Credit text analysis method, credit object auditing method and credit object auditing device

An analysis method and technology of letter of credit, applied in the field of analysis of letter of credit text, can solve the problems of high labor cost and low efficiency

Pending Publication Date: 2022-04-22
CHINA CONSTRUCTION BANK
0 Cites 0 Cited by

AI-Extracted Technical Summary

Problems solved by technology

In the traditional letter of credit review, it is necessary to manually understand the relevant content in the letter of credit or the letter of credit document, and obtain the correspond...
View more

Method used

[0183] The letter of credit language has the problems of high element density, many logical scenarios, and large differences in slot types, making it difficult to use the method based on deep learning models (the demand for samples is huge, and slot definition is difficult). Therefore, the embodiment of the present invention proposes a method for parsing the letter of credit text based on the combination of semantic templates, in which the relationship of multiple elements and logical sentence patterns are all reflected in the short template, which greatly reduces the complexity of parsing. The embodiment of the present invention also proposes a set of semantic algorithms suitable for letter of credit language, which can cover various attributes, attribute booleanization, attribute logical combination and other scenarios, further avoiding the situation that the traditional slot definition cannot handle attribute combination scenarios. The cleverly designed combination of semantic operators and slots in the solution of the embodiment of the present invention can cover richer semantic representations, ...
View more

Abstract

The invention discloses a credit text analysis method and a credit object auditing method and device, and relates to the technical field of knowledge maps. A specific embodiment of the method comprises the steps of obtaining a target credit card text, wherein the target credit card text comprises at least one target element; a target semantic template corresponding to the target credit document text is obtained, the target semantic template comprises a statement template and an upper sharing hierarchy of the statement template, the statement template comprises at least one target slot position, and the target slot position corresponds to the target element; obtaining a full-link slot position of the target credit card text according to the statement template and the upper sharing hierarchy; and matching the target semantic template with the target credit text to obtain a target slot value corresponding to the full-link slot position. According to the implementation mode, the slot position and the slot value corresponding to the element in the credit card can be automatically obtained, and the related content in the credit card text can be analyzed.

Application Domain

FinanceSemantic analysis +3

Technology Topic

EngineeringData mining +4

Image

  • Credit text analysis method, credit object auditing method and credit object auditing device
  • Credit text analysis method, credit object auditing method and credit object auditing device
  • Credit text analysis method, credit object auditing method and credit object auditing device

Examples

  • Experimental program(1)

