Information processing device, information processing method, and information processing program

Abstract

To provide an information processing device capable of accurately presenting a related search query, an information processing method, and an information processing program.SOLUTION: An information processing device comprises: an acquisition unit which acquires a search query input to a user terminal by a user from the user terminal; a conversion unit which converts language representation of the acquired search query to vector representation; an identification unit which identifies vector representation of a related search query similar to vector representation on the basis of the vector representation converted by the conversion unit; an inverse transformation unit which inversely transforms vector representation of the related search query identified by the identification unit to language representation; and a proposal unit which displays search queries inversely transformed by the inverse transformation unit on the user terminal in descending order of similarities.SELECTED DRAWING: Figure 3

Description

Technical Field

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

Background Art

[0002] In an information search service using a search engine, it is considered that the possibility of providing information along with the user's search intention increases by displaying search queries related to the search query.

[0003] For example, Patent Document 1 below discloses a search query extraction means for extracting, as a search query related to a keyword, a search query including the keyword used for the search and one or more keywords other than this keyword from the search log of the search engine, and a keyword search intention candidate specifying means for specifying, as a candidate for the search intention of the keyword, a word obtained by deleting the keyword from the extracted search query. A search intention dictionary creation apparatus characterized by comprising is disclosed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In the technique of presenting related search queries, taking the technique described in Patent Document 1 as an example, it has been a problem to present a wide range of search queries along with the user's search intention. In view of the above problems, an object of the present disclosure is to provide an information processing apparatus, an information processing method, and an information processing program capable of accurately presenting related search queries.

Means for Solving the Problems

[0006] To solve the above-mentioned problems and achieve the objectives, the information processing device according to this disclosure includes: an acquisition unit that acquires a search query entered by the user into the user terminal from the user terminal; a conversion unit that converts the linguistic expression of the acquired search query into a vector expression; an identification unit that identifies vector expressions of related search queries similar to the vector expression based on the vector expression converted by the conversion unit; an inverse conversion unit that converts the vector expressions of the related search queries identified by the identification unit back into a linguistic expression; and a suggestion unit that displays the search queries converted back by the inverse conversion unit on the user terminal in order of decreasing similarity. [Effects of the Invention]

[0007] According to this disclosure, it is possible to provide an information processing device, an information processing method, and an information processing program that can accurately present relevant search queries. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 shows an example of information processing according to the embodiment. [Figure 2] Figure 2 shows an example of the configuration of an information processing system according to the embodiment. [Figure 3] Figure 3 shows an example of the configuration of an information processing device according to the embodiment. [Figure 4] Figure 4 shows an example of information stored in the search query storage unit of the information processing apparatus according to the embodiment. [Figure 5] Figure 5 shows an example of information stored in the related search query storage unit of the information processing device according to the embodiment. [Figure 6] Figure 6 shows an example of information stored in the model storage unit of the information processing device according to the embodiment. [Figure 7] Figure 7 is a flowchart showing the flow of the learning method for the vector search model according to the embodiment. [Figure 8] Figure 8 shows an example of the configuration of a user terminal according to the embodiment. [Figure 9]Figure 9 is a flowchart showing an example of information processing according to the embodiment. [Figure 10] Figure 10 is a hardware configuration diagram showing an example of a computer that implements the functions of an information processing device. [Modes for carrying out the invention]

[0009] The following describes in detail, with reference to the drawings, the embodiments for implementing the information processing device, information processing method, and information processing program relating to this disclosure (hereinafter referred to as "embodiments"). However, these embodiments do not limit the information processing device, information processing method, and information processing program relating to this disclosure.

[0010] (Embodiment) [1. Information processing according to the embodiment] [1-1. An example of information processing according to the embodiment] First, an example of information processing according to the embodiment will be described using Figure 1. Figure 1 is a diagram showing an example of information processing according to the embodiment. In Figure 1, an example is shown in which the information processing according to the embodiment is executed by the information processing device 100 and the user terminal 200. Hereinafter, the information processing according to the embodiment will be described step by step.

[0011] First, the information processing device 100 obtains the search query entered by the user from the user terminal 200 (step S1). Here, a search query means an inquiry made by the user to a search engine. For example, a search query can be expressed as a combination of words or sentences in a given language. To give an example, the information processing device 100 obtains the combination of words "anime A", "character", and "list" as a search query entered by the user from the user terminal 200. Thus, a search query may be a combination of multiple words, a single word, or a combination of multiple sentences. Once the information processing device 100 obtains the search query entered by the user from the user terminal 200, it stores the obtained search query in the search query storage unit 121, which will be described later.

