Information processing device

Abstract

We will improve the technology for searching for products that meet specific specifications based on natural language questions. [Solution] The information processing device 20 registers one or more attributes and attribute values ​​for multiple vehicles in a database. When a question written in natural language is input, the device uses machine learning to identify the search conditions indicated in the question in relation to the attributes, generates a query to select vehicles from the database that match the identified search conditions, and outputs the search results from the database based on the query.

Description

Technical Field

[0001] This disclosure relates to an information processing apparatus.

Background Art

[0002] Patent Document 1 discloses a technique for selecting a document that may contain an answer to a user's question input to a search model and outputting an answer to the user's question.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] There has been room for improvement in the technique of searching for products that meet specific specifications from natural language question sentences.

[0005] <( An object of this disclosure is to improve the technique of searching for products that meet specific specifications from natural language question sentences.

[0006] An information processing apparatus according to an embodiment of this disclosure registers, in a database, one or more attributes in combination with attribute values that are values of the attributes for a plurality of vehicles, when a question sentence in natural language is input, uses machine learning to identify search conditions indicated in the question sentence in relation to the attributes, generates a query for selecting vehicles that meet the identified search conditions from the database, and includes a control unit that performs control to output a search result of the database by the query.

Advantages of the Invention

[0007] According to this disclosure, the technology for searching for products that meet specific specifications from natural language questions will be improved. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the configuration of a system according to an embodiment of this disclosure. [Figure 2] This is a flowchart showing the operation of the information processing device according to the embodiment of this disclosure. [Modes for carrying out the invention]

[0009] The embodiments of this disclosure will be described below with reference to the figures.

[0010] In each figure, identical or corresponding parts are denoted by the same reference numerals. In the description of this embodiment, the description of identical or corresponding parts will be omitted or simplified as appropriate.

[0011] Referring to Figure 1, the configuration of the system 10 according to this embodiment will be described.

[0012] The system 10 comprises at least one information processing device 20 and at least one external device 30.

[0013] The information processing device 20 can communicate with an external device 30 via the network 40.

[0014] The information processing device 20 is a computer, such as a server, belonging to a cloud computing system or other computing system, installed in a facility such as a data center.

[0015] External device 30 is a terminal device installed at a dealership or other location that sells vehicles. External device 30 may be, for example, a mobile phone, smartphone, tablet, or PC. "PC" is an abbreviation for personal computer.

[0016] Network 40 includes the Internet, at least one WAN, at least one MAN, or any combination thereof. "WAN" is an abbreviation for wide area network. "MAN" is an abbreviation for metropolitan area network.

[0017] The outline of this embodiment will be described with reference to Figure 1.

[0018] For example, conversational systems using information retrieval technologies such as RAG are used to answer questions about the specifications of specific vehicles by loading vehicle specification sheets. "RAG" is an abbreviation for Retrieval-Augmented Generation. Using such conversational systems, for example, during a sales negotiation, it is possible to search for vehicle models that match the conditions desired by the customer. However, when a staff member, as a user, inquires about the conditions desired by the customer using natural language questions, the ambiguity of the wording of the question may cause the search condition expression as a query to not be generated correctly. For example, when a customer wants to use specific numerical values ​​for specifications such as "vehicle height," "tire outer width," and "tread width" as search conditions, the units of those numbers may not be spoken in the actual conversation. As a result, the search conditions could not be correctly identified.

[0019] In this embodiment, the information processing apparatus 20 registers, in a database, one or more attributes in combination with attribute values, which are values of the attributes, for a plurality of vehicles. When a question sentence in natural language is input, the information processing apparatus 20 uses machine learning to identify search conditions indicated in the question sentence in relation to the attributes. Any machine learning can be used, for example, any natural language processing technology selected from large language models, NLP, intention recognition models, NER, and semantic search, etc., can be used alone or in combination. "NLP" is an abbreviation for Natural Language Processing. "NER" is an abbreviation for Named Entity Recognition. The information processing apparatus 20 generates a query for selecting vehicles corresponding to the identified search conditions from the database. The information processing apparatus 20 outputs the search result of the database by the query.

