Order information acquisition method and order generation method

Abstract

The embodiment of the invention provides an order information acquisition method, an order generation method, an order generation system, electronic equipment, a storage medium and a computer program product. The order information acquisition method comprises the steps that a communication record for a target order is acquired, and the communication record comprises structured information and / or unstructured information; extracting a communication text fragment of a preset order information category from the unstructured information through an order information extraction model obtained through pre-training; and obtaining order information of the target order based on a preset information extraction rule, the structured information and the communication text fragment. According to the method, the order related effective information in the communication record is intelligently extracted and structured processing is performed to obtain the structured order information, so that the order information acquisition efficiency is improved, and the order information acquisition accuracy and comprehensiveness are improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an order information acquisition method, a training method for an order information extraction model, an order generation method, an order generation system, an electronic device, a storage medium and a computer program product. Background Art

[0002] In e-commerce scenarios, sellers and buyers usually reach an order after multiple rounds of online communication and negotiation. According to the statistical results of online data, 82% of daily paying buyers have communicated in the past 7 days. The content of the negotiation during the communication process includes but is not limited to: products, logistics, address, price and other information. It is not uncommon for an order to be negotiated after several months of online communication. After a long period of online communication and reaching the final order intention, when you want to draft an order for the buyer or seller user, you often need to manually search for key information such as products, quotations, addresses, etc. from the historical communication records as order information when drafting the order. This operation not only takes a long time, but also easily misses key information, and the order information acquisition efficiency is low and the accuracy is low. Summary of the invention

[0003] The embodiment of the present application provides a method for obtaining order information, which can improve the efficiency and accuracy of obtaining order information, thereby improving the efficiency of order generation.

[0004] Correspondingly, the embodiments of the present application also provide a training method for an order information extraction model, an order generation method, an order generation system, an electronic device, a storage medium and a computer program product to ensure the implementation and application of the above-mentioned order information acquisition method.

[0005] In order to solve the above problems, the present application discloses a method for obtaining order information, which includes:

[0006] Acquire a communication record for a target order, wherein the communication record includes: structured information and / or unstructured information;

[0007] Extracting communication text segments of a preset order information category from the unstructured information through a pre-trained order information extraction model;

[0008] Based on the preset information extraction rules, the structured information and the communication text fragment, the order information of the target order is obtained.

[0009] The embodiment of the present application also discloses an order generation method, which is applied to a server. The order generation method includes:

[0010] In response to an order information intelligent extraction request sent by a preset client, obtaining object information carried in the order information intelligent extraction request;

[0011] Acquire a communication record between the current user and the communication object to which the object information belongs regarding the target order, wherein the communication record includes: structured information and / or unstructured information;

[0012] Extracting communication text segments of a preset order information category from the unstructured information through a pre-trained order information extraction model;

[0013] Based on the preset information extraction rules, the structured information and the communication text fragment, obtaining candidate order information of the target order;

[0014] The candidate order information is sent to the preset client, so that the preset client displays the candidate order information and generates the target order based on part or all of the candidate order information.

[0015] The embodiment of the present application also discloses an order generation method, which is applied to a client. The order generation method includes:

[0016] In response to the order information acquisition operation, acquiring object information of the communication object input by the current user;

[0017] Sending an order information intelligent extraction request carrying the object information to a preset server, the order information intelligent extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtaining the communication record between the current user and the communication object for the target order, the communication record including: structured information and / or unstructured information; extracting communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; obtaining candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragments;

[0018] Obtaining and displaying the candidate order information;

[0019] The target order is generated based on part or all of the candidate order information.

[0020] The embodiment of the present application discloses an order generation system, including: a client and a server, wherein:

[0021] The client is used to obtain the object information of the communication object input by the current user in response to the order information acquisition operation;

[0022] The client is further used to send an order information intelligent extraction request carrying the object information to the server;

[0023] The server is used to obtain a communication record between the current user and the communication object regarding a target order, wherein the communication record includes: structured information and / or unstructured information;

[0024] The server is further configured to extract communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model;

[0025] The server is further configured to obtain candidate order information of the target order based on a preset information extraction rule, the structured information and the communication text fragment;

[0026] The client is also used to obtain and display the candidate order information;

[0027] The client is further configured to generate the target order based on part or all of the candidate order information.

[0028] The embodiment of the present application also discloses an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0029] The memory stores computer-executable instructions;

[0030] The processor executes the computer-executable instructions stored in the memory to implement the method described in the embodiment of the present application.

[0031] The embodiment of the present application further discloses a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method described in the embodiment of the present application.

[0032] The embodiment of the present application also discloses a computer program product, including a computer program / computer executable instructions, characterized in that when the computer program / computer executable instructions are executed by a processor in an electronic device, the method described in the embodiment of the present application is implemented.

[0033] Compared with the prior art, the embodiments of the present application have the following advantages:

[0034] By obtaining communication records for a target order, the communication records include: structured information and / or unstructured information; then, extracting communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; finally, based on preset information extraction rules, the structured information and the communication text fragments, obtaining order information of the target order, realizing intelligent extraction of order-related valid information in the communication records and performing structured processing to obtain structured order information, not only improving the efficiency of obtaining order information, but also improving the accuracy and comprehensiveness of order information acquisition, providing effective technical support for applications such as rapid order generation, order submission, and order file generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of the steps of the method for obtaining order information disclosed in the embodiment of the present application;

[0036] Figure 2 is a flowchart of the steps of the order generation method disclosed in the embodiment of the present application;

[0037] Figure 3 It is a schematic diagram of the order information editing interface in the order generation method disclosed in the embodiment of the present application;

[0038] Figure 4 is a flowchart of the steps of the order generation method disclosed in the embodiment of the present application;

[0039] Figure 5 It is a flowchart of the steps of the training method of the order information extraction model disclosed in the embodiment of the present application;

[0040] Figure 6 It is an interactive schematic diagram of the order generation system disclosed in the embodiment of the present application;

[0041] Figure 7 It is a schematic diagram of the structure of an exemplary device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0043] The order information acquisition method disclosed in the embodiment of the present application can be applied to the scenario of drafting an order based on the communication records between the two parties to the transaction, generating orders and other scenarios. In the prior art, the information in the communication process between the two parties to the transaction includes but is not limited to: files in formats such as text, pictures, Excel, etc., but in the order process, it is necessary to fill in the structured address (country, province, city, etc.), the name of the product, the product logo, the price of the product, logistics information and other information in the order data structure. In the prior art, it is usually manually flipping through the communication records, searching for order-related information, and copying and pasting them item by item into the order data structure to generate an order draft. The method of obtaining order data based on communication records in the prior art is not only time-consuming, but also prone to missing information, resulting in inaccurate order information obtained.

[0044] The order information acquisition method disclosed in the embodiment of the present application combines the deterministic order-related information of the communication process (such as product cards, address cards, etc.) and the communication text, communication files, etc. in the chat process, summarizes and refines the effective key information in the communication content through the artificial intelligence big model, and assigns values ​​to the structured order data. It can not only quickly and accurately generate order information, but also further quickly place orders and improve the ordering experience.

[0045] Reference Figure 1 The order information acquisition method disclosed in the embodiment of the present application includes: steps 102 to 106.

[0046] Step 102: Obtain a communication record for the target order, wherein the communication record includes: structured information and / or unstructured information.

[0047] Optionally, the target order may be an order between a buyer and a seller for one or more products. The communication record for the target order is the communication record during the process of the transaction parties reaching the target order. In some optional embodiments, the historical communication record may be queried based on the user information of the transaction parties, and the historical communication record of the transaction parties in the most recent specified time period may be obtained as the communication record of the transaction parties for the target order.

[0048] The structured information may be generated by an order system, a logistics system, an order generation system or other systems, and has a fixed data structure and clear data content. The structured information includes but is not limited to card-type information. Taking the e-commerce scenario as an example, the structured information includes but is not limited to any one of: product card, address card, and freight card. For example, the structured information may be a product card of a specified product shared by either party to the transaction through an e-commerce website during the communication process.

