A method for extracting standardized product units based on commodity details data in invoices

Abstract

The present invention discloses a method for extracting standardized product units based on commodity detail data in an invoice, which belongs to the field of big data processing and specifically comprises: first, defining an industrial chain and determining the scope of the industrial chain; and initially building a knowledge-based industrial chain structure; then, determining the basic SPU of each link in each commodity / service link in the upstream, midstream and downstream of the industrial chain, that is, the common name of the commodity link; then, compiling a policy statement for the commodity / service link and determining the invoice code; in a superset internal system platform, searching for invoice detail data related to the commodity / service link through a MySQL statement, extracting the standardized product unit of the invoice detail data after removing the impurity invoice data, and finally, marking the invoice detail data with an SPU tag; the present invention has the advantages of fast extraction and batch acquisition, and improves data analysis efficiency and increases the significance of data application fields.

Description

Technical Field

[0001] The present invention belongs to the field of commercial or financial data processing and relates to big data processing, and specifically is a method for extracting standardized product units based on commodity detail data in an invoice. Background Art

[0002] Invoices issued between enterprises are a type of third-party evidence used to verify the transaction relationship between enterprises, and are called transaction data. In addition to invoice data, customs data, contract information, winning bid information, listed company announcements, and e-commerce platform data can also reflect inter-enterprise trade and the traded goods involved in the trade. Especially in the field of manufacturing and circulation, invoice data issued between enterprises can better restore the real trade situation, with advantages such as wide coverage and complete commodity details.

[0003] Invoice data used to be mostly paper archives. With the development of the Golden Tax Project, the relevant data has just been digitized in recent years. The product details data in the invoice is a natural language description of the smallest stock keeping unit (SKU), including the brand, color, specification, purpose, model and standardized product unit (SPU) of the product. Since this description is filled out manually by the enterprise during the operation process, the degree of standardization is poor; at the same time, since the business operations of the enterprise involve a very wide range of fields and the product types are extremely complex, it has a strong professional barrier.

[0004] Invoice data is mainly used in the commercial field of the industrial chain and financial institutions. In the past, enterprises in the industrial chain were mounted through well-known well-known enterprises and large enterprises with public information of listed companies. The number of mounted enterprises in each link was dozens, lacking big data support and the presentation of small, medium and micro enterprises. There is an urgent need for a technology that can use invoice data to accurately find the transaction parties, mount tens of thousands of enterprises, mount more and more real industry members for the industrial chain, and cover all types of large, medium, small and micro enterprises.

[0005] Due to the late electronicization of invoice data, the use of big data technology to extract product details data in invoices has not yet been popularized. If product identification is directly performed using big model-related technologies, the results will have a large degree of deviation and cannot be used in application scenarios. Summary of the invention

[0006] In view of the above problems, the present invention proposes a method for extracting standardized product units based on the commodity details data in the invoice. The method adopts big data processing technology to convert non-standardized invoice details data into standardized product units (SPU) through data analysis, data collation and data extraction. The method has the advantages of fast extraction and batch acquisition, as well as improving data analysis efficiency and increasing the significance of data application fields.

[0007] The method for extracting standardized product units based on commodity details data in the invoice has the following specific steps:

[0008] Step 1: Determine and build the industrial chain based on big data models, expert experience and machine learning;

[0009] An industrial chain refers to a structure with a chain-like relationship that includes upstream, midstream and downstream, with core commodities or services as the midstream, extending upward as the upstream, and extending downward as the downstream.

[0010] The upstream and downstream extensions of goods / services are determined through transactions between enterprises, thus forming the upstream and downstream system of the industrial chain, and the goods / services traded between enterprises are defined as SPU.

[0011] Step 2: Obtain synonyms for each product / service link in the upstream, midstream and downstream of the industrial chain through semantic recognition and knowledge learning. Each group of synonyms constitutes a group of SPUs.