[0012] Next, the information processing device 100 converts the acquired linguistic expression of the search query into a vector expression (step S2). For example, the information processing device 100 converts the linguistic expression of the combination of words "anime A", "character", and "list", which is the search query acquired in step S1, into a vector expression. For example, the information processing device 100 may perform processing such as cleaning, morphological analysis, normalization (conversion to standard form), and stop word removal as needed before converting to a vector expression. Note that the conversion of the linguistic expression of the search query into a vector expression may be achieved by using Word2Vec, Retrofitting, Glove, FastText, BERT, etc.

[0013] Next, the information processing device 100 identifies vector representations of related search queries with high similarity based on their vector representations (step S3). For example, the information processing device 100 may calculate the similarity by calculating the cosine similarity of the vector representations of related search queries stored in a database of related search queries (for example, the related search query storage unit 122, which will be described later). Note that the similarity between distributed representations may be calculated based on any index that can be applied as a distance measure between vectors, not limited to cosine similarity. For example, the similarity may be calculated using the Euclidean distance between vector representations, the distance in a non-Euclidean space such as hyperbolic space, the Manhattan distance, the Mahalanobis distance, or a predetermined distance function. Then, based on the calculated similarity, the information processing device 100 may identify, for example, vector representations of related search queries with a similarity of or greater than a predetermined threshold as vector representations of related search queries with high similarity.

[0014] Next, the information processing apparatus 100 converts the specified vector representation into a search query (step S4). For example, the information processing apparatus 100 may convert the vector representation of the relevant search query with a high similarity specified in step S3 into the language expression of the relevant search query based on the correspondence between the language expression of the relevant search query stored in a database (for example, the relevant search query storage unit 122) and the vector representation.

[0015] Next, the information processing apparatus 100 causes the relevant search query to be displayed on the user terminal 200 (step S5). For example, the information processing apparatus 100 may cause the relevant search query converted into the language expression in step S4 to be displayed on the user terminal 200 in descending order of similarity.

[0016] According to this, relevant search queries related to the search query input by the user can be widely extracted based on the similarity of the vector representations, and presented to the user in descending order of similarity. Therefore, it is possible to provide an information processing apparatus that can accurately present relevant search queries.

[0017] [1-2. Another example of information processing according to the embodiment] The information processing apparatus 100 uses a model that has learned the relationship between the language expression and the vector representation of the search query through contrastive learning, with search queries having the same session as positive examples, to convert the obtained language expression of the search query into a vector representation.

[0018] This information processing will be described step by step. First, the information processing apparatus 100 executes the same process as step S1 shown in FIG. 1. Since step S1 is the same as the above-described process, the description is omitted.

[0019] Next, the information processing device 100 converts the linguistic representation of the acquired search query into a vector representation using a model that has learned the relationship between the linguistic representation and the vector representation of a search query through comparative learning, using search queries from the same session as positive examples (Step S2-1). Here, a session refers to the period during which a user has performed web actions consecutively, and if the interval between the user's web actions is within a predetermined period (for example, within 30 minutes), it may be considered the same session. In other words, the information processing device 100 converts the linguistic representation of the search query acquired in Step S1 into a vector representation using a model that has been trained so that the vector representations of search queries in the same session are similar, using search queries that the user has entered consecutively within a predetermined period as positive examples.

[0020] In this case, the learning model can be implemented, for example, by applying SimCSE to BERT and training it. SimCSE is a technique that improves the performance of STS (Semantic Textual Similarity) by performing comparative learning on PLMs (Pre-Trained Language Models) like BERT. Comparative learning is a type of self-supervised learning that can learn from a large amount of data directly by using a mechanism that compares data without labeling.

[0021] Next, the information processing device 100 uses a model that has learned the relationship between search queries and vector representations through comparative learning, with search queries from the same session as positive examples, to convert related search queries into vector representations. The device then refers to the database of related search queries and identifies vector representations of related search queries that are similar to the converted vector representations (Step S3-1). In this case, the vector representations of related search queries stored in the database of related search queries (for example, the related search query storage unit 122) may use the same model that was used in Step S2-1 to convert the linguistic representation of the search query into a vector representation, which is a model that has learned the relationship between the linguistic representation and vector representation of a search query through comparative learning, with search queries from the same session as positive examples.