[0049] The specific contents of structured information and unstructured information are not limited in the embodiments of the present application.

[0050] The unstructured information includes but is not limited to any of the following: communication text, files, pictures, etc. Taking the e-commerce scenario as an example, the communication text may be the online communication content between the two parties regarding the target order, including but not limited to any one or more of the following: logistics information, address information, product information, etc. The file may be one or more of the following: quotation, contract, etc. The picture includes but is not limited to product pictures.

[0051] In the embodiments of the present application, there is no limitation on the specific implementation method of obtaining the communication record for the target order, and there is no limitation on the specific content of the communication record.

[0052] Next, the acquired communication records are further automatically processed. For structured information in the communication records, the formatted information can be parsed according to the inherent format of the structured information, thereby extracting the data content of specific fields in the structured information. For unstructured information,

[0053] Step 104: extracting communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model.

[0054] Next, the acquired communication records are further processed automatically.

[0055] In the embodiments of the present application, artificial intelligence technology is used in combination with rule matching to process the unstructured information in the communication records to extract the order information in text form.

[0056] In some optional embodiments, the order information extraction model obtained through pre-training extracts communication text fragments of a preset order information category from the unstructured information, including: performing text splicing on the communication records in the unstructured information to obtain a communication record spliced ​​text; constructing a prompt text corresponding to the preset order information category based on the communication record spliced ​​text; and calling the order information extraction model based on the prompt text to trigger the order information extraction model to extract communication text fragments of a preset order information category from the communication record spliced ​​text.

[0057] Optionally, the preset order information categories include one or more of the following: address category, logistics category, freight category, product category. In some optional embodiments, a prompt word template can be pre-configured for each preset order information category, and in the actual operation process, the corresponding prompt word template is further instantiated through the communication record splicing text obtained in real time, so as to obtain the prompt text matching the order information category.

[0058] Optionally, the communication records can be spliced ​​in the order of the communication contents in the communication records to obtain the spliced ​​text of the communication records. In some optional embodiments, if the communication records contain files, the files contained in the communication records can be opened first, and the file contents can be read as part of the communication contents. For example, a spliced ​​text of the communication records in the form of: "The buyer said: XXXXX, the seller said: YYYYYYYY, the buyer said: AAAAAA, the buyer said: BBBBB" can be obtained.

[0059] Afterwards, based on the communication record splicing text, a prompt word for instructing the order information extraction model to extract the communication text containing the address information can be generated according to the specified prompt text generation rules. For example, a prompt text including the following content can be generated: "As a transaction expert, describe the communication progress and results of the current transaction parties, and extract the specified relevant information: 1. Whether [stage a] is involved; 1e. If [stage a] is involved, extract the address / other information elements mentioned in [stage a]; 2. Whether [stage b] is involved; 2e. If [stage b] is involved, extract the address / other information elements mentioned in [stage b]; 3. Whether [stage b] is involved; 3e. If [stage c] is involved, extract the address / other information elements mentioned in [stage c]; 4. Summarize the final stages involved; 4e. Summarize the final extracted information elements according to importance. The output format is as follows: [output format example]; The communication content of the transaction parties is as follows: the buyer said: XXXXX, the seller said: YYYYYYYY, the buyer said: AAAAAA, the buyer said: BBBBB".

[0060] Then, the order information extraction model is called based on the generated prompt word, and the order information extraction model is triggered to extract the communication segment that discusses the order element information from the communication record splicing text, wherein the order element information includes but is not limited to: address information, price information, transaction progress, etc. For example, the communication segment that discusses the address information is extracted, wherein the address information includes but is not limited to one or more of the following: country information, postal code, telephone number, geographic location, etc.

[0061] The order information extraction model performs one-stage reasoning based on the progressive understanding ability learned during training to obtain information text of a specified order information category. For example, after reasoning, a sentence containing an address in the communication record concatenation text is identified.

[0062] The training method of the order information extraction model is described below and will not be repeated here.

[0063] Step 106, based on the preset information extraction rules, the structured information and the communication text fragment, obtain the order information of the target order.

[0064] In some optional embodiments, the order information of the target order is obtained based on the preset information extraction rules, the structured information and the communication text fragment, including: based on the preset information extraction rules, structuring the communication text fragment to obtain first structured order information; extracting information from the structured information to obtain second structured order information; generating the order information of the target order based on the first structured order information and the second structured order information.

[0065] In the embodiments of the present application, based on the data structure characteristics of structured information and unstructured communication books, different data processing methods are used to extract information, thereby obtaining various structured data used to constitute order information.

[0066] The following describes the data processing methods for structured information and unstructured communication texts respectively.

[0067] 1. Structured information

[0068] For the structured information in the communication record, the formatted information may be parsed according to the inherent data format of the structured information, thereby extracting the data content of a specific field in the structured information.

[0069] The structured information refers to deterministic information. Optionally, the structured information includes but is not limited to: address cards, product cards, freight cards, etc.

[0070] Taking the structured information as a product card as an example, after extracting the structured information, one or more of the following second structured order information including but not limited to: product image, product name, product introduction, and product price can be obtained.

[0071] 2. Unstructured Communication Text

[0072] For the communication text containing the preset order information category identified by the order information extraction model from the communication records, the key information associated with the order is extracted from it in combination with the rule configuration, and the extracted key information is structured to obtain structured order information.

[0073] Optionally, based on preset information extraction rules, the communication text fragment is structured to obtain the first structured order information, including: identifying index information of the communication text to obtain the index information contained in the communication text; obtaining preset configuration data matching the index information; extracting information from the communication text based on the preset configuration data to obtain the first structured order information. Wherein, the index information includes but is not limited to: country, logistics type, commodity category, etc. Wherein, the index information corresponds to the meaning of the preset configuration information, and the preset configuration data can be retrieved through the index information. The specific content of the index information is determined according to the specific application scenario. Optionally, the preset configuration information includes but is not limited to one or more of the following information: a list of information to be extracted, information extraction rules, information structuring rules, etc. Taking the extraction of addresses in communication texts as an example, the index information can be a country name, and the preset configuration data can be a list of addresses in different countries, address extraction rules, address structuring rules, etc.

[0074] Optionally, entity recognition technology can be used to identify index information in the communication text. After that, text matching technology or artificial intelligence technology can be further used to extract other information to be extracted in the communication text with reference to the information extraction rule. Furthermore, the extracted data can be structured according to the information structuring rule in the preset configuration data to obtain the first structured order information.

[0075] Optionally, the preset order information category includes: address, and the first structured order information includes: structured address information, and the communication text fragments are structured based on the preset information extraction rules to obtain the first structured order information, including: performing country identification on the communication text fragments of the address class to obtain the country information contained in the communication text; obtaining preset address form configuration data matching the country information; extracting address information at the country level for the communication text fragments of the address class based on the preset address form configuration data to obtain hierarchical address information; and performing structured processing on the country information and the hierarchical address information to obtain the structured address information.

[0076] In some optional embodiments, an address form may be configured for each country in advance to obtain address form configuration data. The address form stores address information such as the country's area code, city name, street name, and address resolution rules, address structuring rules, and other information. Entity recognition technology may be used to identify the country name in the communication text, and the address form configured for the country may be retrieved based on the country name.

[0077] In some optional embodiments, the address information at the country level is extracted from the communication text fragment of the address class based on the preset address form configuration data to obtain hierarchical address information, including: identifying the first address information at the country level contained in the communication text fragment of the address class based on the hierarchical address list in the preset address form configuration data; identifying the second address information at the country level contained in the communication text of the address class based on geocoding or a preset artificial intelligence model; and structuring the first address information and the second address information based on the address structure configuration information in the preset address form configuration data to obtain the structured address information.

[0078] The first address information includes but is not limited to: postal code, telephone number, email address, etc., and the second address information includes but is not limited to: city, street, house number, etc. The address structure configuration information includes but is not limited to: the hierarchical structure of the address.