[0012] Step 3: Determine the most simplified name of the goods / services in each group of SPUs as the basic SPU of the group;

[0013] The method to determine the base SPU is as follows:

[0014] (1) Use the results of the SPU strategy test to find the extended basic SPU, and group the extended basic SPU into multiple basic SPUs corresponding to the SPU;

[0015] (2) Using synonyms of goods / services as basic SPU:

[0016] Step 4: Write a policy statement based on the product details of the invoice;

[0017] The invoice includes: the name of the selling company, the name of the purchasing company, the invoice classification code and the product details;

[0018] The specific process is:

[0019] First, write basic policy statements based on product names or basic SPUs;

[0020] It is divided into basic SPU form and modifier form;

[0021] (1) When the basic SPU is not limited, the policy statement directly adopts the basic SPU form:

[0022] (2) When the SPU is in the form of "modifier" + "basic SPU", the policy statement adopts the modifier form;

[0023] Then, the commodity classification code is obtained through the SPU and the corresponding basic SPU, and the basic policy statement is limited according to the commodity classification code to which the SPU belongs.

[0024] Finally, for the products in the industrial chain, the penetration strategy is used to lock the industrial field to which the product is applicable, and the final strategy statement is written in a penetration form:

[0025] 1) Write a policy statement based on the name of the business that buys or sells the SPU;

[0026] 2) Write policy statements based on the products purchased by the SPU enterprise;

[0027] 3) Write policy statements based on the products sold by the purchased SPU enterprises;

[0028] Step 5: For a single invoice, filter it through the policy statement to obtain the SPU that specifically represents the sales of the invoice, that is, extract the standardized product unit of the commodity detail data in the invoice.

[0029] Similarly, for massive invoices, the SPU represented by each invoice can be obtained by filtering based on policy statements, so that the invoices containing the SPU can be aggregated, and then information such as the company that sells / purchases the SPU can be counted.

[0030] The advantages of the present invention are:

[0031] 1) A method for extracting standardized product units based on the commodity details data in the invoice to build a commercialized industrial chain structure, which makes the understanding of the industrial chain more concrete and in line with commercial applications;

[0032] 2) A method for extracting standardized product units based on the commodity details data in the invoice, showing the actual transaction party of a certain commodity based on the invoice data, and providing a more accurate list of corporate members;

[0033] 3) A method for extracting standardized product units based on the commodity details data in the invoice, which can quickly complete the standardized processing of invoice data in batches. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The present invention is a flow chart of a method for extracting standardized product units based on commodity detail data in an invoice. DETAILED DESCRIPTION

[0035] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0036] The present invention discloses a method for extracting standardized product units based on commodity detail data in an invoice. The method adopts big data processing technology and converts non-standardized invoice detail data into standardized product units (SPUs) through data analysis, data sorting and data extraction. The method has the advantages of fast extraction and batch acquisition, improves data analysis efficiency and expands data application fields.

[0037] The method for extracting standardized product units based on commodity details data in the invoice is as follows: Figure 1 As shown, the specific steps are as follows:

[0038] Step 1: Determine and build the industrial chain based on big data models, expert experience and machine learning;

[0039] Different institutions and customer groups have different definitions of the industrial chain. The scope of the industrial chain is determined based on big data models and machine learning to determine the scope for the subsequent structure building process.

[0040] The initial structure of the industrial chain is summarized by using big data, expert experience, and knowledge learning. The industrial chain refers to a structure with a chain-like relationship form, including upstream, midstream, and downstream, with core products or services as the midstream, extending upward as the upstream, and extending downward as the downstream.

[0041] The upstream and downstream extensions of goods / services are determined through transactions between enterprises, thus forming the upstream and downstream system of the industrial chain, and the goods / services traded between enterprises are defined as SPU.

[0042] According to the different types of industrial chains, the upward divergence is raw materials, production equipment, parts or production processes, etc., and the downward extension is the classification, scene application and terminal links of the midstream, etc. By building an industrial chain map based on the characteristics of different industries and upstream, midstream and downstream, we can clearly see the production process of the chemical industry, the composition of automotive parts or the application of lithium batteries.

[0043] For example, a company that sells complete vehicles purchases automobile engines, and the automobile engines are the upstream link of the complete vehicle;

[0044] Step 2: Obtain synonyms for each product / service link in the upstream, midstream and downstream of the industrial chain through semantic recognition and knowledge learning. Each group of synonyms constitutes a group of SPUs.

