Information processing device, information processing method, and program

Abstract

This makes it easy to identify products that are expressed in a format different from the product names pre-registered in the database. [Solution] The information processing device 1 comprises: an acquisition unit 131 that acquires text information; an element identification unit 132 that identifies product identification elements for identifying the product that is the subject of the inquiry in the text information; a conversion unit 133 that generates a group of converted elements, which are strings obtained by converting the strings constituting each product identification element to match the format registered in the product database; a query generation unit 134 that generates a search query for searching a product database that associates a base product identification element, which is an element that identifies a product, with a product code, based on the group of converted elements; and an output unit 135 that outputs the product code associated with the base product identification element in the product database that corresponds to the group of converted elements included in the search query.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] A system for searching for products that match the product names included in text data is known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In exchanges such as emails, products may be specified in a form different from the actual product name. In such cases, in the prior art, it has been impossible to identify products expressed in a form different from the product names registered in the database in advance, and there has been a problem that it takes time for the person in charge who received the product specification to identify the product.

[0005] Therefore, the present invention has been made in view of these points, and an object thereof is to make it possible to easily identify products expressed in a form different from the product names registered in the database in advance.

Means for Solving the Problems

[0006] An information processing apparatus in a first aspect of the present invention includes: an acquisition unit for acquiring text information; an element identification unit for identifying a group of product identification elements for identifying a product that is the subject of a query in the text information; a conversion unit for generating a group of converted elements, which are strings obtained by converting the strings constituting each of the product identification element groups to match the format registered in the product database; a query generation unit for generating a search query for searching a product database that associates a group of reference product identification elements, which are elements for identifying a product, with a product code, based on the group of converted elements; and an output unit for outputting the product code associated with the corresponding group of reference product identification elements in the product database for the group of converted elements included in the search query.

[0007] The aforementioned set of reference product identification elements includes information indicating the attributes of a product, and the query generation unit may identify the element attributes of the converted set of elements and generate the search query for searching for the attributes of the product corresponding to the element attributes and the converted set of elements, and the product code associated with the reference product identification elements.

[0008] The conversion unit may generate the converted element group by inputting an output instruction to the language model to output the strings that constitute each of the product-specific element groups and the converted element group corresponding to those strings, and causing the language model to output the converted element group corresponding to those strings.

[0009] The conversion unit may determine whether the product identification element group matches the format registered in the product database, and if it determines that they do not match, it may generate the converted element group.

[0010] The text information includes customer identification information, which is information for identifying a customer making an inquiry about a product. If the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database, the output unit may output a product code selected from the multiple product codes based on the customer attributes indicated by the customer identification information.

[0011] The acquisition unit acquires customer attribute information indicating the customer attributes shown in the customer identification information, and the information processing device further has a model storage unit that stores a trained model that has been trained using training customer attribute information indicating the attributes of a customer for training, a plurality of training product codes, and the correct product code among the plurality of training product codes as training data, and the output unit may input the customer attribute information and the plurality of product codes into the trained model when there are a plurality of product codes in the converted element group that correspond to the standard product identification element group in the product database, and select the product code output by the trained model.

[0012] The system further includes a rule information storage unit that stores rule information indicating rules for selecting products to be presented from among the products registered in the product database, and the output unit may output a product code selected from the multiple product codes based on the rule indicated by the rule information when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database.

[0013] The rule information is described in natural language, and the information processing device further has a model storage unit that stores a trained model that has been trained using the rule information, a plurality of training product codes, one or more reference product identification element groups associated with each of the plurality of training product codes, and the correct product code as training data, and the output unit may input the rule information and the plurality of product codes into the trained model and output the product code output by the trained model when there are a plurality of product codes in the converted element group that are associated with the corresponding reference product identification element groups in the product database.

[0014] The system may further include a question generation unit that generates a question to identify which of the multiple product codes to select when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database.

[0015] The system further includes a determination unit that analyzes the meaning of the text contained in the text information and determines whether or not the analyzed text information contains the content of an inquiry about a product. If the determination unit determines that the text information contains the content of an inquiry about a product, the element identification unit may identify the group of product identification elements.

[0016] The system may further include a learning unit that trains on the strings constituting each of the aforementioned product identification element groups and the converted element groups generated by the conversion unit for those strings as training data, and generates a name matching model which is a trained model that outputs the converted element groups when the strings constituting each of the product identification element groups are input.