Example Embodiment

[0067] The following illustrative embodiments of the present invention in conjunction with the accompanying drawings, including various details of embodiments of the present invention to facilitate understanding, should be considered merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications may be made to the embodiments described herein, without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and brevity, the description of the well-known functions and structures is omitted from the following description.
[0068] The acquisition, storage, use, processing, etc. of data in this application technical plan comply with the relevant provisions of national laws and regulations.
[0069] To make the embodiment of the present invention embodiments easy to understand, the following will be explained in the terms involved in embodiments of the present invention:
[0070] Knowledge representation: It belongs to the subcategory of the knowledge graph, describes the organizational form of knowledge, and is the cornerstone of the application of the knowledge graph.
[0071] Intention: The core of the sentence wants to express a large type of content, such as "show me a piece of Jay Chou's blue and white porcelain" The intention is "listen to music".
[0072] Slot: An entity category that is defined as a category of related content items, such as slot "Artist", "Song Name".
[0073] Slot value: that is, the entity, is the specific value of the slot in the current statement, such as the slot value "Jay Chou (singer)", "blue and white porcelain (song name)".
[0074] Corresponding to the application of natural language understanding, the terms of the understanding letter of credit can be expressed in the following 4 levels:
[0075] (1) Identify the intention expressed in the clause, such as "prohibiting a certain act", "must display elements", etc.;
[0076] (2) Identifying the slot under the intent in the clause, generally the name of the element, such as "letter of credit number", "signatory", etc.;
[0077] (3) Identify the specific slot value under the joint requirements of the intent and slot in the identification clause, generally a string or numeric value, etc.;
[0078] (4) Assemble the above intentions, slots, and slot values according to the knowledge representation structure defined in the knowledge graph to express the content required by the complete terms.
[0079] Embodiments of the present invention is based on the knowledge representation structure defined in the knowledge graph for the semantic representation of the terms, similar to the general knowledge representation definition, having a multi-level inheritance relationship, the sharing of the basic element structure and other characteristics.
[0080] Features correspond to slots, which are the basic units for semantic identification or resolution. The elements can be the name of the shipper, the name of the country to which the shipper belongs, the name of the beneficiary, the telephone number of the beneficiary, etc. You can < subject, property, object, > examples of basic elements as follows:
[0081] elements used to characterize the name of the shipper;
[0082] elements used to characterize the shipper's address;
[0083] elements used to characterize the name of the country stated by the shipper.
[0084] Start with "." The superior-subordinate relationship of the connecting features, "(*)" a description of the contents of the layer, and the complete feature (slot) is:
[0085] Table 1 Examples of data attribute elements of a letter of credit
[0086] Chinese description Features (slots) Slot category Ask for the name of the shipper on the bill of lading billLadingRequiredClause.shiper.name DP Require the name of the shipper's country on the bill of lading billLadingRequiredClause.shipper.country.name DP Request shipper phone information on the bill of lading billLadingRequiredClause.shipper.contactInfo.phoneNumber DP … Ask for the beneficiary name on the invoice invoiceRequiredClause.beneficiary.name DP Require the name of the beneficiary's country on the invoice invoiceRequiredClause.beneficiary.country.name DP Ask for beneficiary phone information on the invoice invoiceRequiredClause.beneficiary.contactInfo.phoneNumber DP
[0087] Table 1 DP indicates that the slot is a data property. In addition, the terms of the letter of credit will only require the shipper or beneficiary to be displayed in the document, without specifying its specific name, and the corresponding element (slot) will be OP (objectproperty, object attribute).
[0088] Table 2 Examples of attribute elements of some objects in a letter of credit
[0089] Chinese description Features (slots) Slot category Require a shipper on the bill of lading billLadingRequiredClause.shiper OP The state of the shipper on the bill of lading is required billLadingRequiredClause.shipper.country OP Ask for the beneficiary name on the invoice invoiceRequiredClause.beneficiary OP Require beneficiary information on the invoice invoiceRequiredClause.beneficiary.contactInfo OP
[0090] In order to facilitate the identification of the textual information of the letter of credit, the long-sentence clauses in the text need to be split into independent elements. In the closed language scenario of the letter of credit, the expression of business elements is dense and the element units are mostly independent. Accordingly, embodiments of the present invention proposes to split the long sentence clause into independent elements, parse them separately according to their respective semantics, and then merge into a complete semantic structure based on the combination rules. Long sentence clauses can be split into separate units of features.
[0091] Example sentence 1. SIGNED COMMERCIAL INVOICE IN THREE COPIES INDICATEMANUFACTURER'S/EXPORTOR'S NAME AND ADDRESS,LC NO. ABC123456 AND DT 2020 MAY20.
[0092] Example Sentence 1 can be split into the following independent elements, separated by semicolons: SIGNED; COMMERCIALINVOICE; IN THREE COPIES; INDICATE MANUFACTURER'S/EXPORTOR'S NAME AND ADDRESS; LC NO. ABC123456 AND DT 2020 MAY 20。
[0093] Example sentence 2. SIGNED COMMERCIAL INVOICE OR PACKING LIST SHOW LC NUMBER ANDDATE,GOODS AMOUNT,PRICE TERM.
[0094] Example sentence 2 can be split into the following independent elements, separated by semicolons: SIGNED; COMMERCIALINVOICE OR PACKING LIST SHOW; LC NUMBER AND DATE; GOODS AMOUNT; PRICE TERM.