[0045] Through semantic recognition, synonyms of products are obtained. For example, the synonym of water cup is cup. When invoicing, there will be multiple synonyms due to manual writing. By obtaining synonyms, different names of the same product can be locked, and as many invoice-related details as possible can be obtained. For example, chip, integrated circuit and IC are synonyms, which together form a group of SPUs;

[0046] Step 3: Determine the most simplified name of the goods / services in each group of SPUs as the basic SPU of the group;

[0047] Basic SPU is the common name of the product link. For example, chip can also be called integrated circuit, microcircuit, IC. Therefore, "chip", "integrated circuit", "microcircuit" and "IC" can all be used as basic SPUs. For example, the blue glass is the basic SPU. For example, the basic SPU of SPU-car engine is engine.

[0048] The method to determine the base SPU is as follows:

[0049] (1) Use the results of the SPU strategy test to find the extended basic SPU, and group the extended basic SPU into multiple basic SPUs corresponding to the SPU;

[0050] For example: Basic SPU-Engine can find the extended basic SPU as engine assembly or engine group, etc.;

[0051] (2) Using synonyms of goods / services as basic SPU:

[0052] Step 4: Write a policy statement based on the product details of the invoice;

[0053] The invoice includes: the name of the selling company, the name of the purchasing company, the invoice classification code and the product details;

[0054] The strategy anchors the transaction relationship between enterprises at the commodity level. The strategy is written by establishing an association between the commodity name and the invoice details through SQL statements. The strategy includes two situations: "no need to limit the enterprise" and "need to limit the enterprise". "No need to limit the enterprise" includes three forms: basic SPU form, modifier form, noun form and coding form, and "need to limit the enterprise" includes five forms: private sea form, enterprise form, special form, dealer form and subquery form.

[0055] The specific process is:

[0056] First, write basic policy statements based on product names or basic SPUs;

[0057] It is divided into basic SPU form and modifier form;

[0058] (1) When the basic SPU is not limited, the policy statement directly adopts the basic SPU form:

[0059] For example, if SPU is engine, the corresponding basic SPUs are: engine, engine group, engine assembly, as strategy statements;

[0060] (2) When the SPU is in the form of "modifier" + "basic SPU", the policy statement adopts the modifier form;

[0061] For example, if SPU is a car engine, the corresponding basic SPU is engine, engine unit, and engine assembly, and the strategy statement is "car + engine" or "car + engine unit" or "car + engine assembly".

[0062] Then, the commodity classification code is obtained through the SPU and the corresponding basic SPU, and the basic policy statement is limited according to the commodity classification code to which the SPU belongs.

[0063] The commodity classification code uniformly compiled by the tax authorities based on the commodity type; for example, automobile wheels can correspond to both the commodity classification code of automobile parts and the commodity classification code of metal products;

[0064] Finally, for the products in the industrial chain, the penetration strategy is used to lock the industrial field to which the product is applicable, and the final strategy statement is written in a penetration form:

[0065] If the SPU is a universal component or raw material, it is necessary to accurately locate the universal component or raw material in a specific industrial chain through a penetration strategy;

[0066] 1) Write policy statements based on the names of companies that purchase or sell SPUs. For example, the SPU in the automotive industry chain is the engine. To ensure that companies that sell engines meet the requirements of the automotive industry chain, limit the names of companies that sell or purchase engines to include "automobile";

[0067] 2) Write policy statements based on the goods purchased by the companies selling SPUs; for example, the SPU in the bio-based material industry chain is plastic bags. In order to ensure that companies selling plastic bags meet the requirements of the bio-based material industry chain, limit the companies selling plastic bags to those that have purchased bio-based raw materials such as straw, cassava, and corn;

[0068] 3) Write policy statements based on the products sold by the companies that purchase the SPU. For example, if the SPU in the automotive industry chain is the engine, in order to determine that the companies that sell engines meet the requirements of the automotive industry chain, it is limited to companies that purchase the engines sold by the above companies to sell cars.

[0069] Step 5: For a single invoice, filter it through the policy statement to obtain the SPU that specifically represents the sales of the invoice, that is, extract the standardized product unit of the commodity detail data in the invoice.