[0022] Next, the information processing device 100 performs the same process as steps S4 to S5 shown in Figure 1. Steps S4 to S5 are the same as the process described above, so their explanation will be omitted.

[0023] According to this, in extracting related search queries, it is possible to extract highly similar search queries that were entered consecutively in the same session, suggesting similar search intent from the user. Therefore, it is possible to provide an information processing device that can accurately present related search queries.

[0024] [2. Configuration of the Information Processing System] Next, the configuration of the information processing system according to the embodiment will be described using Figure 2. Figure 2 is a diagram showing an example of the configuration of the information processing system according to the embodiment. As shown in Figure 2, the information processing system 1 according to the embodiment has an information processing device 100, a user terminal 200, and a network N. These configurations will be briefly described in order below.

[0025] The information processing device 100 may be, for example, a personal computer (PC), a workstation (WS), or a computer equipped with server functions. The information processing device 100 processes information transmitted from a user terminal 200 or the like via the network N.

[0026] The user terminal 200 is an information processing device used by the user. The user terminal 200 may be, for example, a smartphone, tablet, desktop PC, notebook PC, mobile phone, or PDA (Personal Digital Assistant). In the example shown in Figure 1, the user terminal 200 is a smartphone.

[0027] Network N connects the information processing device 100 and the user terminal 200 so that they can communicate with each other via wired or wireless connection. If Network N is wired, it may be implemented using Ethernet (registered trademark) as defined in IEEE 802.3. If Network N is wireless, it may be implemented using a wireless LAN (Local Area Network) as defined in IEEE 802.11.

[0028] [3. Configuration of the Information Processing Device] Next, the configuration of the information processing device 100 will be explained using Figure 3.

[0029] Figure 3 shows an example of the configuration of an information processing device according to the embodiment. As shown in Figure 3, the information processing device 100 includes a communication unit 110, a storage unit 120, and a control unit 130.

[0030] (Regarding Communications Unit 110) The communication unit 110 is implemented by, for example, a NIC (Network Interface Card), a wireless LAN (Local Area Network) card, etc. The communication unit 110 is connected to the network N by wired or wireless connection and transmits and receives information with the user terminal 200, etc.

[0031] (Regarding memory unit 120) The storage unit 120 includes a main memory and an external storage device. The main memory stores programs executed by the control unit 130, or data processed by the control unit 130. The main memory may be implemented using semiconductor memory elements such as RAM (Random Access Memory), ROM (Read Only Memory), or flash memory. The external storage device stores data processed by the control unit 130. The external storage device may be implemented using, for example, a hard disk, SSD (Solid State Drive), magnetic tape, or optical disc.

[0032] As shown in Figure 3, the storage unit 120 includes a search query storage unit 121, a related search query storage unit 122, and a model storage unit 123. These configurations will be described in detail below.

[0033] (Regarding the search query storage unit 121) The search query storage unit 121 stores the search query entered by the user. Here, an example of the information stored in the search query storage unit 121 will be explained using Figure 4. Figure 4 is a diagram showing an example of the information stored in the search query storage unit of the information processing device according to this embodiment.

[0034] In the example shown in Figure 4, the search query storage unit 121 stores information related to the items "User ID," "Acquisition Date and Time," "Search Query ID," and "Search Query" in an associated manner.

[0035] The "User ID" is an identifier that identifies the user and is represented by a string or number. The "Acquisition Date and Time" indicates the date and time the search query indicated by the "Search Query" was acquired. The "Search Query ID" is an identifier that identifies the search query and is represented by a string or number. The "Search Query" is the search query associated with the "Search Query ID".

[0036] In other words, Figure 4 shows that the search query entered by the user identified by the user ID "User ID #1" was retrieved at the date and time indicated by the retrieval date and time "Retrieval Date and Time #1-1", and that the search query was identified by the search query ID "Search Query ID #1-1" and stored as search query "Search Query #1-1".

[0037] Furthermore, the information stored in the search query storage unit 121 is not limited to information relating to the items "User ID," "Acquisition Date and Time," "Search Query ID," and "Search Query," but may also store other information related to any search query.