[0017] In an information processing apparatus according to a second aspect of the present invention, an acquisition unit acquires text information; an element identification unit identifies a group of product identification elements for identifying a product that is the subject of a query in the text information; a conversion unit generates a group of converted elements, which are strings obtained by converting the strings constituting each of the product identification element groups to match the format registered in the product database; a query generation unit generates a search query for searching a product database that associates a group of reference product identification elements, which are elements for identifying a product, with a product code, based on the group of converted elements; a first output unit outputs a plurality of product codes in the product database that are associated with the group of reference product identification elements that correspond to the group of converted elements included in the search query; and a rule information storage unit stores rule information that describes in natural language the rules for selecting a product to be presented from among the products registered in the product database. The system includes a model storage unit that stores a language model that outputs a response to an output instruction when an output instruction is input, and a second output unit that outputs a product code that is output as a result of inputting an output instruction to the language model that includes a plurality of product codes output by the first output unit and rule information stored in the rule information storage unit, and that outputs a product code that conforms to the rule information from among the plurality of product codes.

[0018] The system may further include a learning unit that generates a trained model that has learned using the multiple product codes output by the first output unit, the rule information stored in the rule information storage unit, and the product codes output by the second output unit as training data.

[0019] A third aspect of the present invention provides an information processing method that includes the steps of: obtaining text information, which is performed by a computer; identifying a group of product identification elements in the text information to identify a product that is the subject of a query; generating a group of converted elements, which are strings obtained by converting each of the strings constituting the group of product identification elements to match the format registered in the product database; generating a search query for searching a product database that associates a group of reference product identification elements, which are elements that identify a product, with a product code, based on the group of converted elements; and outputting the product code associated with the corresponding group of reference product identification elements in the product database for the group of converted elements included in the search query.

[0020] In a program according to a fourth aspect of the present invention, the computer is instructed to perform the following steps: acquire text information; identify a group of product identification elements in the text information for identifying a product that is the subject of a query; generate a group of converted elements, which are strings obtained by converting each of the strings constituting the group of product identification elements to match the format registered in the product database; generate a search query for searching a product database that associates a group of reference product identification elements, which are elements for identifying a product, with a product code, based on the group of converted elements; and output the product code associated with the corresponding group of reference product identification elements in the product database for the group of converted elements included in the search query.

[0021] A fifth aspect of the present invention provides an information processing method, which includes: a step of a computer to acquire text information; a step of identifying a group of product identification elements in the text information to identify a product that is the subject of a query; a step of generating a group of converted elements, which are strings obtained by converting each of the strings constituting the group of product identification elements to match the format registered in the product database; a step of generating a search query for searching a product database that associates a group of reference product identification elements, which are elements that identify a product, with a product code, based on the group of converted elements; a first output step of outputting a plurality of product codes associated with the group of reference product identification elements in the product database that correspond to the group of reference product identification elements in the product database, in the search query; and a second output step of referring to a rule information storage unit that stores rule information in natural language that describes the rules for selecting a product to be presented from among the products registered in the product database, and outputting a product code that is output as a result of inputting an output instruction to a language model that outputs a response to the output instruction, which includes the plurality of product codes output in the first output step and the rule information stored in the rule information storage unit, and which outputs an output instruction that specifies that the product code among the plurality of product codes should be output according to the rule information.

[0022] In the program according to the sixth aspect of the present invention, the computer is caused to execute steps of: acquiring text information; identifying a group of product identification elements for identifying a product that is a target of an inquiry in the text information; generating a group of converted elements, which are strings obtained by converting each string constituting the group of product identification elements into a string that matches the format registered in a product database; generating a search query for searching a product database in which a group of reference product identification elements, which are elements for identifying a product, and a product code are associated based on the group of converted elements; a first output step of outputting a plurality of product codes associated with the group of reference product identification elements corresponding to the group of converted elements included in the search query in the product database; and a second output step of referring to a rule information storage unit that stores rule information describing, in natural language, a rule for selecting a product to be presented from the products registered in the product database, inputting an output instruction for outputting a product code according to the rule information among the plurality of product codes output in the first output step and the rule information stored in the rule information storage unit to a language model that outputs a response to the output instruction, and outputting a product code resulting from the input.

Advantages of the Invention

[0023] According to the present invention, it is possible to easily identify a product expressed in a format different from the product name registered in the database in advance.