[0070] Similarly, for massive invoices, the SPU represented by each invoice can be obtained by filtering based on policy statements, so that the invoices containing the SPU can be aggregated, and then information such as the company that sells / purchases the SPU can be counted.

[0071] Example:

[0072] Step 1: Build an industrial chain;

[0073] Among them, (a) the composition of the upstream structure

[0074] 1. Classification by upstream first level

[0075] (1) Classification by composition:

[0076] If the name of the industrial chain (usually consistent with the name of the midstream link) is a large physical object, the upstream is generally split according to the components of the large physical object;

[0077] ① For example: UAV industry chain (UAV flight control system, UAV transmission system, UAV power system and other parts)

[0078] ② For example: automobile industry chain (automobile engine, automobile chassis, automobile body)

[0079] ③ For example: photovoltaic power generation system industry chain (photovoltaic modules, photovoltaic brackets, photovoltaic inverters, combiner boxes, photovoltaic batteries, photovoltaic cables, MPPT)

[0080] (2) By category:

[0081] If the name of the industrial chain is a general term for a certain type of physical object, the upstream is generally divided according to the focus of the classification of the general term for the physical object;

[0082] ① For example: Hydrogen energy focuses on classification by production method (the first level of the upstream hydrogen energy industry chain is divided into chemical by-product hydrogen production, methanol reforming hydrogen production, fossil fuel hydrogen production, and water electrolysis hydrogen production)

[0083] ② For example: Superhard materials focus on classification by upstream product types (the first level of the upstream of the superhard materials industry chain is divided into natural diamond, artificial diamond, cubic boron nitride, and carbon nitride)

[0084] ③ For example: Refractory materials are mainly classified by production materials and equipment (the first level of the upstream refractory industry chain is divided into refractory materials, auxiliary materials, and refractory processing equipment)

[0085] (3) Division by process

[0086] If the name of the industry chain refers to products / services provided according to a certain production / service process, the upstream is generally divided according to the process;

[0087] ① For example: IC industry chain (IC design, IC manufacturing, IC packaging and testing);

[0088] ②Green and low-carbon steel industry chain (ironmaking, steelmaking, continuous casting, steel rolling)

[0089] (4) Classification by combination of software and hardware

[0090] If the name of the industrial chain is a larger concept, it is generally considered from two aspects: soft (technology / process / software) and hard (composition): judge the importance of "soft". If the degree of softness is relatively light, classify it by composition; if the degree of softness is relatively heavy, combine soft (technology / process / software) and hard (composition) as the first level of classification.

[0091] ① The high-speed railway industry chain (tracks, railway signals, train sets) and the UHV industry chain (power control end, power transmission) are divided according to their composition;

[0092] ② VR industry chain (key technologies, hardware) is divided by technology and composition;

[0093] ③5G industry chain (network planning and design, main equipment, base station system, transmission equipment, tower construction, supporting equipment, network optimization) is divided according to process and composition;

[0094] ④The commercial aerospace industry chain (satellite manufacturing, satellite launch, ground stations, satellite application ground equipment, satellite management application software) is divided according to process, software and composition.

[0095] 2. Composition of downstream structure

[0096] 1. First level division

[0097] The downstream structure is generally divided according to downstream commodities, industries, classifications (purpose, form, channels, application scenarios, terminals), etc.

[0098] (1) By commodity

[0099] When downstream commodities can be exhausted, they are divided by commodities: for example, the pure electric vehicle industry chain (the downstream is divided into pure electric transport vehicles, pure electric passenger vehicles, pure electric washing and sweeping vehicles, pure electric garbage trucks, pure electric engineering vehicles, pure electric buses, pure electric trucks, and pure electric service vehicles); such as the copper-based materials industry chain (the downstream is divided into copper tubes, copper rods, copper wires, copper plates, copper strips, etc.); such as the semiconductor industry chain (the downstream is divided into integrated circuits, discrete devices, optoelectronic devices, sensors, and photovoltaic cells).