[0038] (Regarding the related search query storage unit 122) The related search query storage unit 122 stores related search queries that are related to the search query. Here, an example of the information stored in the related search query storage unit 122 will be explained using Figure 5. Figure 5 is a diagram showing an example of the information stored in the related search query storage unit of the information processing device according to the embodiment.

[0039] In the example shown in Figure 5, the related search query storage unit 122 stores information related to the items "related search query ID," "related search query," and "vector representation" in an associated manner.

[0040] The "Related Search Query ID" is an identifier that identifies a related search query and is represented by a string, number, or other means. The "Related Search Query" is a search query that has been entered by a user as a search query, such as a string or sentence. The "Vector Representation" is a vector representation of the linguistic expression of the string or sentence of the related search query indicated by the "Related Search Query". In this case, the vector representation may be stored as a representation converted from the linguistic expression by a PLM such as BERT, or it may be a representation converted using a model that has improved the conversion performance to vector representation by performing comparative learning with SimCSE on a PLM such as BERT.

[0041] In other words, Figure 5 shows that the related search query identified by the related search query ID "Related Search Query ID #1" is "Related Search Query #1", and the vector representation of that related search query is "Vector Representation #1".

[0042] Furthermore, the information stored in the related search query storage unit 122 is not limited to information relating to the items "related search query ID," "related search query," and "vector representation," but may also store any other information related to related search queries.

[0043] (Regarding the model memory unit 123) The model storage unit 123 stores a model that converts the linguistic representation of a search query into a vector representation. Here, an example of the information stored in the model storage unit 123 will be explained using Figure 6. Figure 6 is a diagram showing an example of the information stored in the model storage unit of the information processing device according to this embodiment.

[0044] In the example shown in Figure 6, the model storage unit 123 stores information related to the items "Model ID" and "Model Data" in an associated manner.

[0045] The "Model ID" is an identifier that identifies the machine learning model that converts the linguistic expression of a search query into a vector representation, and is represented by a string, number, or other means. The "Model Data" refers to the model data of the machine learning model that converts the linguistic expression of a search query into a vector representation. For example, the machine learning model may be a neural network.

[0046] In other words, in Figure 6, the model identified by model ID "M#1" represents the machine learning model M#1. Furthermore, the model data "MDT#1" represents the model data for machine learning model M#1.

[0047] Furthermore, the information stored in the model storage unit 123 is not limited to information related to the items "Model ID" and "Model Data," but may also store other information related to the model that converts the linguistic expression of a search query into a vector expression.

[0048] (Regarding the control unit 130) Next, returning to Figure 3, the control unit 130 will be described. The control unit 130 is a controller that controls the information processing device 100, and is realized by, for example, a CPU (Central Processing Unit) or MPU (Micro Processing Unit) reading various programs stored in the memory unit 120 of the information processing device 100 and executing them using RAM as the working area. The control unit 130 is a controller and may also be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).

[0049] As shown in Figure 3, the control unit 130 includes an acquisition unit 131, a conversion unit 132, a specification unit 133, an inverse conversion unit 134, and a proposal unit 135. The control unit 130 reads a program from the storage unit 120 and executes it using RAM as a working area, thereby realizing or executing the information processing functions and operations described below. These processes will be described in order below.

[0050] (Regarding acquisition unit 131) The acquisition unit 131 acquires the search query entered by the user into the user terminal 200 from the user terminal 200. Here, the search query is represented by a combination of words or sentences in a predetermined language. For example, the acquisition unit 131 acquires the combination of words "anime A", "character", and "list" as the search query entered by the user from the user terminal 200. Thus, the search query may be a combination of multiple words, a single word, or a combination of multiple sentences. Once the acquisition unit 131 has acquired the search query entered by the user from the user terminal 200, it stores the acquired search query in the search query storage unit 121.

[0051] (Regarding the conversion unit 132) The conversion unit 132 converts the linguistic representation of the acquired search query into a vector representation. For example, the conversion unit 132 may perform processing such as cleaning, morphological analysis, normalization (conversion to standard form), and stop word removal as needed before converting to a vector representation. The conversion of the linguistic representation of the search query into a vector representation in the conversion unit 132 may be achieved by using Word2Vec, Retrofitting, Glove, FastText, BERT, etc.