[0079] During the specific implementation process, after obtaining the address form of the target country, the first address information included in the communication record can be further identified according to the address resolution rules configured in the address form, and the second address information included in the communication record can be identified according to the address resolution rules configured in the address form by using geocoding query or artificial intelligence technology. Thereafter, according to the address hierarchy structure configured in the address form, the country information, the first address information and the second address information identified from the communication record are structured to obtain structured address information.

[0080] Taking the communication text "The receipt is No. Y, YY Street, XX District, in Z City" as an example, based on the preset address extraction rules, when performing structured processing on this communication text, the key information associated with the order address "No. Y, YY Street, XX District" and "Z City" can be extracted first. Then, according to the preset format of structured address information, the above key information identified from the communication text is structured to obtain the structured address information "No. Y, YY Street, XX District, Z City".

[0081] Optionally, generating order information of the target order based on the first structured order information and the second structured order information includes: using the first structured order information and the second structured order information as candidate order information of the target order; and selecting part or all of the candidate order information as order information of the target order.

[0082] In some optional embodiments, the server may first further process the structured information extracted from the communication record and the statements containing the order information to generate structured order information. Thereafter, the generated structured order information is sent to the client as candidate order information for display, and the client displays the candidate order information. Then, according to the candidate order information selected by the user, the corresponding fields of the structured order data of the target order are assigned values, thereby obtaining the structured order information of the target order and quickly generating an order.

[0083] In some optional embodiments, the client may also select candidate order information according to the default configuration, automatically assign values ​​to corresponding fields of the structured order data of the target order, thereby obtaining the structured order information of the target order and quickly generating an order.

[0084] In some optional embodiments, after the country information and the hierarchical address information are structured to obtain the structured address information, the method further includes: evaluating the structured address information to obtain an address evaluation result; when the address evaluation result indicates that the structured address information meets the evaluation condition, determining that the structured address information is parsed successfully; when the address evaluation result indicates that the structured address information does not meet the evaluation condition, determining that the structured address information fails to be parsed. Optionally, the evaluation condition includes but is not limited to: the field content in the structured address information is complete.

[0085] In summary, the order information acquisition method disclosed in the example of the present application obtains communication records for a target order, wherein the communication records include: structured information and / or unstructured information; then, a communication text segment of a preset order information category is extracted from the unstructured information through a pre-trained order information extraction model; finally, based on preset information extraction rules, the structured information and the communication text segment, the order information of the target order is obtained, and intelligent extraction of order-related valid information in the communication record and structural processing are performed to obtain structured order information, which not only improves the efficiency of obtaining order information, but also improves the accuracy and comprehensiveness of order information acquisition, and provides effective technical support for applications such as quickly generating orders, submitting orders, and generating order files.

[0086] Based on the above embodiments, the present application also discloses an order generation method, which is applied to a client of an order generation system, such as Figure 2 As shown, the method includes: steps 202 to 208.

[0087] Step 202: In response to the order information acquisition operation, acquire the object information of the communication object input by the current user.

[0088] The order information acquisition operation is triggered by the user through the client of the order generation system. For example, the order information acquisition operation can be an operation in which the user inputs the object information of the communication object in the client interface.

[0089] The client settings of the order generation system are as follows Figure 3 Taking the order information editing interface shown in the figure as an example, an object information input entry of the communication object can be set in the client interface, and the current user can enter the object information of the communication object through the object information input entry. The object information includes but is not limited to any one or more of the following: user name, email address, telephone number, etc. The current user of the client can enter the object information through the object information input entry, thereby triggering the order information acquisition operation.

[0090] Taking the current user as the seller user as an example, the seller user can select the user name of a buyer with whom the user has recently communicated through the buyer user selection entrance set in the client interface, as the object information of the communication object input by the current user; or, the seller user can enter one or more types of user information of the buyer user, such as the buyer user's email address, user name, contact information, etc., through the buyer user information input box set in the client interface, as the object information of the communication object input by the current user.

[0091] In some other application scenarios, the current user may also be a buyer user, and the corresponding communication object input by the current user is a seller user.

[0092] Optionally, the communication object input by the current user is a user in the order generation system who has a historical communication record with the current logged-in user of the client.

[0093] Step 204: Send an intelligent order information extraction request carrying the object information to a preset server, wherein the intelligent order information extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtain the communication record between the current user and the communication object regarding the target order, wherein the communication record includes: structured information and / or unstructured information; extract communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; and obtain candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragments.

[0094] Optionally, the order information intelligent extraction request carries the object information and the user information of the current user.

[0095] Optionally, the candidate order information includes but is not limited to one or more of the following: delivery address, product information, and shipping cost.

[0096] After the client detects the object information of the communication object input by the current user, it further encapsulates the object information of the communication object and the user information of the current user to obtain an order information intelligent extraction request, and sends the order information intelligent extraction request to the server that has pre-established a connection with the client.

[0097] After receiving the order information intelligent extraction request, the server first parses the order information intelligent extraction request to obtain the object information of the communication object input by the current user.

[0098] Afterwards, the server obtains the communication record between the current user and the communication object regarding the target order, wherein the communication record includes: structured information and / or unstructured information. The specific implementation method of obtaining the communication record between the current user and the communication object regarding the target order is described in the above specific implementation method of obtaining the communication record between the transaction parties regarding the target order, which will not be repeated here.

[0099] Next, the server extracts communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model, and obtains the order information of the target order based on preset information extraction rules, the structured information and the communication text fragments.

[0100] Optionally, the order information extraction model obtained through pre-training extracts communication text segments of a preset order information category from the unstructured information, including: performing text splicing on the communication records in the unstructured information to obtain a communication record splicing text; constructing a prompt text corresponding to the preset order information category based on the communication record splicing text; calling the order information extraction model based on the prompt text, triggering the order information extraction model to extract communication text segments of a preset order information category from the communication record splicing text. For the specific implementation method of extracting communication text segments of a preset order information category from the unstructured information by the order information extraction model obtained through pre-training on the server side, please refer to the relevant description in the previous embodiment, which will not be repeated here.

[0101] For the specific implementation method of obtaining the candidate order information of the target order based on the preset information extraction rules, the structured information and the communication text fragments, please refer to the relevant description of obtaining the order information of the target order based on the preset information extraction rules, the structured information and the communication text fragments in the previous embodiment, which will not be repeated here.

[0102] After obtaining the candidate order information of the target order, the server sends the candidate order information to the client for display.

[0103] Step 206: Acquire and display the candidate order information.

[0104] Optionally, the client obtains the candidate order information of the target order obtained by the server and displays it on the order information editing interface of the client. Figure 3 As shown, the content displayed in the order information editing interface includes but is not limited to: order address, product information of the order product.

[0105] Step 208: Generate the target order based on part or all of the candidate order information.

[0106] In some optional embodiments, the order information editing interface further includes: an area for displaying buyer information and an area for displaying order information. The order information includes, but is not limited to, a delivery address and product information of the ordered product. The order information is structured information. For example, the delivery address includes, but is not limited to, one or more of the following fields: telephone number, country, state / province, city, zip code, detailed address, etc.

[0107] In some optional embodiments, the generating the target order based on part or all of the candidate order information includes: automatically filling in the corresponding order information fields of the target order with the default candidate order information in the candidate order information to generate an order to be confirmed; displaying the order information of the order to be confirmed; and generating the target order based on the order information of the order to be confirmed in response to a confirmation operation on the order information of the order to be confirmed. The default candidate order information includes but is not limited to: the first delivery address, the last delivery address, all product information, and the last freight information.

[0108] For example, after the client obtains the candidate order information sent by the server, the client uses the first delivery address in the address information included in the candidate order information to assign values ​​to the various fields of the address information of the order to be confirmed by default. At the same time, the order product list of the order to be confirmed is initialized using the product information of all products in the product information included in the candidate order information. In addition, the last freight information in the freight information included in the candidate order information is used to assign values ​​to the freight information of the order to be confirmed, thereby generating order data for the order to be confirmed.

[0109] In some optional embodiments, generating the target order based on part or all of the candidate order information includes: filling corresponding order information fields of the target order with the candidate order information selected by the current user to generate an order to be confirmed; displaying the order information of the order to be confirmed; and generating the target order based on the order information of the order to be confirmed in response to a confirmation operation on the order information of the order to be confirmed.