[0100] (2) By industry

[0101] When downstream commodities cannot be exhausted, but the application industries can be exhausted, they are divided by industry: for example, the sugar substitute industry chain (the downstream is divided into food and beverage, medicine, and cosmetics); such as the soda ash industry chain (the downstream is divided into glass industry, chemical industry, metallurgical industry, ceramic industry, food industry, pharmaceutical industry, printing and dyeing industry, and analysis and monitoring).

[0102] (3) By category

[0103] When the industry cannot be exhausted, it is divided by classification (purpose, form, channel, application scenario, terminal, etc.).

[0104] For example, the drone industry chain is divided by purpose (the downstream is divided into civil drones and military drones); the photovoltaic power generation system industry chain is divided by form (the downstream is divided into distributed photovoltaic power stations and centralized photovoltaic power stations); the Chinese herbal medicine industry chain is divided by channels (the downstream is divided into hospitals, pharmacies, and clinics); the artificial intelligence industry chain is divided by application scenarios (the downstream is unmanned driving, new retail, Industry 4.0, smart assistants, smart security, smart finance, and smart education); the industrial Internet industry chain is divided by terminal equipment (the downstream is the control of engineering machinery and the control of unattended fixed equipment); the 5G industry chain is divided by application and terminal (the downstream is divided into applications (Internet, enhanced mobile broadband, others), terminals (mobile phones, cars, smart home appliances, smart wearable devices, and electronic industrial equipment)).

[0105] Step 2: Product / Service Strategy Writing

[0106] We develop and build our own database, with the underlying data being invoice details. We use MySQL statements to retrieve, classify, count and test invoice details in the database.

[0107] MySQL statements can compile strategies based on product name, basic SPU, invoice code, transaction parties, etc. The strategy goal is to establish associations between anchored enterprises at the product level. Use strategies to find production, sales, and purchasing companies in the product link. For example, if the product link is "refrigerator", SQL statements can be used to find data containing refrigerators in the invoice. Although the data contains refrigerators, it contains impurity data, and more statements are needed to support more accurate transaction parties.

[0108] 1. Determine the basic SPU;

[0109] The basic SPU is the simplified name of the product / service. For example, if the product name is a double-door refrigerator, the basic SPU is a refrigerator.

[0110] The method to determine the base SPU is as follows:

[0111] (1) Use the results of strategy testing to view the basic SPU. For example, in the above “car window lifter”, find the car window lifter through the statement. After querying, find impurities such as window lifter motor, window lifter right rear window shaker, window lifter switch, window lifter control button, etc. Through all the data related to the car window lifter, perform conditional screening. For example, sort by transaction amount to find several names with large transaction amounts, such as front window lifter, electric window lifter assembly, summarize the commonality of the names, so the common names can be used as the basic SPU. The basic SPU of the car window lifter can be “lifter” and “lifter assembly”.

[0112] (2) Use the common name of the product / service link and its alias or synonyms as the basic SPU: For example, a chip can also be called an integrated circuit, a microcircuit, or an IC, so "chip", "integrated circuit", "microcircuit", or "IC" can all be used as the basic SPU;

[0113] (3) Through the knowledge learning of the product / service link, we can learn about well-known companies and listed companies, and use statements to query the company's transaction records to obtain the names of the corresponding products / services to determine the basic SPU.

[0114] Due to the irregularity of invoice details, the basic SPU is determined first so that more combinations can be matched to the invoice details when writing policy statements.

[0115] 2. Try to write a policy statement based on the name of the product / service link

[0116] (1) When the goods / services do not need to be limited to a certain type of enterprise, the sentence uses the basic SPU form, modifier form, and noun form:

[0117] 1) When there are multiple common terms for goods / services, the basic SPU form can be used. There are three situations:

[0118] A. If the name of the product / service link is a fixed name, such as "chip", then use the SPU policy statement directly. However, because chips are also called integrated circuits, microcircuits, and ICs, "chips", "integrated circuits", "microcircuits", and "ICs" can all be used as basic SPUs;

[0119] B. If the name of the product / service link is long, it can generally be split into several parts. For example, "car window glass lifter" can be split into multiple words such as "car", "window", "glass", and "lifter". However, "window" already covers the meaning of car and glass, and "window" is a modifier of "lifter", so the strategy statement is written as a modifier combination basic SPU;

[0120] C. For some product / service links, if the above method cannot find the corresponding product based on the name or the search data is very small, try to write it with other alternative terms, such as "virtual manufacturing verification". There is less test data. First use sentence testing to determine that the product is in the category of virtual manufacturing verification, and the basic SPU is "system" and "software". The sentence is written in the form of nouns and basic SPU;

[0121] 2) When the product / service link has a fixed name and less impurities, you can write the sentence in noun form instead of basic SPU form: for example, for consulting services, the sentence can be written directly as a noun.