[0052] The conversion unit 132 uses a model that has learned the relationship between the linguistic and vector representations of search queries through comparative learning, using search queries from the same session as positive examples, to convert the linguistic representation of the search query obtained by the acquisition unit 131 into a vector representation. In other words, the conversion unit 132 uses a model that has been trained so that the vector representations of search queries in the same session are similar, using search queries that the user has entered consecutively within a predetermined period as positive examples, to convert the linguistic representation of the acquired search query into a vector representation. In this case, the training model may be implemented, for example, by applying SimCSE to BERT and training it.

[0053] Furthermore, the conversion unit 132 may, in addition to the aforementioned positive examples, use a model trained with negative examples of search queries that have similar literal meanings but different meanings and vector representations, to convert the linguistic representation of the acquired search query into a vector representation. For example, suppose that when the search query "Sanjusangen-do admission fee" is entered into a search engine such as a vector search model, related search queries such as (a) "Kyoto Sanjusangen-do admission fee", (b) "Sanjusangen-do admission fee 2022", and (c) "Sangatsu-do admission fee" are output. Suppose that (a) and (b) were clicked by the user, but (c) was not clicked. In this case, the search queries (a) and (b) are given as positive examples, and the search query (c) is given as a negative example, and SimCSE is applied to train a model such as BERT.

[0054] This learning method will be explained using Figure 7. Figure 7 is a flowchart showing the flow of the learning method for the vector search model according to the embodiment. As shown in Figure 7, the learning method for the vector search model according to the embodiment is divided into a first stage from steps S1 to S3 and a second stage from steps S4 to S5. First, the vector search model (a search engine for related words using a model that has been trained by applying SimCSE to a model such as BERT) is run in a real service to output related words (step S1). Next, from the output related words, those that were clicked or had a high click rate among the related words output by the vector search model are collected as positive examples, and those that were not clicked or had a low click rate among the related words output by the vector search model are collected as negative examples (step S2). Then, these positive and negative examples are saved in the search log (step S3).

[0055] Next, steps S4 and S5 will be described as the second stage of the vector search model learning method according to the embodiment. Positive and negative examples stored in the search log are input into the vector search model for learning (session data may also be added for learning) (step S4). Next, the learned vector search model is put into operation in the actual service (step S5). By repeating the first and second stages thereafter, a vector search model that further reflects the user's selection tendencies for related words can be generated. With this method of learning a vector search model, the vector search model can be adjusted by appropriately reflecting the user's selection tendencies for related words in the actual service and taking into account changes in the user's selection tendencies. Therefore, a vector search model that reflects the changing needs of the user can be generated.

[0056] (Regarding Section 133) The identification unit 133 identifies vector representations of related search queries that are similar to the vector representation based on the vector representation transformed by the transformation unit 132. First, the identification unit 133 calculates the similarity by calculating the cosine similarity of the vector representations of related search queries stored in the related search query database (for example, the related search query storage unit 122). Note that the calculation of similarity is not limited to cosine similarity; the similarity between distributed representations may be calculated based on any index that can be applied as a distance measure between vectors. For example, the similarity may be calculated using the Euclidean distance between vector representations, the distance in a non-Euclidean space such as hyperbolic space, the Manhattan distance, the Mahalanobis distance, or a predetermined distance function. Then, based on the calculated similarity, the identification unit 133 may identify, for example, vector representations of related search queries with a similarity of above a predetermined threshold as vector representations of related search queries with high similarity.

[0057] The identification unit 133 uses a model that has learned the relationship between search queries and vector representations through comparative learning, using search queries from the same session as positive examples, to convert related search queries into vector representations. The identification unit 133 then refers to the database of related search queries (e.g., related search query storage unit 122) and identifies vector representations of related search queries that are similar to the vector representations converted by the conversion unit 132. In other words, the vector representations of related search queries stored in the database of related search queries (e.g., related search query storage unit 122) used by the identification unit 133 to identify related search queries are those in which the linguistic representation of a related search query has been converted into a vector representation using a model that has learned the relationship between the linguistic representation and vector representation of a search query through comparative learning, using search queries from the same session as positive examples.

[0058] (Regarding the inverse conversion unit 134) The inverse transformation unit 134 converts the vector representation of the related search query identified by the identification unit 133 back into a linguistic representation. For example, the inverse transformation unit 134 may convert the vector representation of a related search query with a high similarity identified by the identification unit 133 into a linguistic representation of the related search query, based on the correspondence between the linguistic representation of the related search query stored in the database (e.g., the related search query storage unit 122) and its vector representation.