[0020] According to this embodiment, when a question sentence in natural language is input, the information processing apparatus 20 uses machine learning to identify search conditions indicated in the question sentence in relation to the attributes. For example, when the attribute identified from the question sentence is "vehicle height" and the reference value (the value used as a reference for searching for an attribute value), which is the search condition identified in relation to "vehicle height", is "1.55", the reference value is corrected to "1.55 m". Therefore, even when the unit of the numerical value included in the question sentence is not uttered and the question sentence includes ambiguous expressions, the search conditions can be easily and correctly identified. Therefore, the technology for searching for products corresponding to specific specifications from a question sentence in natural language is improved.

[0021] Referring to FIG. 1, the configuration of the information processing apparatus 20 according to this embodiment will be described.

[0022] The information processing apparatus 20 includes a control unit 21, a storage unit 22, and a communication unit 23. The control unit 21 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof, and executes predetermined processing.

[0023] The storage unit 22 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of these. In the storage unit 22, data used for the operation of the information processing apparatus 20 and data obtained by the operation of the information processing apparatus 20 are stored. In the present embodiment, a database DB is stored in the storage unit 22. The database DB will be described later.

[0024] The communication unit 23 includes at least one communication interface. The communication unit 23 receives data used for the operation of the information processing apparatus 20 and transmits data obtained by the operation of the information processing apparatus 20. In the present embodiment, the communication unit 23 communicates with an external device 30.

[0025] The functions of the information processing apparatus 20 are realized by executing the program according to the present embodiment on a processor as the control unit 21. That is, the functions of the information processing apparatus 20 are realized by software. The program causes a computer to execute the operations of the information processing apparatus 20, thereby causing the computer to function as the information processing apparatus 20. That is, the computer functions as the information processing apparatus 20 by executing the operations of the information processing apparatus 20 according to the program.

[0026] Referring to FIG. 2, the operation of the information processing apparatus 20 according to the present embodiment will be described. This operation corresponds to the information processing method according to the present embodiment. That is, the information processing method according to the present embodiment includes the steps S1 to S7 shown in FIG. 2. Hereinafter, each step of the flowchart is specified by S and a number.

[0027] In S1, the control unit 21 of the information processing device 20 registers one or more attributes for multiple vehicles in the database DB, combined with attribute values. In this embodiment, one or more attributes include vehicle specifications. The control unit 21 can register vehicle specifications for each vehicle by reading the specification tables of multiple vehicles or by referring to publicly available official information. For example, the control unit 21 registers attributes such as "vehicle height," "tire outer width," and "tread width" in the database DB along with their numerical attribute values. The control unit 21 may also calculate attributes not found in the specification tables and their attribute values ​​and register them in the database DB. For example, "tire outer width" is one of the important constraints for multi-story parking garages. However, vehicle catalogs generally do not list "tire outer width." "Tire outer width" is the width of the outer edges of the left and right tires of the vehicle. In this case, the control unit 21 may calculate "tire outer width" based on "tread width" and "tire width (front / rear)" found in the specification tables. "Tire outer width" can be calculated by adding the tread width to the maximum value of the front / rear tire width.

[0028] In S2, the control unit 21 of the information processing device 20 receives a question in natural language. In this embodiment, the question is input to the external device 30. The question may be input as voice or as text. For example, the question may be voice recordings of a business negotiation via a microphone or the like, which serves as the input unit of the external device 30, or text written by a staff member during a business negotiation via a keyboard or touchscreen or the like, which serves as the input unit of the external device 30. The external device 30 transmits the input question to the information processing device 20. The control unit 21 of the information processing device 20 receives the question transmitted from the external device 30 via the communication unit 23. The control unit 21 analyzes the received question. Any method can be used to analyze the question, such as morphological analysis, syntactic analysis, or semantic analysis.