Brief Description of the Drawings

[0024] [Figure 1] It is a diagram for explaining an overview of the information processing system S according to the embodiment. [Figure 2] It is a diagram for explaining an overview of the information processing system S according to the embodiment. [Figure 3] It is a block diagram showing the configuration of the information processing apparatus 1. [Figure 4] It is a diagram showing an example of the processing of the query generation unit 134. [Figure 5] This figure shows an example of the data structure of a product database. [Figure 6] This figure shows an example of the data structure of profile information stored in the memory unit 12. [Figure 7] This figure shows an example of the data structure of the rule information stored in the memory unit 12. [Figure 8] This is a flowchart showing the processing flow in the information processing device 1. [Modes for carrying out the invention]

[0025] [Overview of Information Processing System S] Figures 1 and 2 are diagrams illustrating the overview of an information processing system S according to an embodiment. The information processing system S is a system for supporting sales activities. More specifically, the information processing system S is a system for identifying the product in question when an inquiry is received from customer U1, and for providing customer U1 with information about the identified product. Unless otherwise specified below, "product" includes services. The information processing system S comprises an information processing device 1, an information terminal 2, and a product database.

[0026] Information processing device 1 is a device for identifying the product in question received from customer U1. More specifically, information processing device 1 acquires text information containing the content of the product inquiry and identifies the product in question within the text information. For example, information processing device 1 acquires text information containing the content of a request for a quote for a product. Information processing device 1 may further generate a response regarding the inquiry for the identified product. Customer U1 is a customer who purchases products from the business to which user U2 belongs.

[0027] Information terminal 2 is a terminal used by user U2. Information terminal 2 is, for example, a smartphone, tablet, or personal computer. User U2 is, for example, a person in charge of responding to inquiries from customer U1 or a person in charge of sales or marketing to customer U1.

[0028] The product database is a database for searching for products. As will be explained in detail later, the product database associates elements for identifying a product with a product code. The product database accepts search queries for products and outputs product codes corresponding to the content of the search query. Figure 1 shows an example where the product database and the information processing device 1 are configured separately, but the product database may also be included within the information processing device 1.

[0029] The processing of the information processing system S will be explained with reference to Figure 2. The information processing device 1 acquires text information. The text information is, for example, an email sent by customer U1. The text information includes a group of product identification elements. The group of product identification elements includes one or more product identification elements. A product identification element is information used to identify the product that is the subject of the inquiry in the text information. For example, a product identification element is information that indicates the product name, seller, manufacturer, type of product, usage period, etc. The information processing device 1 may acquire text information from an external mail server (not shown) or from an information terminal 2.

[0030] Information processing device 1 identifies product identification elements contained in text information. For example, information processing device 1 inputs text information into an extraction model and outputs product identification elements contained in the text information. The extraction model is a pre-trained model that takes text information as input and outputs product identification elements contained in the input text information.

[0031] The information processing device 1 generates a set of converted elements based on the set of product identification elements. The set of converted elements is a string obtained by converting the strings that make up the set of product identification elements to match the format registered in the product database (hereinafter referred to as the "predetermined format"). As an example, the information processing device 1 inputs the set of product identification elements into a name matching language model and outputs the set of converted elements.

[0032] The information processing device 1 generates a search query based on the converted element group. The search query is a query for searching the product database. Specifically, the information processing device 1 identifies the element attributes corresponding to each converted element that makes up the converted element group. Element attributes are attributes possessed by the product identification element. The information processing device 1 generates a search query based on the converted element group and element attributes. For example, if the converted element group is "Msoft" and "OfficeXXX", the information processing device 1 identifies the vendor and product name as element attributes for "Msoft" and "OfficeXXX", respectively, and generates the search query "Vendor:Msoft, Product name:OfficeXXX".

[0033] Information processing device 1 searches the product database based on the search query and retrieves product codes that match the conditions indicated by the search query. Information processing device 1 displays the retrieved product codes on information terminal 2.

[0034] The information processing system S is configured to acquire text information, convert a group of product identification elements extracted from the acquired text information into a predetermined format, and generate a search query to search for product codes associated with the converted group of elements. This makes it possible to identify the target product based on the information contained in the text information, and to easily identify products that are expressed in a different format than the product names registered in the database beforehand.