[0059] (Regarding Proposal 135) The proposal unit 135 displays the related search queries that the inverse transformation unit 134 has inverted on the user terminal 200 in order of similarity. For example, suppose a search query is "anime A", "character", "list", and the related search queries identified are combination A, "anime A", "character list", combination B, "anime A", "character", and combination C, "anime A", "character", "popular". Suppose the similarity of combination A is calculated to be 0.8, the similarity of combination B is 0.7, and the similarity of combination C is 0.5. In this case, the proposal unit 135 displays the related search queries on the user terminal 200 in the order of combination A, combination B, and combination C.

[0060] [4. User terminal configuration] Next, the configuration of the user terminal 200 according to the embodiment will be described using Figure 8. Figure 8 is a diagram showing an example of the configuration of a user terminal according to the embodiment. As shown in Figure 8, the user terminal 200 has a communication unit 210, an input unit 220, an output unit 230, and a control unit 240.

[0061] The communication unit 210 is implemented by, for example, a NIC, a wireless LAN card, etc. The communication unit 210 is connected to the network N by wire or wireless connection and transmits and receives various types of information to and from the information processing device 100 via the network N.

[0062] The input unit 220 receives various operation information from the user. For example, the input unit 220 may accept various operations from the user via a touch panel on a display surface (e.g., the output unit 230). Alternatively, the input unit 220 may accept various operations from buttons on the user terminal 200, or from a keyboard or mouse connected to the user terminal 200.

[0063] The output unit 230 is a display screen for a tablet terminal, for example, which is implemented using a liquid crystal display or an organic EL (Electro-Luminescence) display, and is a display device for displaying various information. In other words, if the input unit 220 of the user terminal 200 is a touch panel, the output unit 230, which is the display screen, accepts user input and also outputs to the user. The output unit 230 may also be a speaker, and may output sound through the speaker.

[0064] The control unit 240 is implemented, for example, by a CPU or MPU executing various programs stored in the user terminal 200 using RAM as a working area. Alternatively, the control unit 240 may be implemented by an integrated circuit such as an ASIC or FPGA.

[0065] As shown in Figure 8, the control unit 240 has a reception unit 241 and a proposal unit 242.

[0066] The reception unit 241 receives search queries from users. For example, the reception unit 241 accepts the combination of words "anime A", "character", and "list" as a search query from a user. The reception unit 241 may also receive search queries from users via the input unit 220.

[0067] The suggestion unit 242 suggests related search queries related to the search query entered by the user. For example, suppose the information processing device 100 calculates that the similarity of related search query combination A is 0.8, the similarity of combination B is 0.7, and the similarity of combination C is 0.5. In this case, the suggestion unit 242 displays the related search queries in the order of combination A, combination B, and combination C on the output unit 230, based on the information transmitted from the suggestion unit 135 of the information processing device 100.

[0068] [5. Information Processing Flow] Next, the information processing procedure according to the embodiment will be described using Figure 9. Figure 9 is a flowchart showing an example of information processing according to the embodiment. The information processing procedure according to the embodiment will be described below in accordance with the flowchart shown in Figure 9.

[0069] First, the information processing device 100 obtains the search query entered by the user into the user terminal 200 from the user terminal 200 (step S101). Next, the information processing device 100 converts the linguistic expression of the obtained search query into a vector expression (step S102). Then, based on the converted vector expression, the information processing device 100 identifies vector expressions of related search queries that are similar to the vector expression in question (step S103). Then, the information processing device 100 converts the vector expressions of the identified related search queries back into linguistic expressions (step S104). Finally, the information processing device 100 displays the converted search queries on the user terminal 200 in descending order of similarity (step S105).

[0070] According to this, related search queries associated with the search query entered by the user can be broadly extracted based on the similarity of their vector representations and presented to the user in order of similarity. Therefore, it is possible to provide an information processing device, an information processing method, and an information processing program that can accurately present related search queries.

[0071] [6. Hardware Configuration] Furthermore, the information processing device 100 according to the above-described embodiment is realized by a computer 1000 having a configuration such as that shown in Figure 10. Figure 10 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device. The computer 1000 is connected to an output device 1010 and an input device 1020, and has a configuration in which an arithmetic unit 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected by a bus 1090.

