[0029] Fig.10 An example schematic diagram of a live broadcast scenario is shown.

[0030] Fig.11 An example schematic diagram of an accounting event identification process according to an embodiment of the present specification is shown.

[0031] Fig.12An example schematic diagram of a segment value rule identification process according to an embodiment of the present specification is shown.

[0032] Fig.13 An example schematic diagram of a process for generating business data accounting rules based on retrieval enhancement generation according to an embodiment of the present specification is shown.

[0033] Fig.14 An example block diagram of a business data accounting rule generating device according to an embodiment of this specification is shown.

[0034] Fig.15 An example schematic diagram of a business data accounting rule generation device implemented based on a computer system according to an embodiment of the present specification is shown.

[0035] Fig.16 An example schematic diagram of a business data accounting system implemented based on a computer system according to an embodiment of the present specification is shown. DETAILED DESCRIPTION

[0036] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and is not a limitation of the scope of protection, applicability or examples set forth in the claims. The functions and arrangements of the elements discussed can be changed without departing from the scope of protection of the contents of this specification. Various examples can omit, replace or add various processes or components as needed. For example, the described method can be performed in an order different from the described order, and various steps can be added, omitted or combined. In addition, the features described relative to some examples can also be combined in other examples.

[0037] As used herein, the term "including" and its variations represent open terms, meaning "including but not limited to". The term "based on" means "based at least in part on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other definitions may be included below, whether explicit or implicit. Unless the context clearly indicates otherwise, the definition of a term is consistent throughout the specification.

[0038] The flowcharts used in this specification illustrate the operations implemented by the system according to some embodiments in this specification. It should be clearly understood that the operations of the flowcharts may not be implemented in sequence. On the contrary, the operations may be implemented in reverse order or simultaneously. In addition, one or more other operations may be added to the flowchart. One or more operations may be removed from the flowchart.

[0039] The following will describe a method for generating business data accounting rules based on a large language model, a business data accounting method, an apparatus, and a system according to an embodiment of the present specification with reference to the accompanying drawings.

[0040] Figure 1 An example schematic diagram of a business data accounting system 100 according to an embodiment of the present specification is shown.

[0041] like Figure 1 As shown, the business data accounting system 100 includes a business data accounting rule generating device 110, a business data accounting device 120, a business data accounting rule storage device 130 and a business data accounting engine 140. The business data accounting rule generating device 110, the business data accounting device 120, the business data accounting rule storage device 130 and the business data accounting engine 140 can communicate with each other through a network 150 to transmit data. In some embodiments, the network 150 can be any one or more of a wired network or a wireless network. Examples of the network 150 may include, but are not limited to, a cable network, an optical fiber network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigZee network (ZigZee), a near field communication (NFC), an in-device bus, an in-device line, etc. or any combination thereof. In some embodiments, some or all of the components in the business data accounting rule generating device 110 , the business data accounting device 120 , the business data accounting rule storage device 130 , and the business data accounting engine 140 may communicate directly with each other without the need for the network 150 .

[0042] The business data accounting rule generating device 110 can generate business data accounting rules in the form of structured data according to the business data accounting demand information in the form of natural language. The generated business data accounting rules can be stored in the business data accounting rule storage device 130 for use by the business data accounting device 120 to perform business data accounting. In some embodiments, the business data accounting device 130 can provide the business data to be accounted for to the business data accounting engine 140 to perform business data accounting based on the generated business data accounting rules. In some examples, the business data accounting device 130 can be integrated with the business data accounting engine 140. In other words, the business data accounting device 130 can be implemented as a business data accounting engine 140. In some embodiments, the generated business data accounting rules can also be directly configured in the business data accounting device 120 or the business data accounting engine 140 without the business data accounting rule storage device 130.

[0043] In the embodiments of this specification, the term "business data" may be used to refer to data generated in a business process. Business data may include, for example, form data generated in a business process. For example, when the business process is a financial settlement process, the business data includes financial accounting data, such as bill data, invoice data, and funding data. In this case, the business data accounting engine 140 may be implemented as an accounting engine. The accounting engine is responsible for extracting and converting the financial accounting data, thereby generating accounting entry data.