[0095] Example Sentence 3. IN CASE OF BY COURIER SHIPMENT,BENE’S CERT NOT REQUIRED.
[0096] Example Sentence 3 can be split into the following independent elements, separated by semicolons: IN CASE OF BYCOURIER SHIPMENT; BENE'S CERT NOT REQUIRED..
[0097] As can be seen from the above example, the different element units fully express the required semantic information and are independent of each other. Features may contain logical relationships such as and (AND), OR, or (OR), and IF (IN CASE OF), and their role varies depending on the feature value (if the document name contains logic, the logical relationship covers the entire sentence; other features contain logic, so the logic is limited to that feature).
[0098] Based on the above analysis, embodiments of the present invention provides a method of parsing the text of a letter of credit. Figure 1 is a schematic diagram of a method of parsing the text of a letter of credit provided by a first embodiment of the present invention, e.g., a schematic diagram of the process Figure 1 As shown, the method comprises:
[0099] Step 101: Obtain the target letter of credit text, the target letter of credit text comprises at least one target element.
[0100] The text of the target letter of credit is the text related to the letter of credit. The target letter of credit text can be the text extracted from the letter of credit or letter of credit document.
[0101] Features correspond to slots, which are the basic units for semantic identification or resolution. The elements can be the name of the shipper, the name of the country to which the shipper belongs, the name of the beneficiary, the telephone number of the beneficiary, etc.
[0102] The text of the target letter of credit can be short. Embodiments of the present invention relates to the target letter of credit text is a long text, which may be split into a plurality of independent short text for parsing. Short text includes one element or multiple related elements, and short text can be parsed through a single semantic template, so as to achieve the purpose of parsing the entire long text.
[0103] Step 102: To obtain the target semantic template corresponding to the target letter of credit text, the target semantic template comprises: a statement template and the upper level of the statement template sharing hierarchy.
[0104] The statement template includes at least one target slot that corresponds to the target feature. Using the statement template, matching with the text of the target letter of credit, you can determine the slot value corresponding to the slot. A specific example of a semantic template is given below:
[0105] {address: }>>consignee
[0106] In the above equation, the left side of the ">>" is the statement template part, and the right side is the upper shared level part of the slot. The semantic template section can include one or more target slots. The upper sharing hierarchy section may include one or more upper sharing levels. The upper sharing hierarchy is generally None or OP attributes. May be "intent" may be used to simplify the description of the upper level sharing hierarchy, the description of the embodiment of the present invention will also be through the intention to replace the upper sharing level.
[0107] Step 103: According to the statement template and the upper level sharing hierarchy, the full link slot of the target letter of credit text is obtained.
[0108] The full-link slot is the slot in the text of the target letter of credit that needs to be parsed out, and the full-link slot is the slot corresponding to the target element in the text of the target letter of credit. The semantic template can include semantic operators, through which the combination of the statement template and the upper level sharing level is determined to generate the full link slot of the target letter of credit text.
[0109] Step 104: The target semantic template is matched to the target letter of credit text to obtain the target slot value corresponding to the full link slot.
[0110] In an embodiment of the present invention, according to the target letter of credit text corresponding to the target semantic template in the statement template and the upper level sharing level, to obtain the target letter of credit text of the full link slot. Full-link slots are slots that need to be extracted from the target letter of credit text. Then match the target semantic template with the target letter of credit text to obtain the target slot value corresponding to the full link slot. As a result, the relevant content in the letter of credit text can be automatically identified, reducing the labor cost required in the process of letter of credit review and high efficiency.
[0111] Figure 2 is a second embodiment of the present invention provides a schematic diagram of a method of parsing the text of a letter of credit, e.g., a schematic diagram of the flow Figure 2 As shown, the method comprises:
[0112] Step 201: Obtain the target letter of credit text, the target letter of credit text comprises at least one target element.
[0113] Step 202: to obtain the target semantic template corresponding to the target letter of credit text, the target semantic template comprising: a statement template and the upper sharing level of the statement template, the upper sharing level comprising: the first operator.
[0114] The first operator is used to determine how the upper sharing hierarchy is first combined with the target slot. The first operator includes at least one of the following: no intent, single intent, or intent and output mode.
[0115] Step 203: According to the first operator, determine the first combination of the upper sharing level and the target slot.
[0116] The first combination comprises at least one of the following: no intent mode, a single intent mode, or an intent mode and an output mode.
[0117] Step 204: In the first combined manner, the upper bit sharing level is combined with the target slot to obtain the full link slot of the target letter of credit text.
[0118] Step 205: The target semantic template is matched to the target letter of credit text to obtain the target slot value corresponding to the full link slot.
[0119] In an embodiment of the present invention, according to the first operator in the upper sharing hierarchy, determine the first combination of the upper sharing level and the target slot, and then according to the first combination, the combination of the upper level sharing level and the target slot, can flexibly and conveniently obtain the target letter of credit text of the full link slot.