[0072] The arithmetic unit 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050, as well as programs read from the input device 1020, and executes various processes. The primary storage device 1040 is a memory device, such as RAM, that temporarily stores data used by the arithmetic unit 1030 for various calculations. The secondary storage device 1050 is a storage device that stores data used by the arithmetic unit 1030 for various calculations, and is implemented using ROM (Read Only Memory), HDD (Hard Disk Drive), flash memory, etc.

[0073] Output IF1060 is an interface for transmitting information to be output to output devices 1010, which output various types of information such as monitors and printers. It is implemented using connectors of standards such as USB (Universal Serial Bus), DVI (Digital Visual Interface), and HDMI (High Definition Multimedia Interface). Input IF1070 is an interface for receiving information from various input devices 1020, such as mice, keyboards, and scanners. It is implemented using, for example, USB.

[0074] The input device 1020 may also be a device that reads information from, for example, an optical recording medium such as a CD (Compact Disc), DVD (Digital Versatile Disc), or PD (Phase Change Rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), tape media, magnetic recording media, or semiconductor memory. Furthermore, the input device 1020 may also be an external storage medium such as a USB memory stick.

[0075] Network IF1080 receives data from other devices via network N and sends it to the arithmetic unit 1030, and also transmits data generated by the arithmetic unit 1030 to other devices via network N.

[0076] The arithmetic unit 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070. For example, the arithmetic unit 1030 loads a program from the input device 1020 or the secondary storage device 1050 onto the primary storage device 1040 and executes the loaded program.

[0077] For example, when a computer 1000 functions as an information processing device 100, the arithmetic unit 1030 of the computer 1000 executes a program loaded onto the primary storage device 1040, thereby realizing the functions of the control unit 130 of the information processing device 100.

[0078] [7. Structure and Effects] The information processing device 100 according to this disclosure includes: an acquisition unit 131 that acquires a search query entered by a user into a user terminal 200 from the user terminal 200; a conversion unit 132 that converts the language expression of the acquired search query into a vector expression; an identification unit 133 that identifies vector expressions of related search queries similar to the vector expression based on the vector expression converted by the conversion unit 132; an inverse conversion unit 134 that inversely converts the vector expressions of the related search queries identified by the identification unit 133 back into language expressions; and a suggestion unit 135 that displays the search queries inversely converted by the inverse conversion unit 134 on the user terminal 200 in order of decreasing similarity.

[0079] This configuration allows for the extraction of a wide range of related search queries based on the similarity of their vector representations, and presenting them to the user in order of similarity. Therefore, it is possible to provide an information processing device that can accurately present relevant search queries.

[0080] The conversion unit 132 of the information processing device 100 according to this disclosure uses a model that has learned the relationship between the linguistic representation and the vector representation of a search query by comparative learning, with the same session as a positive example, to convert the linguistic representation of the search query acquired by the acquisition unit 131 into a vector representation.

[0081] This configuration allows for the extraction of related search queries in relation to a given search query. Search queries entered consecutively within the same session, where the user's search intent is considered similar, can be extracted as highly similar search queries. Therefore, the information processing device 100 can accurately present related search queries.

[0082] The identification unit 133 of the information processing device 100 according to this disclosure uses a model that has learned the relationship between search queries and vector representations through comparative learning, with search queries for the same session as positive examples, and after converting related search queries into vector representations, it refers to a database of related search queries and identifies vector representations of related search queries that are similar to the vector representations converted by the conversion unit 132.

[0083] This configuration allows for the extraction of related search queries in relation to a given search query. Search queries entered consecutively within the same session, where the user's search intent is considered similar, can be extracted as highly similar search queries. Therefore, the information processing device 100 can accurately present related search queries.

[0084] The information processing method relating to this disclosure includes the steps of: obtaining a search query entered by a user into the user terminal 200 from the user terminal 200; converting the linguistic expression of the obtained search query into a vector expression; identifying vector expressions of related search queries similar to the vector expression based on the converted vector expression; converting the vector expressions of the identified related search queries back into linguistic expressions; and displaying the back-converted search queries on the user terminal 200 in descending order of similarity.

[0085] This configuration allows for the extraction of a wide range of related search queries based on the similarity of their vector representations, and presenting them to the user in order of similarity. Therefore, it provides an information processing method that can accurately present relevant search queries.