[0044] For the convenience of description, the financial accounting process is taken as an example to illustrate the business data accounting process.

[0045] Figure 2 An exemplary schematic diagram showing an implementation scheme of a financial accounting process is shown.

[0046] The extraction and conversion of financial business data is a technical data processing process. Figure 2 In the financial accounting scheme shown, before financial accounting is performed, the financial accounting requirements (semantic information of financial accounting requirements in natural language) of the financial personnel need to be converted into business data accounting rules and financial accounting rules in the form of structured data by technical personnel. Subsequently, the converted financial accounting rules are configured on the financial accounting engine.

[0047] The financial accounting engine uses financial accounting rules to extract and transform financial business data such as billing data, invoice data, and funding data to generate accounting entries, such as debit entry data and credit entry data.

[0048] Financial accounting rules are used to specify which elements on business documents need to be extracted and which values ​​need to be converted to accounting entries. Financial accounting rules can include, for example, identity configuration (identity accounting rules), element configuration (element accounting rules), and segment value rule configuration. For example, the company segment of an accounting entry can be confirmed based on the payment and collection institutions on the business document. These institution information is usually placed on the context variables $recon_inst or $inst_id of the business document. It is necessary to configure direct value rules on the financial accounting engine to convert it to the company segment. The account segment on the accounting entry can be confirmed based on the associated transaction type ($trade_type) on the business document. When the associated transaction type is an external associated party, it is mapped to accounts receivable-external customer account. When the associated transaction type is an internal associated party, it is mapped to accounts receivable-internal transactions account. If there is no associated party type, it can be uniformly mapped to accounts receivable-external customer account. In addition, the journal description business also expects the expression of the year and month of the account period, which is usually "$YYYYMM" or "$YYYY year $MM month".

[0049] From the above, it can be seen that when the financial accounting needs of financial personnel are transformed into technical semantics, the financial accounting needs semantic information needs to be abstracted into financial accounting rules based on program logic and program variables. The expression of financial accounting rules based on program logic and program variables is highly technical, so that the generation of financial accounting rules needs to be dominated by technical personnel, which in turn makes it impossible for the financial business party to independently perform financial accounting of financial business data.

[0050] In view of this, the embodiments of this specification propose a business data accounting rule generation scheme based on a large language model. Using this business data accounting rule generation scheme, the human-like understanding ability of the large language model can be used to convert business data accounting demand information in natural language form into business data accounting rules in technical language form (i.e., structured data form) for use by the business data accounting device (business data accounting engine) to perform business data accounting processing, thereby achieving de-technification in the business data accounting process, reducing business data accounting costs, and improving the independence and convenience of business data accounting of the business party.

[0051] Figure 3 An example flow chart of a business data accounting method 300 according to an embodiment of the present specification is shown.

[0052] like Figure 3 As shown, at 310, the business data to be calculated is obtained. The business data to be calculated can be reported to the business data calculation device by the business party, or pulled from the business party by the business data calculation device. When the business scenario is a financial data calculation scenario, the business data can include accounting data such as financial document data, such as bill data, invoice data, capital data, etc.

[0053] At 320, business data accounting is performed on the business data to be accounted for using the business data accounting rules. In some examples, a business data accounting engine can be used to perform business data accounting on the business data to be accounted for based on the business data accounting rules. The business data accounting rules can be formed as structured data, for example, the business data accounting rules can be formed as string data in JSON format.

[0054] With the development of Large Language Model (LLM), LLM has the reasoning ability of human-like intelligence. LLM can be used to achieve in-depth understanding and analysis of business logic through natural language processing and machine learning, thereby inferring the business data accounting rules in the form of structured data required by the business data accounting engine based on the business data accounting requirements in natural language form, so that technical personnel do not need to be involved when generating business data accounting rules, thereby achieving de-technification in the business data accounting process, reducing business data accounting costs, and improving the independence and convenience of business data accounting for business parties.

[0055] Large language models, also known as large language models, are language models trained on unlabeled text data, and the data sets involved in the training are usually very large. Large language models mainly use deep learning mechanisms to effectively extract various language logics and language organization rules from massive training data, and then use AI algorithms to understand or generate natural language text in actual user needs.