[0122] 3) When the product / service link consists of "modifiers" + "basic SPU" in a general sense, the result range of the statement written in the basic SPU form is too large and most of them do not meet the requirements, but the result of the statement written in the "modifier" form will be more accurate, you can write the statement in the modifier form instead of the basic SPU form:

[0123] For example, the sonar receiver cannot find the corresponding results with the basic SPU receiver. However, the "sonar" result is more satisfactory when tested with the statement.

[0124] 3. Use of invoice coding statements

[0125] The invoice code is a string of 19 digits. The first digit is a number from 1 to 6. Each digit represents a major category, as shown in Table 1: Major category classification of invoice codes. The following 2 digits form a group, representing a minor category, as shown in Table 2: Minor category classification of invoice codes (part).

[0126] Table 1

[0127]

[0128] Table 2

[0129]

[0130] 1) You can distinguish commodity categories by adding invoice code screening conditions, for example, distinguishing the transaction parties of household appliance refrigerators and toy refrigerators.

[0131] For example, to find "toy refrigerator" by invoice code, the SQL statement adds the precise category of "toy" with invoice code, and combines it with the product name "refrigerator" to get the result;

[0132] 2) Invoice codes can also accurately identify products. After distinguishing by the invoice codes of major categories, products can be accurately divided under the same category. For example, refrigerators and refrigerator accessories can be distinguished by different product codes;

[0133] 3) Some products / services are classified into large categories and have their own invoice codes. In this case, the product nouns are not used, but the invoice codes are used to form the policy statements.

[0134] 4. Write sentences in penetrating form

[0135] For general commodities, such as engines, there are automobile engines and aircraft engines. In addition to writing strategies by adding adjectives such as automobile or aviation to combine the basic SPU, you can also judge by the buyer or seller on the invoice. For example, when looking for automobile engines, you can distinguish the purchased engine as an automobile engine by knowing that the seller or buyer is an automobile-related company.

[0136] 5. Use logical relationships in sentences

[0137] In the policy statement, the relationship between goods and invoices is matched through combined logical expressions such as "yes", "no", "or", or through logical statements such as fuzzy query and multi-condition query.

[0138] In the complex invoices, relevant invoices can be quickly found through policy statements. Impurities can be eliminated and accurately matched to standardized product units through basic SPU, invoice code, transaction penetration and logical relationships.

[0139] Step 3: Mark the invoice details with SPU label

[0140] Each invoice receives an SPU label through a process that determines the Standardized Product Unit.

[0141] The distribution of invoice codes is obtained through product names and related basic SPUs. For example, when determining the invoice code through "automobile", two major categories will appear, one is transportation equipment and the other is toys. At this time, in the automobile industry chain, choose transportation equipment.

[0142] In the internal system platform of Supersetn, MySQL statements are used to search for invoice details related to the goods / services. In the complex invoices, relevant invoices can be quickly found through strategy statements. Although the invoices found at this time are relevant, there are also impurities.

[0143] Pre-process the invoice data by basic SPU, invoice code, transaction parties, and statement logic relationship to remove impurities. Since the invoice details are non-standard statements, you can use logic such as inclusion or exclusion to filter out invoices with strong correlations, and remove impurities in the invoice details.

[0144] Determine the scope of the standardized product unit through dependency analysis, and determine the scope of the standardized product unit through the basic SPU. Due to the irregularity of invoice details, find the relevant invoices through the basic SPU, invoice code and other factors, continue to determine the relevance of the invoice by eliminating impurities, and then form the invoice standardized product unit through dependency analysis.