[0035] [Configuration of Information Processing Device 1] Figure 3 is a block diagram showing the configuration of the information processing device 1. The information processing device 1 includes a communication unit 11, a storage unit 12, and a control unit 13. The control unit 13 includes an acquisition unit 131, an element identification unit 132, a conversion unit 133, a query generation unit 134, an output unit 135, a question generation unit 136, a determination unit 137, and a learning unit 138.

[0036] The communication unit 11 is a communication interface for sending and receiving data with other devices via a network. The storage unit 12 is a storage medium including ROM (Read Only Memory), RAM (Random Access Memory), SSD (Solid State Drive), hard disk drive, etc. The storage unit 12 pre-stores programs to be executed by the control unit 13. The storage unit 12 functions as a model storage unit and a rule information storage unit.

[0037] The memory unit 12 stores the extraction model. The extraction model is a large-scale language model that learns using training text information and a set of product identification elements contained in the training text information as training data.

[0038] The memory unit 12 stores the name matching language model. The name matching language model is a large-scale language model. For example, the name matching language model may have been trained using output instructions containing a set of product identification elements for training, and a set of converted elements obtained by transforming the said set of product identification elements, as training data. The output instructions include content indicating that the input set of product identification elements should be converted into a predetermined format.

[0039] The control unit 13 is a processor, such as a CPU (Central Processing Unit). By executing a program stored in the memory unit 12, the control unit 13 functions as an acquisition unit 131, an element identification unit 132, a conversion unit 133, a query generation unit 134, an output unit 135, a question generation unit 136, a determination unit 137, and a learning unit 138.

[0040] The acquisition unit 131 acquires text information. The acquisition unit 131 may acquire text information from the information terminal 2 or from an external device.

[0041] The element identification unit 132 identifies a group of product identification elements based on the text information acquired by the acquisition unit 131. The element identification unit 132 inputs the text information into an extraction model and outputs the group of product identification elements.

[0042] The conversion unit 133 generates a set of converted elements based on the set of product identification elements. The conversion unit 133 inputs output instructions written in natural language into the name matching language model and generates converted elements by causing the name matching language model to output a set of converted elements corresponding to the product identification elements included in the input instructions. The output instructions include a statement that the strings constituting each of the product identification element groups and the set of converted elements corresponding to those strings should be output.

[0043] The query generation unit 134 generates a search query for searching the product database based on the converted element set. Figure 4 shows an example of the processing of the query generation unit 134. An example of the processing of the query generation unit will be explained with reference to Figure 4. As an example, the storage unit 12 stores a template T configured to generate a search query by applying a product identification element set, and the query generation unit 134 generates a search query by applying the converted element set to template T.

[0044] The query generation unit 134 identifies the element attribute groups of the converted element group. For example, the storage unit 12 stores an element identification model that has been trained to take the converted element group as input and output the element attributes of each converted element that makes up the input converted element group. The query generation unit 134 inputs the converted element group into the element identification model and identifies the element attributes of each converted element that makes up the converted element group. In the example in Figure 4, the query generation unit 134 identifies the element attributes "Vendor," "Product Name," and "Type" for the converted element groups "Msoft," "OfficeXXX," and "Enterprise," respectively.

[0045] The query generation unit 134 generates a search query to search for product codes associated with the identified element attributes and product identification element groups, as well as the element attributes and base product identification element groups corresponding to those elements. The query generation unit 134 applies the identified transformation element groups and element attributes to template T to generate the search query.

[0046] In the product database, a set of standard product identification elements is associated with the product code. The set of standard product identification elements is the set of elements that serve as the basis for identifying the product. For example, the set of standard product identification elements includes information with elemental attributes such as product name, vendor, license type, license period, and version. Figure 5 shows an example of the data structure of the product database. In the product database shown in Figure 5, the product code is associated with the standard product identification elements: product name, vendor, license type, and license period.

[0047] The output unit 135 searches the product database based on the search query and identifies the product code corresponding to the base product identification element group corresponding to the converted element group. The output unit 135 outputs the product code associated with the base product identification element group corresponding to the converted element included in the search query in the product database. The output unit 135 searches the product database based on the search query and obtains the product code corresponding to the converted element group. The output unit 135 outputs the obtained product code to the information terminal 2. Specifically, the output unit 135 may send the obtained product code to the information terminal 2 and display a screen showing the obtained product code on the information terminal 2.