[0056] The typical representative of large language models is GPT3, which is the practice of the Transformer network architecture in the language model scenario. The Transformer architecture is a deep learning model that is mainly used to process sequence data, such as text in natural language processing (NLP). Its core idea is to use the self-attention mechanism to capture long-distance dependencies within the sequence without relying on traditional recurrent neural networks (RNNs) or convolutional neural networks (CNNs).

[0057] Figure 4 An example schematic diagram of a large language model based on the Transformer architecture according to an embodiment of the present specification is shown.

[0058] like Figure 4 As shown in the figure, the Transformer architecture includes an input embedding layer (Input Embeddings), an output embedding layer (Output Embeddings), a positional encoding layer (Positional Encoding), an encoder (Encoder) and a decoder (Decoder).

[0059] The input embedding layer is used to convert each element in the input sequence into an embedding vector (embedded representation) of a fixed size. The output embedding layer is deployed in the decoder part to convert the elements in the output sequence into an embedding vector (embedded representation) of a fixed size.

[0060] Since Transformer does not use a recurrent structure, additional position information is required to maintain the order of sequence words. The position encoding layer is used to add the position information of each sequence word to the corresponding embedding vector.

[0061] The encoder is composed of multiple identical layers stacked together, each layer including a multi-head attention layer (Multi-HeadAttention) and a feed-forward neural network layer (Feed-Forward Neural Network). The multi-head attention layer allows the model to simultaneously focus on information at different positions in the sequence. In some examples, the multi-head attention layer can be implemented as a multi-head self-attention network. The feed-forward network layer is used to further transform the output of the multi-head attention layer. In some examples, the feed-forward network layer is implemented as a feed-forward fully connected network. In addition, the outputs of the multi-head attention layer and the feed-forward network layer are further processed via the Add&Norm layer. The Add&Norm layer includes an Add unit and a Norm unit. The Add unit is used to implement residual connections, and the Norm unit is used to implement layer normalization.

[0062] The decoder consists of multiple identical layers. Each layer contains a masked multi-head attention layer, an encoder-decoder attention layer, and a feedforward network layer. The masked multi-head attention layer is used to prevent future information from being used when predicting the current position. The encoder-decoder attention layer is used to allow the decoder to pay attention to the output of the encoder. In some examples, the encoder-decoder attention layer is implemented as a multi-head attention layer. The feedforward network layer is used to further transform the output of the encoder-decoder attention layer. In addition, the outputs of the masked multi-head attention layer and the feedforward network layer are further processed by the Add&Norm layer. The processing results of the decoder are processed by the linear layer and the softmax layer to generate the final output sequence.

[0063] The reasoning process of the large language model is to predict the next output character given a string of input text, and then combine the output character with the input text as the new input, and repeat the process until the output ends. From the mathematical equivalent expression, given a dictionary V = {deep, learning, neural, network, language, model, ...}, ω 1 ,ω 2 ,…,ω n ∈V, calculate P(ω i |ω 1 ,ω 2 ,…,ω i-1 The conditional probability of .

[0064] The pre-trained large language model can be used to generate business data accounting rules for the business data accounting engine.

[0065] Figure 5 An example flow chart of a business data accounting rule generation process 500 according to an embodiment of the present specification is shown.

[0066] like Figure 5 As shown, at 510, business data accounting requirement information in natural language form is obtained, and the obtained business data accounting requirement information includes business process description information and accounting rule description information involved in business data accounting.

[0067] Figure 6 An example schematic diagram of the process of obtaining business data accounting demand information is shown.

[0068] The business data accounting requirement information can be obtained, for example, from unimodal input data or multimodal input input by the user through the input interface of the business data accounting rule generation device or the large language model. For example, computer vision technologies such as OCR and speech recognition or AI technologies such as large models can be used to obtain business data accounting requirement information from unimodal input data or multimodal input data. In some examples, the multimodal input data can be user accounting requirement data and business data to be processed, for example, input in voice format, image format and / or text format. Figure 6 As shown in the business data accounting requirement information, the business process description information includes the complete process in the business operation process (taking the common scenario of Alipay collecting payments for merchants as an example, the complete process of this scenario includes the merchant's fee bill generation, the merchant's payment and recharge, the merchant's bill write-off, etc.), and the accounting rule description information includes the accounting events (account receivable confirmation, prepayment recharge, account receivable write-off), and the COA segment values ​​(account segment, cost center segment, product segment, etc.) calculated.