[0110] For example, the user can select an address from the address list displayed on the order information editing interface. After the client detects the user's selection operation, it assigns values ​​to the corresponding fields of the order to be confirmed based on the fields in the address information selected by the user, thereby obtaining the structured address information of the order to be confirmed. In the same way, the user can select address information, product information, and freight information from the candidate order information. After that, the client fills the fields of the order to be confirmed displayed on the order information editing interface based on the information associated with the order selected by the user, thereby assigning values ​​to the fields of the order to be confirmed and obtaining the order information of the order to be confirmed.

[0111] After assigning values ​​to the fields of the order to be confirmed, the client refreshes and displays the order information of the order to be confirmed in real time.

[0112] In some optional embodiments, the order information editing interface may also include an application entry for order information, such as an order submission button, an order confirmation button, etc. The user may trigger the order submission button to perform a confirmation submission operation on the order to be confirmed, triggering the client to submit the target order based on the order content of the order to be confirmed. Alternatively, the user may trigger the order confirmation button to cause the client to generate order data for the target order based on the order data of the order to be confirmed, and store the target order.

[0113] In some optional embodiments, after displaying the order information of the order to be confirmed, the method further includes: in response to the current user's editing operation on the order to be confirmed, updating the order information of the corresponding order information field of the order to be confirmed, wherein the editing operation includes but is not limited to one or more of the following: modifying the delivery address, adding a new address, adding a new product, and deleting a product. That is, the user can reselect an address in the address list in the candidate order information as the delivery address, or reselect an order product in the candidate order information to add to the order products of the order to be confirmed, or delete the order products of the order to be confirmed, etc.

[0114] In summary, the order generation method disclosed in the embodiment of the present application obtains the object information of the communication object input by the current user through the client in response to the order information acquisition operation; then, an order information intelligent extraction request carrying the object information is sent to the preset server, and the order information intelligent extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtain the communication record between the current user and the communication object for the target order, and the communication record includes: structured information and / or unstructured information; extract the communication text fragment of the preset order information category from the unstructured information through the pre-trained order information extraction model; based on the preset information extraction rules, the structured information and the communication text fragment, obtain the candidate order information of the target order; then, the client obtains and displays the candidate order information; and based on part or all of the candidate order information, generates the target order, shortening the order placement time and fully improving the order generation efficiency. With this method, order information can be brought in with one click to quickly generate orders.

[0115] Based on the above embodiments, the present application also discloses an order generation method, which is applied to the server of the order generation system. Figure 4 As shown, the method includes: step 402 to step 410.

[0116] Step 402: In response to the order information intelligent extraction request sent by the preset client, obtain the object information carried in the order information intelligent extraction request.

[0117] The method for generating the order information intelligent extraction request is described above and will not be repeated here.

[0118] After receiving the order information intelligent extraction request, the server parses the order information intelligent extraction request and obtains the object information carried in the order information intelligent extraction request and the user information of the current user who logs in to the client.

[0119] Step 404: obtaining a communication record regarding the target order between the current user and the communication object to which the object information belongs, wherein the communication record includes: structured information and / or unstructured information.

[0120] Step 406: extract communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model.

[0121] Step 408, based on the preset information extraction rules, the structured information and the communication text fragment, obtain the candidate order information of the target order.

[0122] Step 410: Send the candidate order information to the preset client, so that the preset client displays the candidate order information and generates the target order based on part or all of the candidate order information.

[0123] The specific implementation of step 404 to step 410 refers to the relevant description of the previous embodiment, which will not be repeated here.

[0124] In summary, in the order generation method disclosed in the embodiment of the present application, the server side responds to the order information intelligent extraction request sent by the preset client, obtains the object information carried in the order information intelligent extraction request; obtains the communication record between the current user and the communication object to which the object information belongs for the target order, and the communication record includes: structured information and / or unstructured information; extracts communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; obtains candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragments; sends the candidate order information to the preset client, so that the preset client displays the candidate order information, and generates the target order based on part or all of the candidate order information, thereby shortening the order placement time and fully improving the order generation efficiency.

[0125] In order to implement the above embodiment, the present application also discloses a training method for an order information extraction model, such as Figure 5 As shown, the method includes: steps 502 to 508.

[0126] Step 502, based on the first preset task description, triggering the preset teacher model to extract information from the historical communication records to obtain a symbolic state described in natural language, wherein the symbolic state includes: a communication record segment, and an order information state that matches the communication record segment.

[0127] In the embodiment of the present application, the order information extraction model is built on the basis of a large language model and trained in a semi-supervised manner. In the process of training the order information extraction model, a small sample active learning mechanism is adopted. First, a teacher model built on a large language model is trained based on a small number of labeled samples, and the teacher model extracts the input communication records. After that, the extraction results output by the teacher model are used as model training supervision signals to guide the student model to learn information extraction capabilities.

[0128] Optionally, the preset teacher model and the preset student model are both large language models.

[0129] The historical communication record may be a communication record between the transaction parties regarding the transaction order. The first preset task description may be an instruction for calling the preset student model, which is used to describe the task to be performed by the preset teacher model, the format of input data and output data.

[0130] In a specific implementation, based on the first preset task description, a preset teacher model is called to extract information from the historical communication record, and a symbolic state of the natural language description output by the preset teacher model is obtained. Optionally, the symbolic state includes the following information: one or more communication record segments involving order elements included in the historical communication record, and the order information status matched by each communication record segment.

[0131] In some optional embodiments, the first preset task description heuristically calls the preset teacher model to extract the communication record fragment of the preset order stage, and the order information status matched by the communication record fragment. The order information status is used to indicate the order element information involved in the communication record fragment. The first preset task description calls the preset teacher model through guided instructions.

[0132] For example, taking the first preset task description as input, calling the preset teacher model, so that the preset teacher model extracts information from a set of communication records of the current order according to the first preset task description, and one or more communication record fragments associated with different order generation stages of the current order can be obtained, and each of the communication record fragments includes order element information of an order information status. Among them, the order generation stage is determined according to the specific application scenario. Optionally, the order generation stage includes but is not limited to one or more of the following: transaction not started, order not paid, payment completed, etc. The order element information includes but is not limited to one or more of the following: address, logistics information, freight, price, product name, and order progress, etc. The order information status includes but is not limited to one or more of the following: the two parties to the transaction discussed the price, the two parties to the transaction discussed the address, the transaction was reached, etc.

[0133] Optionally, the content of the first preset task description includes but is not limited to one or more of the following texts: problem description text, communication record text, output description text. For example, the content of the first preset task description can be expressed in the following format: "As a transaction expert, describe the communication progress and results of the current transaction parties, and extract the specified relevant information: the communication record is as follows: [the communication record splicing text of the transaction parties]; 1. Whether it involves [stage a]; 1e. If it involves [stage a], extract the address / other information elements mentioned in [stage a]; 2. Whether it involves [stage b]; 2e. If it involves [stage b], extract the address / other information elements mentioned in [stage b]; 3. Whether it involves [stage b]; 3e. If it involves [stage c], extract the address / other information elements mentioned in [stage c]; 4. Summarize the final stages involved; 4e. Summarize the final extracted information elements according to importance, and the output format is as follows: [output format example]".

[0134] The preset teacher model extracts information from the communication record text in the first preset task description according to the questions raised in the problem description text in the first preset task description, obtains the communication record segments that mention the order element information in the communication record text, and determines the order stage corresponding to each communication record segment, and the order information status matched by each communication record segment. For example, the order element information of the preset order stage extracted by the preset teacher model can be expressed as follows: communication record segment [i] - the transaction parties discussed [address information]; communication record segment [ii] - the transaction parties discussed the price; communication record segment [iii] - the two parties reached a transaction.

[0135] In the embodiment of the present application, guided instructions are used to allow the large language model to jointly describe the current state and results, without relying on corresponding symbolic actions, and to heuristically generate the corresponding symbolic state. Furthermore, based on the symbolic state, the order element information of the preset order stage can be obtained.