[0029] In S3, the control unit 21 of the information processing device 20 uses machine learning to identify the search conditions indicated in the question. Specifically, the control unit 21 extracts one or more keywords to define the vehicle specifications as search conditions. The control unit 21 extracts one or more keywords, which include vehicle attributes, reference values ​​associated with those attributes, and comparison conditions with the reference values. Examples of keywords extracted as "vehicle attributes" include "vehicle height," "tire outer width," and "tread width." Keywords extracted as "reference values" are arbitrary numerical values. Examples of keywords extracted as "comparison conditions" include "equal to," "not equal to," "greater than or equal to," "less than or equal to," and "less than."

[0030] In S4, the control unit 21 of the information processing device 20 adjusts the extracted reference value based on the attribute associated with the reference value. Specifically, the control unit 21 determines whether the extracted reference value has a unit, and if it determines that it does not, it estimates the unit of the reference value based on the attribute and the number of digits of the reference value. For example, if the attribute is "vehicle height", the control unit 21 estimates the unit to be [mm] if the reference value has 4 digits, and to be [m] if the reference value has 1 digit. Here, the number of digits is the number of digits after truncating the decimal part. As an example, suppose in S2 the question "Tell me about vehicles with a vehicle height of 1.55 or less" is entered. Here, "example" is not to limit the present disclosure, but to help understand this embodiment. In this case, in S3, the control unit 21 of the information processing device 20 extracts the keywords "vehicle height", "1.55", and "less than or equal to" as search conditions, as vehicle attributes, reference values ​​associated with attributes, and comparison conditions for the reference value. Here, "1.55" is extracted as a reference value associated with the attribute "vehicle height". In this case, the control unit 21 determines that the number of digits in the reference value "1.55" is 1 by truncating the decimal part, and estimates that the unit of the reference value "1.55" is "m". Alternatively, suppose in S2 the question "Tell me about vehicles with a vehicle height of 1550 or less" is entered. In this case, in S3, the control unit 21 of the information processing device 20 extracts the keywords "vehicle height", "1550", and "less than or equal to" as search conditions, as vehicle attributes, reference values ​​associated with attributes, and comparison conditions for the reference values. Here, "1550" is extracted as a reference value associated with the attribute "vehicle height". In this case, the control unit 21 determines that the number of digits in the reference value is 4, and estimates that the unit of the reference value "1550" is "mm".

[0031] As one modification of this embodiment, the control unit 21 of the information processing device 20 may convert the unit of the reference value to a unit registered in the database DB. Specifically, suppose the attribute value for "vehicle height" is registered in meters in the database DB, and the unit of the reference value estimated in S4 is "mm". In this case, the control unit 21 may convert the reference value to meters. For example, the control unit 21 may convert the reference value "1550 mm" to "1.55 m".

[0032] In S5, the control unit 21 of the information processing device 20 generates a query to select vehicles from the database DB that match the specified search conditions. Specifically, the control unit 21 generates an SQL query to the database DB based on the search conditions specified in S3 and the reference value adjusted in S4. "SQL" is an abbreviation for Structured Query Language. For example, suppose in S3 the keywords "vehicle height" (vehicle attribute), "1.55" (reference value), and "less than or equal to" (comparison condition with reference value) are extracted as search conditions, and in S4 the reference value "1.55" is adjusted to "1.55m". In this case, the control unit 21 generates a query that means "select vehicles from the database DB whose vehicle height is 1.55m or less".

[0033] In S6, the control unit 21 of the information processing device 20 searches the database DB using the query generated in S5. Specifically, the control unit 21 executes the query generated in S5 against the database DB and obtains vehicle data D1 that matches the query. Data D1 contains information indicating the make and model of one or more vehicles that match the search criteria. Data D1 may also contain information indicating the year of manufacture, price, mileage, etc., for each vehicle.