[0069] In some embodiments, the business data accounting requirement information may be formed as a business data accounting requirement template. When obtaining the business data accounting requirement information in natural language form from the multimodal input data, the template element value of the template element in the business data accounting requirement template may be obtained from the multimodal input data, and then the obtained template element value is filled into the business data accounting requirement template to generate the business data accounting requirement information.

[0070] At 520, the business data accounting requirement information is provided to the accounting rule generation model to generate business data accounting rules for use by (e.g., a business data accounting engine) to perform business data accounting. The accounting rule generation model can be implemented as a large language model. The accounting rule generation model can be pre-trained using a natural language corpus. Figure 6An example schematic diagram of training sample data of a calculation rule generation model according to an embodiment of this specification is shown. After obtaining business data calculation requirement information, the business data calculation requirement information is provided to the trained large language model to generate business data calculation rules. Figure 7 An example schematic diagram of a business data accounting rule according to an embodiment of this specification is shown.

[0071] In some embodiments, the accounting rule generation model may include a multi-level large language model, and the business data accounting demand information is subjected to multi-level reasoning processing via the multi-level large language model to obtain the business data accounting rules. For example, the reasoning results of each level of the large language model may be combined to obtain the final business data accounting rules.

[0072] In some embodiments, when the business scenario is a financial accounting data calculation scenario, the calculation rule generation model includes a business scenario recognition model, an accounting event recognition model, and a segment value rule recognition model. The business scenario recognition model, the accounting event recognition model, and the segment value rule recognition model can all be implemented using a large language model and pre-trained.

[0073] Figure 8 FIG. 8 is a flow chart showing an example of a process 800 for generating business data accounting rules in a financial accounting scenario according to an embodiment of the present specification. Figure 8 In the example, the business data accounting is financial accounting, the business data accounting requirement information is financial accounting requirement information, and the business data accounting rule is a financial accounting rule.

[0074] like Figure 8 As shown, at 810, the financial accounting demand information is provided to the business scenario identification model to identify the transaction relationship composition and the settlement feature information under each group of transaction relationships.

[0075] Fig. 9 FIG. 1 shows an example schematic diagram of a business scenario identification process according to an embodiment of the present specification. Fig. 9In the example shown, financial accounting demand information is submitted to the business scenario recognition model in the form of natural language description or in the form of picture description, and the submitted financial accounting demand information includes business process description information and billing / settlement rule description (i.e., accounting rule description information). In addition, the submitted financial accounting demand information may also include business background description information. The model type of the business scenario recognition model can be selected according to the business scenario involved in the submitted financial accounting demand information. The business scenario recognition model infers a group of transaction relationships and settlement feature information under each group of transaction relationships based on the business process description information and the billing / settlement rule description information. The settlement feature information may include, for example, settlement subject information and settlement object information. In addition, the business scenario recognition model can also infer business feature information for each group of transaction relationships. The inferred transaction relationships, settlement feature information under each group of transaction relationships, and business feature information for each group of transaction relationships are represented by structured data.

[0076] The following uses the live broadcast scenario as an example to illustrate the business scenario identification process.

[0077] Fig.10 An example schematic diagram of a live broadcast scenario is shown.

[0078] like Fig.10 As shown in the figure, in the live broadcast scenario where ordinary users give tips to anchors while watching live broadcasts, there are four roles: live broadcast operation platform, ordinary users, Apple channels and anchors. These four roles form the following three transaction relationships:

[0079] Transaction relationship 1: When ordinary users watch live broadcasts, they will top up and purchase props, thus forming a transaction relationship in which the live broadcast operation platform charges ordinary users;

[0080] Transaction relationship 2: After the anchor receives the reward, a paid transaction relationship is formed between the live broadcast operation platform and the anchor;

[0081] Transaction relationship 3: Since the live broadcast is on Apple phones, Apple will also deduct part of the channel costs, thus forming a paid transaction relationship between the live broadcast operation platform and the Apple channel.