[0048] The information processing device 1 is configured to convert a group of product identification elements extracted from text information into a predetermined format and generate a search query to search for product codes associated with the converted group of elements. This makes it easy to identify products that are expressed in a format different from the product names registered in the database beforehand.

[0049] Note that the number of elements in the product identification element group and the converted element group do not necessarily have to correspond one-to-one. That is, the conversion unit 133 may generate a converted element group that includes the supplemented elements based on the product identification element group. In this case, the name matching language model learns training data that associates the training product identification element group with the converted element group that includes the supplemented product identification elements. For example, the name matching language model may be trained to output "Msoft", "OfficeXXX", and "Enterprise" for the input product identification element group "Msoft" and "OfficeXXX".

[0050] If there is insufficient information in the text data to identify a product code, multiple product candidates may be identified. In such cases, the information processing device 1 may be configured to identify the product based on the attributes of customer U1.

[0051] The text information may include customer identification information so that the information processing device 1 can identify a product based on the attributes of customer U1. Customer identification information is information used to identify customer U1 who is making an inquiry about a product. Examples of customer identification information include customer U1's email address, customer U1's name, or customer U1's user ID. The element identification unit 132 identifies the customer identification information contained in the text information. For example, the extraction model may be configured to output the customer identification information contained in the text information.

[0052] The acquisition unit 131 acquires customer attribute information that indicates the attributes of customer U1 as shown by the customer identification information. For example, the storage unit 12 stores profile information that associates the customer identification information and the customer attribute information. Figure 6 is a diagram showing an example of the data structure of the profile information stored by the storage unit 12. The acquisition unit 131 acquires customer attribute information that corresponds in the profile information to the customer identification information identified by the element identification unit 132. In the profile information shown in Figure 6, the customer attribute information includes the number of employees and the industry of the customer.

[0053] The output unit 135 outputs a selected product code from among the multiple product codes, based on the attributes of customer U1 indicated by the customer identification information, when the converted element has multiple product codes associated with the corresponding standard product identification element group in the product database.

[0054] As an example, the information processing device 1 selects a product code using a trained model that has been trained to output a product code to present to customer U1 based on the attributes of customer U1. The storage unit 12 stores the trained model, which has been trained using training customer attribute information indicating the attributes of customer U1 for training, multiple training product codes, and the correct product code from among the multiple training product codes as training data. When the trained model receives customer attribute information and multiple product codes as input, it outputs a product code.

[0055] The output unit 135, when the converted element group contains multiple product codes associated with the corresponding standard product identification element group in the product database, inputs customer attribute information and multiple product codes into the trained model and displays a screen on the information terminal 2 showing the product codes output by the trained model.

[0056] The output unit 135 is configured to select a product code based on the attributes of customer U1, which has the effect of making it possible to select a product that matches the attributes of customer U1 who made the inquiry about the product.

[0057] In some cases, rules for determining which products to offer to customer U1 are predefined. In such cases, the information processing device 1 may be configured to select a product code based on the defined rules.

[0058] In this case, the storage unit 12 stores rule information that indicates the rules for selecting which products to present from among the products registered in the product database. Figure 7 shows an example of the data structure of the rule information stored in the storage unit 12. The rules in the rule information are written in natural language. As an example, the rule information is set for each business to which customer U1 belongs. The rule information may also be set for each product code.

[0059] The output unit 135 outputs a product code selected from multiple product codes based on the rule indicated by the rule information when the converted element has multiple product codes associated with the corresponding reference product identification element group in the product database. In this case, the trained model has been trained using the rule information, multiple training product codes, one or more reference product identification element groups associated with each of the multiple training product codes, and the correct product code as training data, and outputs a product code with the rule information and multiple product codes as input.

[0060] The output unit 135, when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database, inputs the rule information and the multiple product codes into the trained model and outputs the product codes output by the trained model.

[0061] The information processing device 1 may be configured to use a large-scale language model to select a product code from among multiple product codes that matches the rule indicated by the rule information. In this case, the storage unit 12 stores the selection language model. The selection language model is trained to output a product code, taking as input an output instruction that includes multiple product codes, the rule indicated by the rule information, and a statement to output a product that conforms to the rule from among the multiple product codes. In this case, the selection language model is trained based on training data, which is an output instruction for training that associates a product code, the rule indicated by the rule information, a statement to output a product that conforms to the rule from among multiple product codes, and the correct product code.