[0120] In one embodiment of the present invention, the upper sharing level comprises: a first sharing level, a second sharing level and a second operator; said in the first combination, combined with the target slot, to obtain the target letter of credit text of the full link slot, comprising: in the first combination, the combination of the first sharing level and the target slot, to obtain the first link slot; in the first combination, the second sharing level and the target slot, To obtain a second link slot; according to the second operator, determine the first link slot and the second link slot second combination; in the second combined manner, the first link slot and the second link slot are combined to obtain the full link slot of the target letter of credit text.
[0121] The second operator is located in the upper sharing hierarchy portion of the target semantic template, and the second operator can be used to connect multiple upper sharing levels. The second operator includes at least one of the following: and, or, and the output mode. The second combination comprises at least one of the following: and mode, or mode and output mode. According to the first operator and the second operator, the full link slot of the target letter of credit text can be obtained flexibly and conveniently.
[0122] In one embodiment of the present invention, the statement template comprises: a first target slot, a second target slot and a third operator; the full link slot of the target letter of credit text is obtained according to the statement template and the upper level sharing level, comprising: combining the upper layer sharing level with the first target slot, to obtain a third link slot; combining the upper layer sharing level with the second target slot, to obtain a fourth link slot; according to the third operator, Determine the third link slot and the fourth link slot in combination with the fourth link slot; in the third combination, combine the third link slot with the fourth link slot to obtain the full link slot of the target letter of credit text.
[0123] The third operator is located in the statement template section of the target semantic template. The third operator can be used to connect multiple target slots. The third operator comprises at least one of the following: and, or, and substitution. The third combination comprises at least one of the following: and the manner, or manner and substitution. According to the third operator, the full link slot of the target letter of credit text can be obtained flexibly and conveniently.
[0124] To facilitate the understanding of embodiments of the present invention. The following is a concrete example to explain the concepts of semantic templates, semantic operators, etc. Among them, semantic operators include: the first operator, the second operator and the third operator. Based on the inheritance and sharing characteristics of the knowledge representation structure, and the semantic independence of the element unit in the long sentence clause, the embodiment of the present invention proposes a semantic template-based element analysis rules, and defines a set of semantic operators and representation requirements thereof.
[0125] Semantic Template 1: IN? {totalCopy: }[COPIES| COPY]>>None
[0126] Semantic Template 2: [CONSIGNEED| CONSIGNEE]AND[NOTIFIED| NOTIFY]{name: }[,|#]{address: }>>AND_consignee_notifyParty
[0127] In the above semantic template 1 and semantic template 2, the left side of the ">>" is the statement template, and the right side is the upper shared level of the slot, which is simplified with "intent", generally noone or OP attribute. "*?” Indicates that the word may exist or not; "{}" defines the finest-grained slot, ":" preceded by the slot name, followed by a specific slot value, "< *>" is a wildcard character for a class of slot values, generally a preprocessing result or a scene where the dictionary can replace the expression, such as Then correspond to a company name library; "[*|*]" describes synonyms or other scenes of the same kind that can be replaced without affecting semantics, and is divided by a vertical line, where "#" indicates that the word to which the position belongs can be nullable, and "? "The role is consistent. Therefore, for the following long sentences to be parsed, the full semantic meaning can be expressed through 5 semantic templates.
[0128] Long sentence to be parsed: SIGNED COMMERCIAL INVOICE IN THREE COPIES INDICATEMANUFACTURER'S/EXPORTOR'S NAME AND ADDRESS,LC NO. ABC123456 AND DT 2020 MAY20.
[0129] Semantic Template 1: ALL? DULY? {show_signer:SIGNED}>>None
[0130] Semantic Template 2: {docTitle: }>>None
[0131] Semantic Template 3: IN? {totalCopy: }[COPIES| COPY]>>None
[0132] Semantic Template 4: ACTUAL? [MANUFACTURER'S| EXPORTER'S][OR|/][MANUFACTURER'S| EXPORTER'S]{show_name:Name}[AND|,]{show_address:ADDRESS}>>OR_exporter_manufacturer
[0133] Semantic Template 5: [DOCUMENTARY CREDIT| LETTER OF CREDIT| LC|DC][NUMBER|NO.| NO]{creditNumber: }[AND|,][DATED| DT| DD]{lcIssueDate: }>>None
[0134] Suppose the above long sentence to be parsed belongs to the category of organismRequiredClause, and replace the right "intent" with the left slot "." All the semantics of the whole sentence obtained after connection are as follows, the left side of the colon is the slot, and the right side of the colon is the slot value obtained by using the semantic template to match the long sentence to be parsed:
[0135] invoiceRequiredClause.signer:”
[0136] invoiceRequiredClause.docTitle:COMMERCIAL INVOICE
[0137] invoiceRequiredClause.totalCopy:3
[0138] (invoiceRequiredClause.exporter.name:”&invoiceRequiredClause.exporter.address:”)| (invoiceRequiredClause.manufacturer.name:”&invoiceRequiredClause.manufacturer.address:”)
[0139] invoiceRequiredClause.creditNumber:123456
[0140] invoiceRequiredClause.lcIssueDate:2020 MAY 20
[0141] Based on the elements required by the language of the letter of credit and their relationships, the following semantic operators (first operator, second operator and third operator) are defined:
[0142] AND/OR in the first place of intent or intent;
[0143] ">>OR" expresses the left slot OR relationship;
[0144] ">>AND/OR_A" expresses the intention that A is AND/OR after stitching with the left slot;