[0082] For each transaction relationship, the business characteristic information under the transaction relationship can also be inferred. For example, for transaction relationship 1, the inferred business characteristic information may include demand name, demand description, department, business scenario, product code, business identity code, settlement subject, settlement object, settlement method, payment method, settlement time, whether prepaid, whether provisioned, product category, etc.

[0083] At 820, the transaction relationship composition and the settlement characteristic information under each group of transaction relationships are provided to the accounting event recognition model to identify the accounting event descriptions of the accounting events involved in the business data accounting.

[0084] Fig.11 FIG. 2 shows an example schematic diagram of an accounting event identification process according to an embodiment of the present specification. Fig.11 As shown, based on the transaction relationship and settlement characteristics identified by the business scenario identification process, the accounting event identification model can infer the accounting events involved. Taking "live broadcast" as an example, transaction relationship 1 is that the live broadcast operation platform charges user C. Combined with the settlement feature information inferred by the business scenario identification process, the rules formulated in the rule engine can be used to infer all accounting events. The accounting event identification results can include, for example, transaction relationships, settlement actions, accounting events, rule expressions, etc.

[0085] In some embodiments, the accounting event recognition model can be implemented using a large language model. In some embodiments, since the Key and Value of the business feature information are fixed enumeration values, a rule engine can also be selected to implement the accounting event recognition model.

[0086] At 830, the accounting event description information is provided to a segment value rule identification model to identify segment value rules for each accounting event.

[0087] In the accounting event recognition process, all accounting events are inferred. The inferred accounting events only have a data structure framework of structured data. The specific element value of the data element in each data structure needs to be further deduced based on the business characteristic information of the accounting events and business scenarios.

[0088] Fig.12 FIG. 2 shows an example schematic diagram of a segment value rule recognition process according to an embodiment of the present specification. Fig.12 As shown, for each accounting event, the segment value rule recognition model can recommend corresponding flexible segment rules under the current business characteristic values ​​by learning the business characteristic values ​​of historical accounting plans.

[0089] In some embodiments, in order to enhance the accuracy and richness of the generated business data accounting rules, a business data accounting rule method based on retrieval enhancement generation can be used to generate business data accounting rules. In the process of generating business data accounting rules based on retrieval enhancement generation, business data accounting requirement information and similar business data accounting requirement information retrieved from the business data accounting requirement information library are provided to the accounting rule generation model to generate business data accounting rules in the form of structured data. In this business data accounting rule generation method, not only does it rely on the knowledge of the accounting rule generation model itself, but it also enhances the accuracy and richness of the generated content by retrieving relevant information from an external knowledge base.

[0090] Fig.13FIG. 1 shows an example schematic diagram of a process for generating business data accounting rules based on retrieval enhancement generation according to an embodiment of this specification. Fig.13 As shown, the business data accounting rule generation process based on retrieval enhancement generation can include four components, namely, the historical sample library generation process, the current sample generation process, the vectorized storage process and the large model reasoning process. Here, the historical sample is equivalent to the sample constructed based on the historical business data accounting demand information, and the current sample is equivalent to the sample constructed based on the current business data accounting demand information.

[0091] In the process of generating the historical sample library, a historical sample library that can be used for learning is generated based on the historical business data accounting requirements, business data, and historical business data accounting rules. Specifically, the historical business data accounting requirements are processed in blocks (for example, segmented by paragraphs) to screen out useful text blocks, and then the screened text blocks are integrated with the business data (for example, form data of financial accounting data, etc.) and the historical business data accounting rules to form historical samples. When the historical business data accounting requirements are processed in blocks, at least one of character blocks (CharacterTextSplitter), paragraph blocks (ParagraphTextSplitter), Markdown text blocks (MarkdownTextSplitter) and recursive blocks (RecursiveCharacterTextSplitter) can be used. If the text blocks obtained after these block processing are not cohesive, some text segmentation algorithms can be introduced to perform further text segmentation, such as: sentence-level sequence annotation text segmentation model, etc. For the segmented text blocks, corresponding text summaries can also be generated, and then the text blocks, text summaries, business data, and historical business data accounting rules can be flattened in the transaction dimension, thereby forming historical knowledge for use in large language model reasoning.