[0136] Step 504: construct a simulated interactive environment based on the communication record segment and the order information status, and obtain a feedback result through the simulated interactive environment, wherein the feedback result includes: a conversion signal corresponding to the communication record segment.

[0137] The conversion signal may be one or more application data indicators describing the transaction process or transaction result. For example, the conversion signal may be an indicator indicating that a transaction has been completed and an order has been placed, an indicator indicating that a transaction has been completed and a deep cooperative relationship has been reached, etc.

[0138] In the autonomous learning process of the large language model, the symbolic actions interact with the interactive environment, and corresponding feedback results can be obtained, which are used as supervisory signals for model updates. In the self-training paradigm of the prior art, each symbolic state output by the large language model needs to be executed in the ENV (i.e., interactive environment) to obtain feedback results.

[0139] In an embodiment of the present application, the communication record fragments and the matching order information status of the preset order stage output by the preset teacher model are the symbolic states output by the preset teacher model. According to the predefined conversion signal, an initial reward model is constructed to obtain a simulated interactive environment.

[0140] Optionally, the value of the conversion signal can be obtained through the order system. The order system determines the symbolic atomic state associated with the order according to a preset method, and sets the real conversion signal enumeration value corresponding to each symbolic atomic state. Among them, the symbolic atomic state can be obtained by abstracting the order state corresponding to the historical communication case. For example, for the symbolic atomic state "the two parties to the transaction discussed the address information", the value of its conversion signal can be set to 0.5, and for the symbolic atomic state "the two parties reached a transaction", the value of its conversion signal can be set to 1. The symbolic atomic state matched by a communication record fragment reflects the importance of the order element information contained in the communication record fragment. Therefore, the value of the conversion signal corresponding to the communication record fragment can be used as a supervisory signal for the optimization of the information extraction model in the future.

[0141] In the embodiments of the present application, there is no restriction on the types of symbolized atomic states. For example, the symbolized atomic states may include but are not limited to one or more of the following: the transaction parties discussed address information, the transaction parties discussed price, the transaction was concluded, etc. The real conversion signal enumeration value corresponding to the symbolized atomic state is determined according to the specific application scenario. In the embodiments of the present application, there is no restriction on the real conversion signal enumeration value corresponding to the symbolized atomic state.

[0142] Furthermore, by matching the order information state matched by the communication record fragment extracted from the communication record with the symbolic atomic state, a symbolic atomic state set corresponding to the communication record in the simulated interactive environment can be obtained. For example, a set of symbolic atomic states corresponding to a group of communication records can be represented as follows: communication record fragment [i] - the transaction parties discussed [address information]; communication record fragment [ii] - the transaction parties discussed the price; communication record fragment [iii] - the two parties reached a transaction. Afterwards, according to the true value of the conversion signal corresponding to the symbolic atomic state in the order system, the conversion signal value corresponding to the communication record can be obtained.

[0143] Optionally, the conversion signal value of the communication record obtained in the simulated interaction environment is used as a feedback result. The feedback result is used as a supervision signal for subsequent supervision tasks (ie, reward model).

[0144] In the embodiment of the present application, instead of relying on the interaction between the large language model itself and the actual environment, an evolutionary mechanism is created, and interactive situations are simulated with the help of communication history to achieve the transformation of the large language model from weak to strong, train the specific capabilities of the large language model, and solve the problem of information extraction in order scenarios.

[0145] Step 506, based on the communication record fragment, the order information status and the conversion signal, the preset teacher model is optimized in combination with the reward model.

[0146] After obtaining the communication record segments included in the historical communication records, the order information status matching each communication record segment, and the conversion signal corresponding to the historical communication record, the subsequent model optimization steps can be performed by referring to the self-training method of the large language model in the prior art.

[0147] In some optional embodiments, the preset teacher model is optimized based on the communication record fragment, the order information status and the conversion signal in combination with a reward model, including: based on the feedback result, obtaining a first supervision signal of the supervision task; based on the first supervision signal, the communication record fragment and the order information status, executing the supervision task corresponding to the second preset task description to generate a second supervision signal; based on the second supervision signal, updating the historical communication record; based on the historical communication record and a third preset task description, generating the communication record fragment of the preset order stage, the order information status corresponding to the communication record fragment, and the chain deduction conclusion through the updated preset teacher model to optimize the preset teacher model.

[0148] As a task corresponding to the reward model in the active learning architecture, the supervision task needs to fine-tune the large language model in combination with the supervision signal. In some optional embodiments, the conversion signal corresponding to each communication record segment obtained in the above steps can be used as the first supervision signal of the supervision task.

[0149] As mentioned above, the conversion signal corresponding to the communication record fragment is calculated according to a preset method based on the order-related data corresponding to the historical communication record, and can be used to describe one or more application data indicators of the transaction process or transaction results. Therefore, the first supervision signal can be used to evaluate the accuracy of order information extraction in the communication record.

[0150] Next, the supervision task corresponding to the second preset task description is executed based on the first supervision signal, the communication record segment and the order information status to generate a second supervision signal.

[0151] In an embodiment of the present application, the supervisory signal of the supervision task includes two parts: a manually input supervisory signal and a self-supervisory signal. Among them, the manual supervision signal is the sample label of the labeled training sample in the training sample, which is used as a part of the input supervisory signal of the supervision task; the automatic supervision signal is dynamically generated according to the communication record data associated with the training sample. During the model training process, intermediate information such as status is input as a self-supervisory signal through the supervision task (i.e., the reward model).

[0152] Optionally, the manually input supervision signal is a data pair of a communication record and a true value of a conversion signal, which can be expressed as follows, for example: [([communication record segment i_1], [true value of conversion signal s_1]), ([communication record segment i_2], [true value of conversion signal s_2]), ..., ([communication record segment i_n], [true value of conversion signal s_n])]. The method for obtaining the true value of the conversion signal is described above and will not be repeated here.

[0153] Optionally, the second preset task description is used to describe the task that triggers the preset teacher model to generate a supervision signal. In an embodiment of the present application, the second preset task description includes but is not limited to the following contents: question text, input data, and output description text. The input data is not limited to: communication record fragments and order information status extracted from historical communication records in the aforementioned steps. The content of the second preset task description is determined according to the specific application scenario. In the implementation of this application, the specific content of the second preset task description is not limited.

[0154] Optionally, the question text in the second preset task description is expressed as follows: As a trading expert, please evaluate the accuracy of the following extracted information for the trading process [communication record fragment i], and the evaluation criteria are as follows: [evaluation criterion criterion]… At the same time, in the second preset task description, the input data includes: manual supervision signal and self-supervision signal, wherein the self-supervision signal further includes: communication record fragment, order information status and communication evaluation result. For example, the self-supervised signal can be expressed as: [([communication record fragment i_1], [order information status e_1], [communication evaluation v_1]), ([communication record fragment i_2], [order information status e_2], [communication evaluation v_2]),..., ([communication record fragment i_n], [order information status e_n], [communication evaluation v_n])]; the artificial supervision signal can be expressed as: [([communication record fragment i_1], [communication conversion signal s_1]), ([communication record fragment i_2], [communication conversion signal s_2]),..., ([communication record fragment i_n], [communication conversion signal s_n])].

[0155] The preset teacher model performs the supervision task according to the question text and input data in the second preset task description, and generates a second supervision signal in a specified output format. Optionally, the second supervision signal is used to indicate the communication progress and results.

[0156] Optionally, the second supervisory signal includes but is not limited to: the conversion signal. The specific implementation of the preset teacher model generating the second supervisory signal according to the instruction can be found in the prior art and will not be described in detail here.

[0157] Afterwards, the accuracy of the communication record segment information extraction is determined based on the second supervisory signal, and the model parameters of the preset teacher model are updated.

[0158] Next, based on the historical communication records and the third preset task description, the communication record segments of the preset order stage, the order information status corresponding to the communication record segments, and the chain deduction conclusions are generated through the updated preset teacher model to optimize the preset teacher model.