[0086] The information processing program relating to this disclosure causes a computer to perform the following steps: obtain a search query entered by the user into the user terminal 200 from the user terminal 200; convert the linguistic expression of the obtained search query into a vector expression; identify vector expressions of related search queries similar to the vector expression based on the converted vector expression; convert the vector expressions of the identified related search queries back into linguistic expressions; and display the back-converted search queries on the user terminal 200 in descending order of similarity.

[0087] This configuration allows for the extraction of a wide range of related search queries based on the similarity of their vector representations, and presenting them to the user in order of similarity. Therefore, it is possible to provide an information processing program that can accurately present relevant search queries.

[0088] Although embodiments of the present application have been described in detail based on the drawings, these are illustrative examples, and the present invention can be implemented in various other forms, including those described in the disclosure section of the invention, based on the knowledge of those skilled in the art.

[0089] Furthermore, the terms "section, module, unit" mentioned above can be replaced with "means" or "circuit." For example, the acquisition unit 131 can be replaced with acquisition means or acquisition circuit. [Explanation of symbols]

[0090] 100 Information Processing Devices 110 Communications Department 120 Storage section 121 Search query storage unit 122 Related Search Query Storage 123 Model Memory Unit 130 Control Unit 131 Acquisition Department 132 Conversion section 133 Specific part 134 Inverse Transform Section 135 Proposal Department 200 user terminals 210 Communications Department 220 Input section 230 Output section 240 Control Unit 241 Reception Department 242 Proposal Department N Network

Claims

1. An acquisition unit that obtains search queries, which are combinations of words or sentences in a specified language, entered by the user into the user terminal, from the user terminal, The acquisition unit acquires the linguistic representation of the search query, and the transformation unit converts it into a vector representation using a model that has learned the relationship between the linguistic representation and the vector representation of the search query through comparative learning, with the search query from the same session as a positive example. Based on the vector representation converted by the conversion unit, the identification unit identifies a vector representation of a related search query, which is a linguistic representation of a string or sentence, that is similar to the vector representation. The aforementioned identification unit converts the vector representation of the related search query identified by the identification unit back into a linguistic representation based on a predetermined correspondence between the linguistic representation and the vector representation of the related search query that is stored in advance, and the inverse conversion unit converts the vector representation of the related search query identified by the identification unit back into a linguistic representation, The system includes a suggestion unit that displays the search queries in reverse transformed by the reverse transformation unit on the user terminal in order of similarity. Information processing device.

2. The identification unit uses a model that has learned the relationship between search queries and vector representations through comparative learning, with search queries for the same session as positive examples, and refers to a database of related search queries that stores the vector representations of related search queries to identify vector representations of related search queries that are similar to the vector representation transformed by the transformation unit. The information processing apparatus according to claim 1.

3. An information processing method performed by an information processing device, The steps include obtaining a search query from the user's terminal, which is a combination of words or sentences in a specified language entered by the user into the user's terminal, The steps include: converting the linguistic representation of the retrieved search query into a vector representation using a model that has learned the relationship between the linguistic representation and vector representation of the search query through comparative learning, with the search query from the same session as a positive example; and The steps include identifying a vector representation of a related search query, which is a linguistic representation of a string or sentence, that is similar to the converted vector representation, based on the converted vector representation, The steps include: converting the vector representation of the identified related search query back into a linguistic representation based on a predetermined correspondence between the linguistic representation and vector representation of the related search query that has been stored in advance; The steps include displaying the reverse-transformed search queries on the user's terminal in order of similarity, and Information processing methods including

4. The steps include obtaining a search query from the user's terminal, which is a combination of words or sentences in a specified language entered by the user into the user's terminal, The steps include: converting the linguistic representation of the retrieved search query into a vector representation using a model that has learned the relationship between the linguistic representation and vector representation of the search query through comparative learning, with the search query from the same session as a positive example; and The steps include identifying a vector representation of a related search query, which is a linguistic representation of a string or sentence, that is similar to the converted vector representation, based on the converted vector representation, The steps include: converting the vector representation of the identified related search query back into a linguistic representation based on a predetermined correspondence between the linguistic representation and vector representation of the related search query that has been stored in advance; The steps include displaying the reverse-transformed search queries on the user's terminal in order of similarity, and An information processing program that causes a computer to execute something.

Images