[0092] In the current sample generation process, based on the current business data accounting requirements and current business data, the current sample for business data accounting rule reasoning is generated. The current sample generation process is basically similar to the historical sample library generation process, except that the historical business data accounting rules are not used.

[0093] In the vectorized storage process, the historical samples and current samples in the historical sample library are vectorized one by one, and the obtained vectors are stored in the vector library.

[0094] During the large model reasoning process, historical samples similar to the current sample are recalled from the historical sample library, and the recalled historical samples and the current sample are spliced ​​into the model prompt words of the large language model. In the spliced ​​model prompt words, the recalled historical samples have historical accounting rules, while the current sample has no accounting rules. Subsequently, the large language model generates business data accounting rules corresponding to the current business data accounting needs based on the model prompt words enhanced by historical samples.

[0095] In some embodiments, vector similarity may be calculated based on vectors of samples stored in a vector library, and historical samples similar to the current sample may be recalled from the historical sample library based on the calculated vector similarity. The calculated vector similarity may include, for example, cosine similarity.

[0096] Optionally, in some embodiments, after the business data accounting rules are generated via the large language model, the generated business data accounting rules may be stored in a business data accounting rule configuration library of the business data accounting engine after engineering processing.

[0097] Fig.14 FIG. 1 shows an example block diagram of a business data accounting rule generating device 1400 according to an embodiment of this specification. Fig.14 As shown, the business data accounting rule generating device 1400 includes an accounting requirement information acquiring unit 1410 , an accounting rule generating unit 1420 and an accounting rule storing unit 1430 .

[0098] The accounting requirement information acquisition unit 1410 is configured to acquire business data accounting requirement information in natural language form, and the acquired business data accounting requirement information includes business process description information and accounting rule description information involved in business data accounting. The operation and function of the accounting requirement information acquisition unit 1410 can refer to the above reference Figure 5 The operation and functionality of 510 are described.

[0099] The accounting rule generation unit 1420 is configured to provide the business data accounting requirement information to the accounting rule generation model to generate business data accounting rules in the form of structured data for use by the business data accounting engine to perform business data accounting. The operation and function of the accounting rule generation unit 1420 can be referred to in the above reference. Figure 5 520 and Figure 8-Figure 13 Describes the operation and functionality.

[0100] The calculation rule storage unit 1430 is used to store the generated business data calculation rules in the business data calculation rule library after engineering processing. In some embodiments, the calculation rule storage unit may not be required.

[0101] As above Figures 1 to 14 , describes a method for generating business data accounting rules and a device for generating business data accounting rules, as well as a business data accounting method and a business data accounting system according to an embodiment of this specification. The above business data accounting rule generating device and business data accounting system can be implemented by hardware, or by software or a combination of hardware and software.

[0102] Fig.15 FIG. 1 shows an example schematic diagram of a business data accounting rule generating device 1500 implemented based on a computer system according to an embodiment of the present specification. Fig.15 As shown, the business data accounting rule generating device 1500 may include at least one processor 1510, a memory (e.g., a non-volatile memory) 1520, a memory 1530, and a communication interface 1540, and the at least one processor 1510, the memory 1520, the memory 1530, and the communication interface 1540 are connected together via a bus 1560. At least one processor 1510 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above-mentioned element implemented in the form of software).

[0103] In one embodiment, computer executable instructions are stored in a memory, which when executed cause at least one processor 1510 to: obtain business data accounting requirement information in natural language form, the business data accounting requirement information including business process description information and accounting rule description information involved in business data accounting; and provide the business data accounting requirement information to an accounting rule generation model to generate business data accounting rules in structured data form for use by a business data accounting engine to perform business data accounting, the accounting rule generation model including a large language model.

[0104] It should be understood that the computer executable instructions stored in the memory, when executed, cause at least one processor 1510 to perform the above combined operations in various embodiments of the present specification. Figure 4-Figure 14 Describes the various operations and functions.