[0159] The third preset task description is used to describe the optimization task of the preset teacher model. Optionally, the third preset task description includes one or more of the following contents: question text, input data, and output description text. Among them, the input data includes: updated historical communication records as training samples.

[0160] The third preset task description can be expressed as follows: As a transaction expert, describe the communication progress and results between the two parties of the current transaction, which is performed in two steps, extracting i) symbolic state natural language (NL) description; ii) chain deduction conclusion, you need to extract the specified relevant information: 1. Whether it involves [stage a]; 1e. If it involves [stage a], extract the address / other information elements mentioned in [stage a]; 2. Whether it involves [stage b]; 2e. If it involves [stage b], extract the address / other information elements mentioned in [stage b]; 3. Whether it involves [stage b]; 3e. If it involves [stage c], extract the address / other information elements mentioned in [stage c]; 4. Summarize the final stages involved; 4e. Summarize the final extracted information elements according to importance. The output format is as follows: [output format example].

[0161] The specific implementation scheme of the preset teacher model generating corresponding output data based on the input third preset task description can be found in the text generation scheme of the large language model, which will not be repeated here.

[0162] After the aforementioned steps, the preset teacher model fully learns the ability to extract order element information from communication records by performing self-training.

[0163] In the embodiments of the present application, other methods may also be used to optimize the preset teacher model, which will not be listed one by one in the embodiments of the present application.

[0164] Step 508: jointly train a preset student model based on the optimized preset teacher model, and use the trained preset student model as the order information extraction model.

[0165] For example, the symbolic state output by the preset teacher model when performing the aforementioned task (such as a communication record fragment, and the order information state matched by the communication record fragment) is used as the output supervision signal of the preset student model, the student model is trained, and the trained student model is used as the order information extraction model.

[0166] For the specific implementation method of jointly training the preset student model based on the preset teacher model obtained by optimization, please refer to the relevant scheme of jointly training the teacher model and the student model in the self-training architecture in the prior art, which will not be repeated in the embodiments of this application.

[0167] In summary, the training method of the order information extraction model disclosed in the embodiment of the present application triggers the preset teacher model based on the first preset task description to extract information from the historical communication records, and obtains the symbolic state of the natural language description, wherein the symbolic state includes: a communication record segment, and an order information status matching the communication record segment; and a simulated interaction environment is formed based on the communication record segment and the order information status, and feedback results are obtained through the simulated interaction environment, wherein the feedback results include: a conversion signal corresponding to the communication record segment; then, based on the communication record segment, the order information status and the conversion signal, the preset teacher model is optimized in combination with a reward model; finally, the preset student model is jointly trained based on the optimized preset teacher model, and the trained preset student model is used as the order information extraction model, so that the model extracts the order element information in the communication record based on the understanding of the transaction flow, thereby further improving the accuracy of the order element information extraction.

[0168] In order to implement the above embodiment, the present application also discloses an order generation system, such as Figure 6 As shown, the order generation system includes: a client 602 and a server 604, wherein:

[0169] The client 602 is used to obtain the object information of the communication object input by the current user in response to the order information acquisition operation;

[0170] The client 602 is further used to send an order information intelligent extraction request carrying the object information to the server 604;

[0171] The server 604 is used to obtain the communication record between the current user and the communication object regarding the target order, wherein the communication record includes: structured information and / or unstructured information;

[0172] The server 604 is further configured to extract communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model;

[0173] The server 604 is further configured to obtain candidate order information of the target order based on a preset information extraction rule, the structured information and the communication text segment;

[0174] The client 602 is also used to obtain and display the candidate order information;

[0175] The client 602 is further configured to generate the target order based on part or all of the candidate order information.

[0176] The specific implementation methods of the client and the server refer to the relevant description in the previous embodiments and will not be repeated here.

[0177] In summary, the embodiment of the present application also discloses an order generation system, which obtains the object information of the communication object input by the current user in response to the order information acquisition operation by the client; and sends the server an order information intelligent extraction request carrying the object information; triggers the server to obtain the communication record between the current user and the communication object for the target order, and the communication record includes: structured information and / or unstructured information; then, extracts the communication text fragments of the preset order information category from the unstructured information through the pre-trained order information extraction model; and obtains the candidate order information of the target order based on the preset information extraction rules, the structured information and the communication text fragments; then, the client obtains and displays the candidate order information, and generates the target order based on part or all of the candidate order information, thereby shortening the order placement time and fully improving the order generation efficiency.

[0178] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0179] On the basis of the above embodiment, this embodiment further provides an order information acquisition device, the device comprising:

[0180] A communication record acquisition module, used to acquire communication records for a target order, wherein the communication records include: structured information and / or unstructured information;

[0181] A communication record segment extraction module, used to extract communication text segments of a preset order information category from the unstructured information through a pre-trained order information extraction model;

[0182] The order information acquisition module is used to acquire the order information of the target order based on a preset information extraction rule, the structured information and the communication text fragment.

[0183] Optionally, the order information extraction model obtained through pre-training extracts communication text segments of a preset order information category from the unstructured information, including:

[0184] Performing text splicing on the communication records in the unstructured information to obtain a communication record splicing text;

[0185] Constructing a prompt text corresponding to the preset order information category based on the communication record splicing text;

[0186] The order information extraction model is called based on the prompt text, and the order information extraction model is triggered to extract communication text segments of a preset order information category from the communication record splicing text.

[0187] Optionally, the preset order information categories include one or more of the following: address category, logistics category, freight category, and product category.

[0188] Optionally, acquiring the order information of the target order based on a preset information extraction rule, the structured information, and the communication text fragment includes:

[0189] Based on a preset information extraction rule, the communication text fragment is structured to obtain first structured order information;

[0190] Extracting information from the structured information to obtain second structured order information;

[0191] Based on the first structured order information and the second structured order information, order information of the target order is generated.

[0192] Optionally, the preset order information category includes: address, the first structured order information includes: structured address information, and the communication text fragment is structured based on the preset information extraction rule to obtain the first structured order information, including:

[0193] Performing country identification on the communication text segment of the address class to obtain country information contained in the communication text;

[0194] Obtaining preset address form configuration data matching the country information;

[0195] Extracting address information at the country level from the communication text fragment of the address class based on the preset address form configuration data to obtain hierarchical address information;

[0196] The country information and the hierarchical address information are structured to obtain the structured address information.

[0197] The order information acquisition device disclosed in the embodiment of the present application is used to implement the above-mentioned order information acquisition method. The specific implementation methods of each module of the device refer to the specific implementation methods of the corresponding steps in the above-mentioned method embodiment, which will not be repeated here.

[0198] In summary, the order information acquisition device disclosed in the example of the present application obtains communication records for a target order, wherein the communication records include: structured information and / or unstructured information; then, a communication text segment of a preset order information category is extracted from the unstructured information through a pre-trained order information extraction model; finally, based on preset information extraction rules, the structured information and the communication text segment, the order information of the target order is obtained, and intelligent extraction of order-related valid information in the communication record and structural processing are performed to obtain structured order information, which not only improves the efficiency of obtaining order information, but also improves the accuracy and comprehensiveness of order information acquisition, and provides effective technical support for applications such as quickly generating orders, submitting orders, and generating order files.

[0199] On the basis of the above embodiment, this embodiment further provides an order generation device, which is applied to a client, and the device includes:

[0200] A communication object information acquisition module, used to obtain object information of a communication object input by a current user in response to an order information acquisition operation;

[0201] An order information intelligent extraction request sending module is used to send an order information intelligent extraction request carrying the object information to a preset server, and the order information intelligent extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtaining the communication record between the current user and the communication object for the target order, and the communication record includes: structured information and / or unstructured information; extracting communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; based on preset information extraction rules, the structured information and the communication text fragments, obtaining candidate order information of the target order;

[0202] A candidate order information acquisition module, used to acquire and display the candidate order information;

[0203] An order generation module is used to generate the target order based on part or all of the candidate order information.