[0145] ">>AND/OR_A_B" expresses the relationship between the intention A and B as AND/OR, and then splices with the left slot;
[0146] Or slot first, expressing multiple or defined slots have or relationships, such as "ISSUED BY{or_name: }OR{or_name: }>>issuer”;
[0147] None of the intent, the content of the terms is expressed directly by the left slot;
[0148] Show is at the top of the intent/slot, converting the original non-boolean attribute to a boolean intent
[0149] ">>show_A" expresses the document display slot A, output {A:"};
[0150] ">>show_A_B" expresses that multiple slots A and B values are boolean, and are AND relationships;
[0151] ">>showOR_A_B" expresses that multiple slots A and B values are boolean and are OR relationships;
[0152] replace_a@b, the value of the expression slot refers to the scene and needs to be replaced by the specific value of "b" after "@", such as {replace_issuer.name@seller.name} the output is {issuer.name:ABC LTD} where "ABC LTD" is the seller.name.
[0153] a_b express multiple slots to share a slot value, such as slot {issuer.name_seller.name:ABC LTD} output is {issuer.name:ABC LTD}, {seller.name:ABC LTD}.
[0154] Embodiments of the present invention proposes a complex clause analysis and structural reduction scheme based on a combination of templates. Figure 3 is a third embodiment of the present invention provides a schematic diagram of a method of parsing the text of a letter of credit, e.g., a schematic diagram of the flow Figure 3 As shown, the input target semantic template, the target semantic template comprising: a statement template and the upper level of the statement template, the statement template includes at least one target slot, the upper level sharing level can be used to represent the intent. Subsequent stitching operations are determined based on the form of intent and the accounting included in the semantic template. Traverse the slot, according to the slot form and operator, and the intention to do splicing, to get the full link slot. Depending on the position of the OR operator, determine whether to extend the OR intent to the right clause. The target semantic template is used to match the target letter of credit text and output the semantic information of the target letter of credit text. According to the semantic template and semantic operator defined above, the business meaning of the terms of the letter of credit can be restored according to the intention, slot, and slot value, and directly mapped to the knowledge representation of the knowledge graph. In the above different scenarios, the combination of intent and slot is different.
[0155] Table 3 Structural reduction logic based on intent slots and semantic operators
[0156]
[0157] Table 3 shows the structural reduction logic based on the intent slot and the semantic operator, and the semantic information in the letter of credit text can be parsed by using the logical relationship shown in Table 3. where A or B can be used to represent intent or slot. The intent in Table 3 can also be understood as the upper level of sharing.
[0158] In Table 3, the None intent corresponds to the intentlessness of the first operator. A single intent corresponds to a single intent of the first operator. OR_ the "OR" in intent corresponds to the third operator or way. The "OR" in the OR_A corresponds to the first operator or intent. The "OR" in the OR_A_B corresponds to the or way of the second operator. The "AND" in the AND_A_B corresponds to the sum of the second operator. The "show" in the Show_A corresponds to the output mode of the first operator. The "show" in the Show_A can also correspond to the output mode of the third operator. For showOR_A_B, if A and B are the intent, showOR corresponds to the output mode of the first operator, or the intention; if A and B are slots, the showOR corresponds to the output mode or mode of the third operator. The "replace" in the Replace_A@B corresponds to the substitution of the third operator.
[0159] In order to facilitate understanding, take the Letter of Credit Requirement Clause InvestmentRequiredClause as an example of the semantic restoration results of each scenario (you can also refer to the semantic analysis and structure reduction results of the above complete long sentences):
[0160] Text A: SIGNED OR STAMPED BY BENEFICIARY
[0161] >OR_signer_stamper
[0162] The "OR" of ">>OR_signer_stamper" in template A corresponds to the second operator, or mode. Extract the full link slots and slot values in text A from template A:
[0163] InvoiceRequiredClause.signer.partyIndicator:BENEFICIARY| InvoiceRequiredClause.stamper.partyIndicator:BENEFICIARY
[0164] Text B: SIGNED BY BENEFICIARY OR ABC LTD
[0165] Template B: ISSUED BY THE? {partyIndicator:BENEFICIARY} [OR|/] {name: }>>OR_issuer
[0166] [OR|/] in template B corresponds to the or mode of the third operator. Extract the full link slots and slot values in text B by template B:
[0167] InvoiceRequiredClause.issuer.partyIndicator:BENEFICIARY| InvoiceRequiredClause.issuer.name:ABC LTD
[0168] Text C: LC NUMBER AND DATE
[0169] >show_creditNumber_lcIssueDate
[0170] The "show" of the "show_creditNumber_lcIssueDate" in template C corresponds to the output mode of the second operator. Extract the full link slots and slot values in text C through template C:
[0171] creditNumber:”&lcIssueDate:”
[0172] Text D: ISSUED BY THE SELLER
[0173] Template D:
[0174] ISSUED BY THE? {replace_name@seller.name:SELLER}>>issuer
[0175] The "issuer" in template D corresponds to the single intent of the first operator. Extract the full link slots and slot values in text D by template D:
[0176] issuer.name: ABC LTD, where "ABC LTD" is seller.name
[0177] Text E:INVOICE OR PACKING LIST MUST INDICATE GOODS ORIGIN.
[0178] Template E1:{docTitle: }OR{originCountry: }>>OR
[0179] >show_originCountry
[0180] The first "OR" in template E1 corresponds to the or mode of the third operator. The "show" in template E2 corresponds to the output mode of the first operator. Extract the full link slots and slot values in text E by template E1 and template E2:
[0181] (InvoiceRequiredClause.docTitle:INVOICE&InvoiceRequiredClause.originCountry:”)| (PackinglistRequiredClause.docTitle:PACKING LIST&PackinglistRequiredClause.originCountry:”)
[0182] Wherein, text A to text D is short text, and short text is text that can be parsed through a single semantic template. Text E is long text, and long text needs to be split into multiple independent short text for parsing.