[0105] Fig.16 FIG. 1 shows an example schematic diagram of a business data accounting system 1600 implemented based on a computer system according to an embodiment of the present specification. Fig.16 As shown, the business data accounting system 1600 may include at least one processor 1610, a memory (e.g., a non-volatile memory) 1620, a memory 1630, and a communication interface 1640, and the at least one processor 1610, the memory 1620, the memory 1630, and the communication interface 1640 are connected together via a bus 1660. At least one processor 1610 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above-mentioned element implemented in the form of software).

[0106] In one embodiment, computer executable instructions are stored in a memory, which, when executed, cause at least one processor 1610 to: obtain business data to be accounted for; and perform business data accounting on the business data to be accounted for using business data accounting rules via a business data accounting engine, wherein the business data accounting rules are generated according to the business data accounting rule generation method as described above.

[0107] It should be understood that the computer executable instructions stored in the memory, when executed, cause at least one processor 1610 to perform the above combined operations in various embodiments of the present specification. Figure 3 Describes the various operations and functions.

[0108] According to one embodiment, a program product such as a machine-readable medium (e.g., a non-transitory machine-readable medium) is provided. The machine-readable medium may have instructions (i.e., the above-mentioned elements implemented in the form of software), which, when executed by a machine, causes the machine to perform the above-mentioned combination of various embodiments of this specification. Figure 1-Figure 14 Specifically, a system or device equipped with a readable storage medium may be provided, on which a software program code implementing the functions of any of the above-mentioned embodiments is stored, and a computer or processor of the system or device reads and executes the instructions stored in the readable storage medium.

[0109] In this case, the program code itself read from the machine-readable medium can realize the function of any one of the above-mentioned embodiments, and thus the machine-readable code and the machine-readable storage medium storing the machine-readable code constitute a part of the present invention.

[0110] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer or a cloud via a communication network.

[0111] According to one embodiment, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the processor performs the above combination of the various embodiments of this specification. Figure 1-Figure 14 Describes the various operations and functions.

[0112] Those skilled in the art should understand that the various embodiments disclosed above can be modified and altered in various ways without departing from the essence of the invention. Therefore, the protection scope of the present invention should be defined by the appended claims.

[0113] It should be noted that not all steps and units in the above-mentioned processes and system structure diagrams are necessary, and some steps or units can be ignored according to actual needs. The execution order of each step is not fixed and can be determined as needed. The device structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0114] In the above embodiments, the hardware unit or module can be realized by mechanical or electrical means. For example, a hardware unit, module or processor can include permanent dedicated circuit or logic (such as special processor, FPGA or ASIC) to complete the corresponding operation. The hardware unit or processor can also include programmable logic or circuit (such as general-purpose processor or other programmable processor), which can be temporarily set by software to complete the corresponding operation. Specific implementation (mechanical method or dedicated permanent circuit or temporary circuit) can be determined based on cost and time consideration.

[0115] The specific embodiments described above in conjunction with the accompanying drawings describe exemplary embodiments, but do not represent all embodiments that can be implemented or fall within the scope of protection of the claims. The term "exemplary" used throughout this specification means "used as an example, instance or illustration" and does not mean "preferred" or "having advantages" over other embodiments. For the purpose of providing an understanding of the described technology, the specific embodiments include specific details. However, these technologies can be implemented without these specific details. In some instances, in order to avoid making the concepts of the described embodiments difficult to understand, well-known structures and devices are shown in block diagram form.

[0116] The above description of the present disclosure is provided to enable any person of ordinary skill in the art to implement or use the present disclosure. Various modifications to the present disclosure will be apparent to those of ordinary skill in the art, and the general principles defined herein may be applied to other variations without departing from the scope of protection of the present disclosure. Therefore, the present disclosure is not limited to the examples and designs described herein, but is consistent with the widest range of principles and novel features disclosed herein.

Claims

1. A method for generating business data accounting rules based on a large language model, comprising: Acquire business data accounting requirement information in natural language format, wherein the business data accounting requirement information includes business process description information and accounting rule description information involved in business data accounting; as well as The business data accounting requirement information is provided to an accounting rule generation model to generate business data accounting rules in the form of structured data for use by a business data accounting engine to perform business data accounting. The accounting rule generation model includes a large language model.