[0204] Optionally, generating the target order based on part or all of the candidate order information includes:

[0205] Automatically fill in the corresponding order information fields of the target order with the default candidate order information in the candidate order information to generate an order to be confirmed;

[0206] Displaying order information of the order to be confirmed;

[0207] In response to a confirmation operation on the order information of the order to be confirmed, the target order is generated based on the order information of the order to be confirmed.

[0208] Optionally, generating the target order based on part or all of the candidate order information includes:

[0209] Filling the corresponding order information field of the target order with the candidate order information selected by the current user to generate an order to be confirmed;

[0210] Displaying order information of the order to be confirmed;

[0211] In response to a confirmation operation on the order information of the order to be confirmed, the target order is generated based on the order information of the order to be confirmed.

[0212] Optionally, the candidate order information includes but is not limited to one or more of the following: delivery address, product information, and shipping cost.

[0213] The order generation device disclosed in the embodiment of the present application is used to implement the above-mentioned order generation method. The specific implementation methods of each module of the device refer to the specific implementation methods of the corresponding steps in the above-mentioned method embodiment, which will not be repeated here.

[0214] In summary, the order generation device disclosed in the example of the present application obtains the object information of the communication object input by the current user through the client in response to the order information acquisition operation; then, sends an order information intelligent extraction request carrying the object information to the preset server, and the order information intelligent extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtain the communication record between the current user and the communication object for the target order, and the communication record includes: structured information and / or unstructured information; extract the communication text fragment of the preset order information category from the unstructured information through the pre-trained order information extraction model; based on the preset information extraction rules, the structured information and the communication text fragment, obtain the candidate order information of the target order; then, the client obtains and displays the candidate order information; and based on part or all of the candidate order information, generates the target order, shortening the order placement time and fully improving the order generation efficiency. With this method, order information can be brought in with one click to quickly generate an order.

[0215] On the basis of the above embodiment, this embodiment further provides an order generation device, which is applied to a server, and the device includes:

[0216] A communication object information acquisition module, used to respond to an order information intelligent extraction request sent by a preset client, and acquire the object information carried in the order information intelligent extraction request;

[0217] A communication record acquisition module, used to acquire a communication record between the current user and the communication object to which the object information belongs regarding the target order, wherein the communication record includes: structured information and / or unstructured information;

[0218] A communication text segment extraction module, used to extract communication text segments of a preset order information category from the unstructured information by using a pre-trained order information extraction model;

[0219] A candidate order information acquisition module, used to acquire candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragment;

[0220] The candidate order information sending module is used to send the candidate order information to the preset client, so that the preset client displays the candidate order information and generates the target order based on part or all of the candidate order information.

[0221] The order generation device disclosed in the embodiment of the present application is used to implement the above-mentioned order generation method. The specific implementation methods of each module of the device refer to the specific implementation methods of the corresponding steps in the above-mentioned method embodiment, which will not be repeated here.

[0222] In summary, in the order generation device disclosed in the example of the present application, the server side responds to the order information intelligent extraction request sent by the preset client, obtains the object information carried in the order information intelligent extraction request; obtains the communication record between the current user and the communication object to which the object information belongs for the target order, and the communication record includes: structured information and / or unstructured information; extracts communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; obtains candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragments; sends the candidate order information to the preset client, so that the preset client displays the candidate order information, and generates the target order based on part or all of the candidate order information, thereby shortening the order placement time and fully improving the order generation efficiency.

[0223] On the basis of the above embodiment, this embodiment further provides a training device for an order information extraction model, the device comprising:

[0224] A symbolic state generation module is used to trigger a preset teacher model to extract information from historical communication records based on a first preset task description to obtain a symbolic state described in natural language, wherein the symbolic state includes: a communication record segment, and an order information state matched by the communication record segment;

[0225] A feedback result acquisition module, configured to form a simulated interactive environment based on the communication record segment and the order information status, and acquire feedback results through the simulated interactive environment, wherein the feedback results include: a conversion signal corresponding to the communication record segment;

[0226] A model optimization module, configured to optimize the preset teacher model based on the communication record fragment, the order information status and the conversion signal in combination with a reward model;

[0227] A joint training module is used to jointly train a preset student model based on the optimized preset teacher model, and use the trained preset student model as the order information extraction model.

[0228] Optionally, the first preset task description heuristically calls the preset teacher model to extract the communication record fragment of the preset order stage, and the order information status matched by the communication record fragment.

[0229] The training device for the order information extraction model disclosed in the embodiment of the present application is used to implement the training method of the above-mentioned order information extraction model. The specific implementation methods of each module of the device refer to the specific implementation methods of the corresponding steps in the aforementioned method embodiment, which will not be repeated here.

[0230] In summary, the training device of the order information extraction model disclosed in the embodiment of the present application triggers the preset teacher model based on the first preset task description to extract information from the historical communication records, and obtains the symbolic state of the natural language description, wherein the symbolic state includes: a communication record segment, and an order information status matching the communication record segment; and a simulated interaction environment is formed based on the communication record segment and the order information status, and feedback results are obtained through the simulated interaction environment, wherein the feedback results include: a conversion signal corresponding to the communication record segment; then, based on the communication record segment, the order information status and the conversion signal, the preset teacher model is optimized in combination with the reward model; finally, the preset student model is jointly trained based on the optimized preset teacher model, and the trained preset student model is used as the order information extraction model, so that the model extracts the order element information in the communication record based on the understanding of the transaction flow, thereby further improving the accuracy of the order element information extraction.

[0231] The embodiment of the present application also provides a non-volatile readable storage medium, which stores one or more modules (programs). When the one or more modules are applied to a device, the device can execute instructions (instructions) of each method step in the embodiment of the present application.

[0232] The embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method described in the embodiment of the present application.

[0233] The embodiment of the present application also provides an electronic device, comprising: a processor, and a memory connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method described in the embodiment of the present application. In the embodiment of the present application, the electronic device includes a server, a terminal device, and other devices.

[0234] The embodiment of the present application also discloses a computer program product, including a computer program / computer executable instructions, characterized in that when the computer program / computer executable instructions are executed by a processor in an electronic device, the method described in the embodiment of the present application is implemented.

[0235] The embodiments of the present disclosure may be implemented as a device configured as desired using any appropriate hardware, firmware, software, or any combination thereof, and the device may include electronic devices such as a server (cluster), a terminal, etc. Figure 7 An exemplary apparatus 700 that can be used to implement various embodiments described in this application is schematically illustrated.

[0236] For one embodiment, Figure 7 An exemplary apparatus 700 is shown having one or more processors 702, a control module (chip set) 704 coupled to at least one of the processor(s) 702, a memory 706 coupled to the control module 704, a non-volatile memory (NVM) / storage device 708 coupled to the control module 704, one or more input / output devices 710 coupled to the control module 704, and a network interface 712 coupled to the control module 704.

[0237] The processor 702 may include one or more single-core or multi-core processors, and the processor 702 may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). In some embodiments, the device 700 can be used as a server, terminal, or other device described in the embodiments of the present application.

[0238] In some embodiments, the apparatus 700 may include one or more computer-readable media (e.g., memory 706 or NVM / storage device 708) having instructions 714 and one or more processors 702 combined with the one or more computer-readable media and configured to execute the instructions 714 to implement a module to perform the actions described in the present disclosure.

[0239] For one embodiment, the control module 704 may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) 702 and / or any suitable device or component in communication with the control module 704 .

[0240] The control module 704 may include a memory controller module to provide an interface to the memory 706. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0241] The memory 706 may be used, for example, to load and store data and / or instructions 714 for the device 700. For one embodiment, the memory 706 may include any suitable volatile memory, such as a suitable DRAM. In some embodiments, the memory 706 may include double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0242] For one embodiment, control module 704 may include one or more input / output controllers to provide an interface to NVM / storage device 708 and input / output device(s) 710 .

[0243] For example, NVM / storage 708 may be used to store data and / or instructions 714. NVM / storage 708 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable non-volatile storage device(s) (e.g., one or more hard disk drives (HDDs), one or more compact disk (CD) drives, and / or one or more digital versatile disk (DVD) drives).