[0062] The output unit 135 inputs an output instruction to the selected language model, which includes multiple product codes identified based on the search query, the rule indicated by the rule information, and a statement that products conforming to the rule should be output from among the multiple product codes, and outputs product codes as a response to the output instruction.

[0063] With the information processing device 1 configured in this way, it becomes possible to identify products that conform to defined rules in response to inquiries.

[0064] The information processing device 1 may be configured to determine whether the set of product identification elements extracted from text information matches the format registered in the product database, and if they do not match, to generate a converted set of elements.

[0065] In this case, the conversion unit 133 determines whether the group of product identification elements identified by the element identification unit 132 matches a predetermined format, and generates a converted group of elements if it determines that they do not match. For example, the storage unit 12 stores the allowed formats for each element attribute. For example, the storage unit 12 defines the number of digits and data format for each element attribute. The conversion unit 133 determines whether each product identification element included in the group of product identification elements identified by the element identification unit 132 matches a predetermined format. If the conversion unit 133 determines that the group of product identification elements does not match a predetermined format, it generates a converted group of elements based on the group of product identification elements.

[0066] The information processing device 1 is configured to determine the format of the extracted product identification element group and generate a converted element group if the formats do not match, thereby enabling the identification of product codes with simple processing.

[0067] If there are multiple product codes corresponding to a set of product identification elements, the information processing device 1 may be configured to generate a question for user U2 to identify the product code. To do this, the question generation unit 136 generates a question to identify which of the multiple product codes to select when there are multiple product codes in the converted element set that are associated with the corresponding standard product identification element set in the product database.

[0068] Specifically, the memory unit 12 stores a question generation model. The question generation model is a trained model that takes a set of product-specific elements for each of several products as input and outputs a question asking the user U2 which elements are needed to identify a particular product. The question generation model is a large-scale language model. As an example, the question generation model is trained using a set of product-specific elements associated with each of several product codes used for training, and a question sentence used for training, as training data. For example, if the product name, vendor, and type associated with each of several product codes identified based on a search query are the same, but the license periods differ ("1 year" and "6 months"), the question generation model will generate a question asking, "Is the license period 1 year?"

[0069] The question generation unit 136 refers to the product database and obtains a set of product identification elements corresponding to each of the multiple product codes identified based on the search query. The question generation unit 136 inputs the obtained set of product identification elements corresponding to each of the multiple product codes into the question generation model and outputs a question. The question generation unit 136 displays the generated question on the information terminal 2. The acquisition unit 131 obtains the answer from the information terminal 2 as text information. Subsequently, the information processing device 1 identifies the set of product identification elements and the set of converted elements contained in the text information obtained as the answer to the question, and performs processing to identify the product code based on the already obtained set of converted elements.

[0070] The information processing device 1 is configured to generate a question for identifying the product code, making it possible to identify the product code even when product identification elements are missing in the text information.

[0071] The information processing device 1 may be configured to determine whether text information is an inquiry about a product and to identify the product being inquired about. In this case, the determination unit 137 analyzes the meaning of the text contained in the text information and determines whether the analyzed text information contains the content of an inquiry about a product. For example, the determination unit 137 determines that the text contains the content of an inquiry about a product if it contains a string that means an inquiry, such as "Please give me a quote." If the determination unit 137 determines that the text contains the content of an inquiry about a product, the element identification unit 132 identifies a group of product identification elements.

[0072] Alternatively, the memory unit 12 may store a pre-trained model, which is a model trained to determine whether or not a text information is an inquiry about a product, and the system may determine whether or not a text information is an inquiry about a product by inputting it into the determination model.

[0073] By configuring the information processing device 1 to determine whether or not the text information is an inquiry about a product, it is possible to avoid the process of identifying a product code for text information that is not intended as an inquiry.

[0074] The information processing device 1 may be configured to generate a set of converted elements using a trained model that has learned from the set of converted elements output by the name matching language model.

[0075] The learning unit 138 generates a name matching model by training it with the product identification element group and the converted element group output by the name matching language model, which takes the product identification element group as input, as training data. In other words, it generates a name matching model that is trained with the strings that make up each product identification element group and the converted element group generated by the conversion unit 133 for those strings as training data. The name matching model is a trained model that outputs the converted element group when the strings that make up each product identification element group are input. The learning unit 138 generates the name matching model by training a machine learning model that is not intended for natural language processing with training data. The conversion unit 133 inputs the product identification element group into the name matching model generated by the learning unit 138 and outputs the converted element group.