2. The method for generating business data accounting rules according to claim 1, wherein: The business data accounting rules include string data in JSON format.

3. The method for generating business data accounting rules according to claim 1, wherein: The business data accounting demand information obtained in natural language format includes: Business data accounting requirement information in natural language form is obtained from multimodal input data, where the multimodal input data includes user accounting requirement data and business data to be processed input in voice format, image format and / or text format.

4. The method for generating business data accounting rules according to claim 1, wherein: The business data accounting requirement information is formed into a business data accounting requirement template, and the business data accounting requirement information in natural language form is obtained from the multimodal input data, including: The template element values ​​of the template elements in the business data accounting requirement template are obtained from the multimodal input data, and the obtained template element values ​​are filled into the business data accounting requirement template to generate business data accounting requirement information.

5. The method for generating business data accounting rules according to claim 1, wherein: Providing the business data accounting requirement information to the accounting rule generation model to generate business data accounting rules in the form of structured data includes: The business data accounting requirement information and similar business data accounting requirement information retrieved from the business data accounting requirement information library are provided to the accounting rule generation model to generate business data accounting rules in the form of structured data.

6. The method for generating business data accounting rules according to claim 1, further comprising: The generated business data accounting rules are processed by engineering and stored in the business data accounting rule configuration library of the business data accounting engine.

7. The method for generating business data accounting rules according to any one of claims 1 to 6, wherein: The business data calculation includes financial accounting data calculation, and the business data calculation engine includes an accounting engine.

8. The method for generating business data accounting rules according to claim 7, wherein: The accounting rule generation model includes a business scenario recognition model, an accounting event recognition model and a segment value rule recognition model. Providing the business data accounting requirement information to the accounting rule generation model to generate business data accounting rules in the form of structured data includes: Providing the business data accounting demand information to the business scenario recognition model to identify the transaction relationship composition and the settlement feature information under each group of transaction relationships; Providing the transaction relationship composition and the settlement characteristic information under each group of transaction relationships to the accounting event recognition model to identify the accounting event descriptions of the accounting events involved in the business data accounting; and The accounting event description information is provided to a segment value rule recognition model to identify the segment value rules of each accounting event.

9. A business data accounting method based on a large language model, comprising: Obtain the business data to be calculated; as well as The business data to be calculated is calculated by the business data calculation engine using the business data calculation rules. The business data accounting rules are generated according to the business data accounting rule generation method as described in any one of claims 1 to 8.

10. A device for generating business data accounting rules based on a large language model, comprising: An accounting requirement information acquisition unit is configured to acquire business data accounting requirement information in a natural language format, wherein the business data accounting requirement information includes business process description information and accounting rule description information involved in business data accounting; as well as The accounting rule generation unit provides the business data accounting requirement information to the accounting rule generation model to generate business data accounting rules in the form of structured data for use by the business data accounting engine to perform business data accounting. The accounting rule generation model includes a large language model.

11. A business data accounting system based on a large language model, comprising: A business data acquisition device for acquiring business data to be calculated; The business data accounting rule generating device as claimed in claim 10; and The business data accounting device uses the business data accounting rules generated by the business data accounting rule generating device to perform business data accounting on the business data to be accounted for.

12. A device for generating business data accounting rules based on a large language model, comprising: at least one processor; a memory coupled to the at least one processor; as well as The computer program stored in the memory, and the at least one processor executes the computer program to implement the business data accounting rule generation method as described in any one of claims 1 to 8.

13. A business data accounting system based on a large language model, comprising: at least one processor; a memory coupled to the at least one processor; as well as The computer program stored in the memory, and the at least one processor executes the computer program to implement the business data accounting method as claimed in claim 9.

14. A computer-readable storage medium storing executable instructions, which, when executed, enable a processor to execute the business data accounting rule generation method as described in any one of claims 1 to 8 or the business data accounting method as described in claim 9.

15. A computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the business data accounting rule generation method as described in any one of claims 1 to 8 or the business data accounting method as described in claim 9.

Images