[0183] The letter of credit language has the problems of large feature density, many logical scenes, and large differences in slot types, and it is difficult to use the deep learning model-based method (the sample demand is extremely large, and the slot definition is difficult). Thus, embodiments of the present invention proposes a method of parsing the text of the letter of credit based on the combination of semantic templates, wherein the relationship between multiple elements and logical sentences, are reflected in the short template, greatly reducing the complexity of the analysis. Embodiments of the present invention also proposes a set of semantic algorithms suitable for the language of letters of credit, which can cover various attributes, attribute boolean, attribute logic combinations and other scenarios, further avoiding the traditional slot definition can not handle the attribute combination scenario. Embodiments of the present invention embodiments of the cleverly designed semantic operator and the combination of slots, can cover a richer semantic representation, to avoid the traditional analysis of slots can only be divided into scenarios redundant definition of the problem.
[0184] Further, embodiments of the present invention proposes a "short semantic template parsing" method of long sentences, complex letter of credit requirements clause analysis problems split into sub-problems of parsing each short sentence, reducing the difficulty of parsing. In addition, a cleverly designed set of semantic operators can cover complex property value recognition, non-Boolean property Boolean ("show_A" form or "show_A_B form, etc.), shared property upper bit words, and identification of complex properties with logic and or (AND/OR), etc., avoiding the redundancy and complexity of defining slots for each type of attribute in traditional slot recognition. Finally, through the corresponding operation process of the semantic operator, the short sentence parsing results are combined to obtain the semantic representation of the complex clauses of the complete long sentence. Accordingly, the embodiments of the present invention have the following technical effects:
[0185] (1) Long sentence parsing is decoupled into short sentence parsing according to its independent semantic meaning, and a parsing expression is designed to cover synonym substitution, alternative words, slot extraction, and intent (slot sharing upper words).
[0186] (2) A set of semantic operators has been designed to cover the slot recognition problems of various attributes, attribute booleanization, attribute logic combinations and other special scenarios. In addition, semantic operators are bound to their operational processes, and the resulting combination of short sentences can be restored to the full semantics of long sentences according to combinatorial logic.
[0187] (3) Through the combination of slots and operators, the automatic recognition of various complex semantics of the letter of credit text is realized.
[0188] Figure 4 Is a fourth embodiment of the present invention provides a schematic diagram of the process of reviewing the object of a letter of credit. as Figure 4 As shown, the method comprises:
[0189] Step 401: to obtain the first letter of credit text of the first letter of credit object, using the parsing method of the above letter of credit text, to obtain the first text slot of the first letter of credit text and the corresponding slot value of the first text slot.
[0190] The first letter of credit is related to the letter of credit. The first letter of credit can be a letter of credit, a letter of credit document, a letter of credit clause, etc. The semantics in the text of the first letter of credit can be parsed by the slot value corresponding to the first text slot and the first text slot.
[0191] Step 402: to obtain the second letter of credit text of the second letter of credit object, using the parsing method of the above letter of credit text, to obtain the second text slot of the second letter of credit text and the corresponding slot value of the second text slot.
[0192] The second letter of credit is related to the letter of credit. The object of the second letter of credit may be a letter of credit, a letter of credit document, a letter of credit clause, etc. The semantics in the text of the second letter of credit can be parsed by the slot value corresponding to the second text slot and the second text slot.
[0193] Step 403: According to the slot value corresponding to the first text slot and the slot value corresponding to the second text slot, determine whether the first letter of credit object and the second letter of credit object is consistent.
[0194] Specifically, the first letter of credit object may be determined by the following means: for each of the first text slots: to determine the current first text slot corresponding to the current second text slot; respectively, to determine the first slot value corresponding to the current first text slot and the current second text slot corresponding to the second slot value; to determine whether the first slot value and the second slot value matches; if not matched, then determine the current first text slot as a fault slot Determine whether the first text slot contains a fault slot; if so, determine that the object of the first letter of credit is inconsistent with the object of the second letter of credit; if not, determine that the object of the first letter of credit is consistent with the object of the second letter of credit.
[0195] By the above method, the first slot value corresponding to each first slot in the first letter of credit object is compared with the second slot value corresponding to the second slot, and if there is a situation where the first slot value does not match the second slot value, it means that the first letter of credit object and the second letter of credit object are inconsistent. If all first slot values have corresponding second slot values, the first letter of credit object is consistent with the second letter of credit object.