[0034] In S7, the control unit 21 of the information processing device 20 outputs the search results from the database DB based on the query. Specifically, the control unit 21 controls the display of the data D1 obtained in S6 on the external device 30. The data D1 may be displayed in any procedure, but for example, it may be displayed in the following procedure: The control unit 21 obtains a set of data that matches the query from the database DB as data D1. The control unit 21 transmits data D1 to the external device 30 via the communication unit 23. The external device 30 receives the data D1 transmitted from the information processing device 20 and displays the search results indicated by data D1 on a display or the like, which acts as an output unit. Alternatively, the control unit 21 may input the query received in S2 and the data D1 obtained in S5 into a large-scale language model, obtain a natural language response, and output this response.

[0035] According to this embodiment, when querying for products that meet specific specifications using a natural language query, even if the unit of the numerical reference value included in the query is unknown, the correct search condition expression can be automatically generated. Therefore, the technology for searching for products that meet specific specifications from a natural language query is improved.

[0036] In this embodiment, the external device 30 is installed outside the information processing device 20, and the information processing device 20 and the external device 30 perform processing while sending and receiving various data via the network 40. However, some or all of the functions of the information processing device 20 may be incorporated into the external device 30.

[0037] In this embodiment, if the control unit 21 of the information processing device 20 determines that the unit of the reference value is clear when it specifies the search condition in S3, it may omit the process in S4. Alternatively, even if the unit of the reference value is clear when it specifies the search condition in S3, the control unit 21 may, in S4, refer to the database DB and check the numerical value and / or unit registered for the attribute associated with the reference value to correct the numerical value or unit of the reference value included in the question. For example, if a user mistakenly says "1.55 millimeters" instead of "1.55 meters" as the reference value for the attribute "vehicle height," the control unit 21 will refer to the database DB and check the numerical value registered for "vehicle height" to detect the unnatural change in unit. This configuration allows for the correction of errors in numerical units due to slips of the tongue or input errors. The control unit 21 may also perform control to display a prompt to ask the user for confirmation when an error in numerical units is detected. For example, the control unit 21 may perform control to display the message "Is it 1.55 meters, not 1.55 millimeters?" on the external device 30.

[0038] This disclosure is not limited to the embodiments described above. For example, multiple blocks described in the block diagram may be combined, or a single block may be divided. Instead of executing multiple steps described in the flowchart in chronological order as described, they may be executed in parallel or in a different order, depending on the processing capacity of the device performing each step, or as necessary. Other modifications are possible without departing from the spirit of this disclosure. [Explanation of Symbols]

[0039] 10 Systems 20 Information Processing Devices 21 Control Unit 22 Memory section 23 Communications Department 30 External devices 40 Networks

Claims

1. For multiple vehicles, one or more attributes are registered in the database in combination with attribute values, which are the values ​​of those attributes. When a question in natural language is input, machine learning is used to identify the search conditions indicated in the question in relation to the attribute. A query is generated to select vehicles from the database that match the specified search criteria. An information processing device comprising a control unit that controls the output of search results from the database based on the aforementioned query.

2. The one or more attributes mentioned above include the specifications of the automobile. The information processing apparatus according to claim 1, wherein the control unit extracts one or more keywords for defining the specifications as the search conditions.

3. The control unit extracts one or more keywords, which include vehicle attributes, reference values ​​associated with those attributes, and comparison conditions with the reference values. The information processing apparatus according to claim 2, wherein the control unit adjusts the extracted reference value based on the attribute associated with the extracted reference value.

4. The information processing apparatus according to claim 3, wherein the control unit determines whether the extracted reference value has units, and if it determines that there are no units, it estimates the units of the reference value based on the attributes and the number of digits of the reference value.

5. The information processing apparatus according to claim 3 or 4, wherein the control unit converts the unit of the reference value to a unit registered in the database.

Images