[0244] NVM / storage device 708 may include storage resources that are part of the device on which apparatus 700 is installed, or it may be accessible to the device without being part of the device. For example, NVM / storage device 708 may be accessed via (one or more) input / output devices 710 over a network.

[0245] (One or more) input / output devices 710 may provide an interface for the apparatus 700 to communicate with any other appropriate device, and the input / output device 710 may include a communication component, an audio component, a sensor component, etc. The network interface 712 may provide an interface for the apparatus 700 to communicate through one or more networks, and the apparatus 700 may wirelessly communicate with one or more components of a wireless network according to any of one or more wireless network standards and / or protocols, for example, accessing a wireless network based on a communication standard, such as Bluetooth, WiFi, 2G, 3G, 4G, 5G, etc., or a combination thereof for wireless communication.

[0246] For one embodiment, at least one of the processor(s) 702 may be packaged together with the logic of one or more controllers (e.g., a memory controller module) of the control module 704. For one embodiment, at least one of the processor(s) 702 may be packaged together with the logic of one or more controllers of the control module 704 to form a system-in-package (SiP). For one embodiment, at least one of the processor(s) 702 may be integrated on the same die with the logic of one or more controllers of the control module 704. For one embodiment, at least one of the processor(s) 702 may be integrated on the same die with the logic of one or more controllers of the control module 704 to form a system-on-chip (SoC).

[0247] In various embodiments, the apparatus 700 may be, but is not limited to, a terminal device such as a server, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). In various embodiments, the apparatus 700 may have more or fewer components and / or a different architecture. For example, in some embodiments, the apparatus 700 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0248] Among them, the main control chip can be used as a processor or control module in the detection device, sensor data, location information, etc. are stored in a memory or NVM / storage device, the sensor group can be used as an input / output device, and the communication interface may include a network interface.

[0249] The embodiment of the present application also provides an electronic device, including: a processor; and a memory, on which executable code is stored, and when the executable code is executed, the processor executes one or more methods as described in the embodiment of the present application. In the embodiment of the present application, the memory can store various data, such as target files, file and application association data, and user behavior data, so as to provide a data basis for various processing.

[0250] The embodiments of the present application also provide one or more machine-readable media on which executable codes are stored. When the executable codes are executed, the processor executes one or more methods described in the embodiments of the present application.

[0251] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0252] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0253] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0254] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0255] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable terminal device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0256] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0257] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0258] The above is a detailed introduction to an order information acquisition method, an order generation method, an order generation system, an order information extraction model training method, an electronic device, a storage medium and a computer program product provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for obtaining order information, characterized in that: The method comprises: Acquire a communication record for a target order, wherein the communication record includes: structured information and / or unstructured information; Extracting communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model; Based on the preset information extraction rules, the structured information and the communication text fragment, the order information of the target order is obtained.

2. The method according to claim 1, characterized in that The order information extraction model obtained by pre-training extracts communication text segments of a preset order information category from the unstructured information, including: Performing text splicing on the communication records in the unstructured information to obtain a communication record splicing text; Constructing a prompt text corresponding to the preset order information category based on the communication record splicing text; The order information extraction model is called based on the prompt text, and the order information extraction model is triggered to extract communication text segments of a preset order information category from the communication record splicing text.

3. The method according to claim 2, characterized in that The preset order information categories include one or more of the following: address category, logistics category, freight category, and product category.

4. The method according to claim 2, characterized in that: The acquiring the order information of the target order based on the preset information extraction rule, the structured information and the communication text fragment includes: Based on a preset information extraction rule, the communication text fragment is structured to obtain first structured order information; Extracting information from the structured information to obtain second structured order information; Based on the first structured order information and the second structured order information, order information of the target order is generated.

5. The method according to claim 4, characterized in that The preset order information category includes: address, the first structured order information includes: structured address information, and the communication text fragment is structured based on the preset information extraction rule to obtain the first structured order information, including: Performing country identification on the communication text segment of the address class to obtain country information contained in the communication text; Obtaining preset address form configuration data matching the country information; Extracting address information at the country level from the communication text fragment of the address class based on the preset address form configuration data to obtain hierarchical address information; The country information and the hierarchical address information are structured to obtain the structured address information.

6. An order generation method, applied to a client, characterized in that: The method comprises: In response to the order information acquisition operation, acquiring object information of the communication object input by the current user; Sending an order information intelligent extraction request carrying the object information to a preset server, the order information intelligent extraction request is used to trigger the preset server to perform the following order information acquisition operation: obtaining the communication record between the current user and the communication object for the target order, the communication record including: structured information and / or unstructured information; extracting communication text fragments of a preset order information category from the unstructured information through a pre-trained order information extraction model; obtaining candidate order information of the target order based on preset information extraction rules, the structured information and the communication text fragments; Obtaining and displaying the candidate order information; The target order is generated based on part or all of the candidate order information.

7. The method according to claim 6, characterized in that The generating the target order based on part or all of the candidate order information includes: Automatically fill in the corresponding order information fields of the target order with the default candidate order information in the candidate order information to generate an order to be confirmed; Displaying order information of the order to be confirmed; In response to a confirmation operation on the order information of the order to be confirmed, the target order is generated based on the order information of the order to be confirmed.

8. The method according to claim 6, characterized in that The generating the target order based on part or all of the candidate order information includes: Filling the corresponding order information field of the target order with the candidate order information selected by the current user to generate an order to be confirmed; Displaying order information of the order to be confirmed; In response to a confirmation operation on the order information of the order to be confirmed, the target order is generated based on the order information of the order to be confirmed.

9. The method according to claim 6, characterized in that The candidate order information includes but is not limited to one or more of the following: delivery address, product information, and freight.

10. An order generation method, applied to a server, characterized in that: The method comprises: In response to an order information intelligent extraction request sent by a preset client, obtaining object information carried in the order information intelligent extraction request; Acquire a communication record between the current user and the communication object to which the object information belongs regarding the target order, wherein the communication record includes: structured information and / or unstructured information; Extracting communication text segments of a preset order information category from the unstructured information through a pre-trained order information extraction model; Based on the preset information extraction rules, the structured information and the communication text fragment, obtaining candidate order information of the target order; The candidate order information is sent to the preset client, so that the preset client displays the candidate order information and generates the target order based on part or all of the candidate order information.

11. A training method for an order information extraction model, characterized in that: The method comprises: Based on the first preset task description, the preset teacher model is triggered to extract information from the historical communication records to obtain a symbolic state described in natural language, wherein the symbolic state includes: a communication record segment, and an order information state matched by the communication record segment; A simulated interactive environment is formed based on the communication record fragment and the order information status, and a feedback result is obtained through the simulated interactive environment, wherein the feedback result includes: a conversion signal corresponding to the communication record fragment; Based on the communication record fragment, the order information status and the conversion signal, the preset teacher model is optimized in combination with the reward model; The preset teacher model obtained based on the optimization is jointly trained with the preset student model, and the trained preset student model is used as the order information extraction model.

12. The method according to claim 11, characterized in that The first preset task description heuristically calls the preset teacher model to extract the communication record fragment of the preset order stage, and the order information status matched by the communication record fragment.

13. An order generation system, characterized in that: The system comprises: a client and a server, wherein: The client is used to obtain the object information of the communication object input by the current user in response to the order information acquisition operation; The client is further used to send an order information intelligent extraction request carrying the object information to the server; The server is used to obtain a communication record between the current user and the communication object regarding a target order, wherein the communication record includes: structured information and / or unstructured information; The server is further configured to extract communication text segments of a preset order information category from the unstructured information using a pre-trained order information extraction model; The server is further configured to obtain candidate order information of the target order based on a preset information extraction rule, the structured information and the communication text fragment; The client is also used to obtain and display the candidate order information; The client is further configured to generate the target order based on part or all of the candidate order information.

14. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method according to any one of claims 1 to 12.

16. A computer program product comprising a computer program / computer executable instructions, characterized in that: When the computer program / computer executable instructions are executed by a processor in an electronic device, the method according to any one of claims 1 to 12 is implemented.

Images