[0196] The embodiment of the present invention provides a method of parsing the text of the letter of credit, applied to the letter of credit review, the system can automatically obtain the corresponding slot and slot value of the elements in the letter of credit object, thereby identifying the semantic content contained in the letter of credit object, reducing the labor cost required in the process of letter of credit review and high efficiency.
[0197] Figure 5 Is an embodiment of the present invention provides a schematic structural diagram of a letter of credit text parsing apparatus. as Figure 5 As shown, the apparatus comprises:
[0198] Text acquisition module 501 for obtaining the target letter of credit text, the target letter of credit text comprises at least one target element;
[0199] Template acquisition module 502, for obtaining the target semantic template corresponding to the target letter of credit text, the target semantic template comprising: a statement template and the upper level of the statement template sharing hierarchy, the statement template comprises at least one target slot, the target slot corresponds to the target element;
[0200] Slot determination module 503, for sharing the hierarchy according to the statement template and the upper level, to obtain the full link slot of the target letter of credit text;
[0201] Value determination module 504, for matching the target semantic template to the target letter of credit text to obtain the target slot value corresponding to the full link slot.
[0202] Alternatively, the upper sharing hierarchy comprises: a first operator;
[0203] The slot determination module 503 is specifically for:
[0204] According to the first operator, determine the first combination of the upper sharing level and the target slot;
[0205] In the first combinatorial manner, the upper level sharing level is combined with the target slot to obtain the full link slot of the target letter of credit text.
[0206] Alternatively, the second operator comprises at least one of the following: and, or, and the output mode.
[0207] Alternatively, the first operator comprises at least one of the following: no intent, single intent, or intent and output mode.
[0208] Alternatively, the statement template comprises: a first target slot, a second target slot and a third operator;
[0209] The slot determination module 503 is specifically for:
[0210] Combine the upper layer sharing hierarchy with the first target slot to obtain a third link slot;
[0211] Combine the upper shared level with the second target slot to obtain a fourth link slot;
[0212] According to the third operator, determine the third link slot and the third combination of the fourth link slot;
[0213] In the third combination, the third link slot is combined with the fourth link slot to obtain the full link slot of the target letter of credit text.
[0214] Alternatively, the third operator comprises at least one of the following: and, or, and substitution.
[0215] Figure 6 is an embodiment of the present invention provides a schematic structural diagram of a letter of credit object audit apparatus. as Figure 6 As shown, the apparatus comprises:
[0216] The first acquisition module 601, for obtaining the first letter of credit text of the first letter of credit object, using the parsing method of the above letter of credit text, to obtain the first text slot of the first letter of credit text and the corresponding slot value of the first text slot;
[0217] The second acquisition module 602, for obtaining the second letter of credit text of the second letter of credit object, using the parsing method of the above letter of credit text, to obtain the second text slot of the second letter of credit text and the corresponding slot value of the second text slot;
[0218] Audit module 603, for determining whether the first letter of credit object is consistent with the second letter of credit object according to the slot value corresponding to the first text slot and the corresponding slot value of the second text slot.
[0219] Alternatively, the first letter of credit text corresponds to at least one first text slot;
[0220] Audit module 603 is specifically used for:
[0221] For each of the first text slots: determine the current second text slot corresponding to the current first text slot; respectively, determine the first slot value corresponding to the current first text slot and the second slot value corresponding to the current second text slot; determine whether the first slot value matches the second slot value; if not matched, then determine the current first text slot as a fault slot;
[0222] Determine whether the first text slot contains a fault slot;
[0223] If so, it is determined that the object of the first letter of credit is inconsistent with the object of the second letter of credit;
[0224] If not, it is determined that the object of the first letter of credit is consistent with the object of the second letter of credit.
[0225] Embodiments of the present invention provides an electronic device, comprising:
[0226] One or more processors;
[0227] a storage device for storing one or more programs,
[0228] When one or more programs are executed by one or more processors, such that one or more processors implement the method of any of the above embodiments.
[0229] Embodiments of the present invention provides a computer program product, including a computer program, the computer program is executed by the processor to implement a method of parsing the text of the letter of credit in an embodiment of the present invention or a method of reviewing the object of the letter of credit.
[0230] Refer to below Figure 7 , which shows a schematic structural diagram of a computer system 700 suitable for implementing an embodiment of the present invention. Figure 7 The terminal apparatus shown is merely an example and should not bring any limitations to the function and scope of use of the embodiments of the present invention.
[0231] as Figure 7 As shown, the computer system 700 includes a central processing unit (CPU) 701, which may perform various appropriate actions and processing according to the program stored in the read-only memory (ROM) 702 or loaded from the storage portion 708 into the random access memory (RAM) 703. In RAM 703, there are also various programs and data required for the operation of the system 700. CPU 701, ROM 702 and RAM 703 are connected to each other via bus 704. Input / output (I / O) interface 705 is also connected to bus 704.