[0145] Each invoice obtains the SPU label through the process of determining the standardized product unit. The subsequent industrial chain construction and the standards of the company's main products can be quickly obtained through the SPU label.

[0146] Step 4: Load the main products through transaction data and other data dimensions

[0147] Summarize the invoice details of the enterprise and determine the basic SPU classification of the enterprise through the invoice details. For example, the enterprise's invoices include refrigerators, accommodation, catering, etc., and the main products of the enterprise are judged through invoice transaction data. For example, if the invoice transaction amount of refrigerators exceeds a certain threshold, it can be judged that refrigerators are the main products of the enterprise (the threshold is set according to the industry and region of the enterprise).

[0148] Extracting standardized product units based on invoice details can quickly build a multi-layer structure of the industrial chain, specifically form commodities in each link, and accurately mount enterprises in each commodity link. At present, Daokou Jinke Company has built 163 industrial chains based on the extraction of standardized product units based on invoice details, involving new generation information technology, metallurgical industry, chemical industry, machinery and equipment industry, biological industry, high-end equipment manufacturing industry, service industry, energy conservation and environmental protection industry, food processing industry, textile and clothing industry, real estate and construction industry, new materials industry, agriculture, forestry and animal husbandry and other industries; a total of more than 6,500 commodities and more than 20 million enterprises;

[0149] For example, Daokou Jinke Company extracts standardized product units based on invoice details to form a total of 7 layers of the automotive industry chain, 160 commodity links, and 718,636 enterprises. Compared with the construction of other automotive industry chains, there are 22 commodity links and more than 120 members of the industry chain. The number of commodity links and the number of enterprises are far ahead.

Claims

1. A method for extracting standardized product units based on commodity details data in an invoice, characterized in that: The specific steps are as follows: Step 1: Determine and build the industrial chain based on big data models, expert experience and machine learning; The industrial chain refers to a structure that includes upstream, midstream and downstream, with core products or services as the midstream, upstream extending upwards and downstream extending downwards in a chain-like relationship form; The upstream and downstream extension of goods / services is determined through transactions between enterprises, thus forming the upstream and downstream system of the industrial chain, and the goods / services traded between enterprises are defined as SPU; Step 2: Obtain synonyms for each product / service link in the upstream, midstream and downstream of the industrial chain through semantic recognition and knowledge learning. Each group of synonyms constitutes a group of SPUs. Step 3: Determine the most simplified name of the goods / services in each group of SPUs as the basic SPU of the group; Step 4: Write a policy statement based on the product details of the invoice; The specific process is: First, write basic policy statements based on product names or basic SPUs; It is divided into basic SPU form and modifier form; (1) When the basic SPU is not limited, the policy statement directly adopts the basic SPU form: (2) When the SPU is in the form of "modifier" + "basic SPU", the policy statement adopts the modifier form; Then, the commodity classification code is obtained through the SPU and the corresponding basic SPU, and the basic policy statement is limited according to the commodity classification code to which the SPU belongs; Finally, for the products in the industrial chain, the penetration strategy is used to lock the industrial field to which the product is applicable, and the final strategy statement is written in a penetration form: Step 5: For a single invoice, filter it through the policy statement to obtain the SPU that specifically represents the sales of the invoice, that is, extract the standardized product unit of the commodity detail data in the invoice.

2. The method according to claim 1, characterized in that In step 3, the method for determining the basic SPU is as follows: (1) Use the results of the SPU strategy test to find the extended basic SPU, and group the extended basic SPU into multiple basic SPUs corresponding to the SPU; (2) Use synonyms of goods / services as basic SPU.

3. The method according to claim 1, characterized in that In step 4, the final policy statement includes: (1) Write a policy statement based on the name of the business that buys or sells the SPU; (2) Write policy statements based on the products purchased by the SPU enterprise; (3) Write policy statements based on the products sold by the purchased SPU companies.

4. The method according to claim 1, characterized in that In step 5, for the massive invoices, the SPU represented by each invoice is obtained by screening based on the policy statement, so as to aggregate the invoices containing the SPU, and then count the enterprise information that sells / purchases the SPU.

Images