[0076] The information processing device 1 is configured to output the transformed element set using a trained model, which is a non-verbal model that has learned from the transformed element set output by the name matching language model. This makes it possible to generate the transformed element set with simple processing.

[0077] The information processing device 1 may be configured to select a product code using a trained model that has learned the product codes output by the selection language model. In this case, the learning unit 138 generates a selection model that has learned training data associating multiple product codes identified based on a search query, rule information, and product codes output by the selection language model based on an output instruction that includes the multiple product codes and rule information. The selection model outputs a product code when it receives multiple product codes and rule information as input. The learning unit 138 generates a selection model by training a machine learning model not intended for natural language processing with training data. The output unit 135 inputs multiple product codes identified based on a search query and rule information into the selection model and causes it to output a product code.

[0078] By configuring the information processing device 1 to select a product code using a selection model, which is a non-verbal model that has learned the information output by the selection language model, it becomes possible to select a product code to present to customer U1 with a simple process.

[0079] [Processing flow in information processing device 1] Figure 8 is a flowchart showing the processing flow in the information processing device 1. The flowchart shown in Figure 8 starts from the point when text information becomes available for acquisition. The acquisition unit 131 acquires the text information (S01). The element identification unit 132 identifies a group of product identification elements based on the text information (S02). The conversion unit 133 converts the group of product identification elements into a group of converted elements (S03).

[0080] The query generation unit 134 generates a search query based on the converted element set (S04). The output unit 135 searches the product database using the generated search query and identifies the product code (S05). The output unit 135 outputs the product code (S06). The information processing device 1 then terminates processing.

[0081] [Effects of Information Processing Device 1] The information processing device 1 is configured to acquire text information, convert a group of product identification elements extracted from the acquired text information into a predetermined format, and generate a search query to search for product codes associated with the converted group of elements. This makes it easy to identify products that are expressed in a format different from the product names registered in the database beforehand.

[0082] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of its gist. For example, all or part of the apparatus can be configured by functionally or physically distributing and integrating in any unit. Furthermore, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of the present invention. The effects of the new embodiments resulting from the combinations are combined with the effects of the original embodiments. [Explanation of symbols]

[0083] 1. Information Processing Device 2 Information terminals 11 Communications Department 12 Storage section 13 Control Unit 131 Acquisition Department 132 Element Identification Unit 133 Conversion section 134 Query Generation Unit 135 Output section 136 Question generation part 137 Judgment section 138 Learning Department

Claims

1. A unit that acquires text information, An element identification unit identifies a group of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, A conversion unit generates a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database. A query generation unit generates a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the aforementioned converted set of elements. An output unit that outputs product codes associated with the corresponding standard product identification elements in the product database for the converted elements included in the search query, An information processing device having

2. The aforementioned set of criteria product identification elements includes information indicating the attributes of the product, The query generation unit identifies the element attributes of the converted element group and generates a search query for searching for the element attributes, the attributes of the product corresponding to the converted element group, and the product code associated with the reference product identification element group. The information processing apparatus according to claim 1.

3. The conversion unit inputs an output instruction to the language model to output the strings that constitute each of the product-specific element groups and the converted element groups corresponding to those strings, and generates the converted element groups by outputting the converted element groups corresponding to those strings. The information processing apparatus according to claim 1.

4. The conversion unit determines whether the product identification element group matches the format registered in the product database, and if it determines that they do not match, it generates the converted element group. The information processing apparatus according to claim 1.

5. The aforementioned text information includes customer identification information, which is information used to identify customers who make inquiries about products. The output unit, when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database, outputs a product code selected from the multiple product codes based on the customer attributes indicated by the customer identification information. The information processing apparatus according to claim 1.

6. The acquisition unit acquires customer attribute information that indicates the attributes of the customer indicated by the customer identification information, The information processing device further includes a model storage unit that stores a trained model, using training customer attribute information indicating the attributes of a customer for training, a plurality of training product codes, and the correct product code among the plurality of training product codes as training data. The output unit, when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database, inputs the customer attribute information and the multiple product codes into the trained model and selects the product code output by the trained model. The information processing apparatus according to claim 5.

7. It further includes a rule information storage unit that stores rule information indicating the rules for selecting which products to present from among the products registered in the product database, The output unit outputs a product code selected from the multiple product codes based on the rule information when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database. The information processing apparatus according to claim 1.