[0232] The following components are connected to the I / O interface 705: comprising a keyboard, mouse, etc. input portion 706; comprising an output portion such as a cathode ray tube (CRT), a liquid crystal display (LCD) and the like and a speaker, etc.; including a storage portion 708 of a hard disk, etc.; and a communication portion 709 comprising a network interface card such as a LAN card, a modem, etc. Communication portion 709 performs communication processing via a network such as the Internet. Driver 710 is also connected to the I / O interface 705 as needed. Removable media 711, such as disks, optical disks, magnetic optical disks, semiconductor memory, etc., as needed installed on the drive 710, in order to facilitate the readout of the computer program is installed into the storage portion 708 as needed.
[0233] In particular, according to an embodiment of the disclosure of the present invention, the process described above with reference to the flowchart may be implemented as a computer software program. For example, embodiments disclosed in the present invention include a computer program product comprising a computer program hosted on a computer-readable medium, the computer program comprising a program code for executing the method shown in the flowchart. In such embodiments, the computer program may be downloaded and installed from the network by communication portion 709, and / or installed from removable media 711. When the computer program is executed by the central processing unit (CPU) 701, the above functions are performed in the system of the present invention is defined.
[0234] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above. Computer-readable storage mediums may, for example, be - but are not limited to - electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, the computer-readable storage medium may be any tangible medium comprising or storing a program, the program may be used by or in combination with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium may include a data signal propagated in the baseband or as part of a carrier, which carries computer-readable program code. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, the computer-readable medium may be transmitted, propagated or transmitted for use by or in combination with the instruction execution system, apparatus or device program. The program code contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the above.
[0235] Flowcharts and block diagrams in the drawings illustrate the architecture, functions and operations of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram may represent a module, program segment, or part of code, the module, program segment, or part of the code comprises one or more executable instructions for implementing a prescribed logic function. It should also be noted that in some implementations as substitutions, the functions noted in the box may also occur in a different order than those noted in the drawings. For example, two boxes represented consecutively can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the function involved. It should also be noted that each box in a block diagram or flowchart, as well as a combination of boxes in a block diagram or flowchart, may be implemented in a dedicated hardware-based system that performs a prescribed function or operation, or may be implemented in combination with a combination of specialized hardware and computer instructions.
[0236] The modules described in embodiments of the present invention may be implemented by means of software, or may be implemented by means of hardware. The described module may also be disposed in the processor, for example, may be described as: text acquisition module, template acquisition module, slot determination module and value determination module. Wherein, the name of these modules does not in some case constitute a limitation of the module itself, for example, the text acquisition module may also be described as "to obtain the target letter of credit text, the target letter of credit text comprises at least one target element of the module".
[0237] As another aspect, the present invention further provides a computer-readable medium, the computer-readable medium may be contained in the apparatus described in the above embodiments; may also be present alone, and not assembled into the apparatus. The computer-readable medium carries one or more programs, when one or more of the above programs are executed by one of the devices, such that the device comprises:
[0238] Obtain the target letter of credit text, the target letter of credit text includes at least one target element;
[0239] To obtain the target semantic template corresponding to the target letter of credit text, the target semantic template comprising: a statement template and the upper level of the statement template sharing hierarchy, the statement template comprises at least one target slot, the target slot corresponds to the target element;
[0240] According to the statement template and the upper level sharing hierarchy, the full link slot of the target letter of credit text is obtained;
[0241] The target semantic template is matched to the target letter of credit text to obtain the target slot value corresponding to the full link slot.
[0242] According to the technical solution of the embodiment of the present invention, according to the target letter of credit text corresponding to the target semantic template in the statement template and the upper level sharing level, to obtain the target letter of credit text of the full link slot. Full-link slots are slots that need to be extracted from the target letter of credit text. Then match the target semantic template with the target letter of credit text to obtain the target slot value corresponding to the full link slot. As a result, the relevant content in the letter of credit text can be automatically identified, reducing the labor cost required in the process of letter of credit review and high efficiency.
[0243] The above specific embodiments do not constitute a limitation on the scope of protection of the present invention. Those skilled in the art should understand that depending on design requirements and other factors, various modifications, combinations, subcombinations and substitutions may occur. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the invention shall be included in the scope of protection of the present invention.

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products