8. The aforementioned rule information is written in natural language, The information processing device further includes a model storage unit that stores a trained model, which is trained using the rule information, a plurality of training product codes, one or more reference product identification element groups associated with each of the plurality of training product codes, and the correct product code as training data. The output unit, when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database, inputs the rule information and the multiple product codes to the trained model and outputs the product codes output by the trained model. The information processing apparatus according to claim 7.

9. The system further includes a question generation unit that generates a question to identify which of the multiple product codes to select when the converted element group has multiple product codes associated with the corresponding standard product identification element group in the product database. The information processing apparatus according to claim 1.

10. The system further includes a determination unit that analyzes the meaning of the text contained in the aforementioned text information and determines whether or not the analyzed text information contains the content of an inquiry about a product. If the determination unit determines that the inquiry contains information about a product, the element identification unit identifies the product identification element group. The information processing apparatus according to claim 1.

11. The system further includes a learning unit that trains on the strings constituting each of the aforementioned product identification element groups and the converted element groups generated by the conversion unit for those strings as training data, and generates a name matching model which is a trained model that outputs the converted element groups when the strings constituting each of the product identification element groups are input. The information processing apparatus according to claim 1.

12. A unit that acquires text information, An element identification unit identifies a group of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, A conversion unit generates a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database. A query generation unit generates a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the aforementioned converted set of elements. A first output unit outputs a plurality of product codes associated with the corresponding standard product identification element group in the product database to the converted element group included in the search query, A rule information storage unit stores rule information that describes in natural language the rules for selecting which products to present from the products registered in the product database, A model storage unit that stores a language model that outputs a response to an output instruction when an output instruction is input, A second output unit outputs a product code that includes a plurality of product codes output by the first output unit and rule information stored in the rule information storage unit, and outputs a product code that is output as a result of inputting an output instruction to the language model to output a product code that conforms to the rule information from among the plurality of product codes; An information processing device having

13. The system further includes a learning unit that generates a trained model that has learned using the multiple product codes output by the first output unit, the rule information stored in the rule information storage unit, and the product codes output by the second output unit as training data. The information processing apparatus according to claim 12.

14. A computer executes Steps to obtain text information, The steps include identifying a set of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, The steps include generating a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database, The steps include generating a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the converted set of elements, The steps include outputting product codes associated with the corresponding standard product identification elements in the product database for the converted elements included in the search query, An information processing method having

15. On the computer, Steps to obtain text information, The steps include identifying a set of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, The steps include generating a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database, The steps include generating a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the converted set of elements, The steps include outputting product codes associated with the corresponding standard product identification elements in the product database for the converted elements included in the search query, A program to execute.

16. A computer executes Steps to obtain text information, The steps include identifying a set of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, The steps include generating a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database, The steps include generating a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the converted set of elements, A first output step which outputs a plurality of product codes associated with the corresponding base product identification element group in the product database to the converted element group included in the search query, A second output step involves referring to a rule information storage unit that stores rule information, which describes in natural language the rules for selecting which products to present from among the products registered in the product database, and, when an output instruction is input, inputting an output instruction that includes the multiple product codes output in the first output step and the rule information stored in the rule information storage unit, and that outputs a product code from among the multiple product codes that conforms to the said rule information, the second output step outputs a product code that is output as a result of inputting to a language model that outputs a response to the said output instruction. An information processing method having

17. On the computer, Steps to obtain text information, The steps include identifying a set of product identification elements in the aforementioned text information to identify the product that is the subject of the inquiry, The steps include generating a set of converted elements, which are strings obtained by converting the strings constituting each of the aforementioned product-specific elements to match the format registered in the product database, The steps include generating a search query for searching a product database that associates a set of reference product identification elements, which are elements that identify a product, with a product code, based on the converted set of elements, A first output step which outputs a plurality of product codes associated with the corresponding base product identification element group in the product database to the converted element group included in the search query, A second output step, upon input of an output instruction, refers to a rule information storage unit that stores rule information in natural language describing the rules for selecting which products to present from among the products registered in the product database, and outputs a product code that is the result of inputting into a language model that outputs a response to the output instruction, which includes the multiple product codes output by the first output step and the rule information stored in the rule information storage unit, and which indicates that it will output a product code from among the multiple product codes that conforms to the said rule information. A program to execute.

Images