KR102661768B1 - Method for generateing recommended query - Google Patents

Abstract

According to some embodiments of the present disclosure, a method for generating recommended queries performed by a processor of a computing device is disclosed. The method includes: filtering N candidate questions (N is a natural number of 1 or more) corresponding to a user input from among a plurality of candidate questions based on the first method; Classifying the N candidate queries into at least one semantic group; Based on the second method, selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group; Based on the third method, selecting L query sentences (L is a natural number of 1 or more) from each of the at least one semantic group; and determining the K questions and the L questions as recommended questions.

Description

How to generate a recommended query {METHOD FOR GENERATEING RECOMMENDED QUERY}

The present invention relates to a method for generating recommended questions, and specifically, to a method for generating and providing recommended questions in a chatbot system.

In a chatbot system, the chatbot's response method is to search the chatbot sentence database for similar sentences to the user's input question and output the corresponding answer. Basically, this method builds a query database consisting of a query that serves as the standard for the content (standard query) and a query that has the same meaning as the standard query but has a different expression (modified query). The main method is to search for sentences that correspond to the customer's inquiry.

Chatbot systems can have a variety of purposes. For example, a chatbot system may include a free conversation chatbot to replace a user's conversation partner, or a business chatbot to handle a specific task. Here, it can be judged that responding to various inquiries is an important factor for free conversation chatbots. Meanwhile, for business chatbots, it may be judged that providing accurate responses to inquiries related to specific tasks is a more important factor than responding to a variety of inquiries.

Since these business chatbots are specialized for specific tasks, it may be difficult for users who are using business chatbots for the first time to figure out in advance what tasks the chatbot can handle.

Accordingly, there may be a demand for a function that can notify users in advance of work areas that business chatbots can handle.

Republic of Korea Patent Publication No. 10-2008169

This disclosure was developed in response to the above-described background technology, and informs users in advance of work areas that a business chatbot can handle through a method for generating recommended queries.

The technical problems of the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

According to some embodiments of the present disclosure for solving the above-mentioned problems, a method for generating recommended queries performed by a processor of a computing device is disclosed. The method includes: filtering N candidate questions (N is a natural number of 1 or more) corresponding to a user input from among a plurality of candidate questions based on the first method; Classifying the N candidate queries into at least one semantic group; Based on the second method, selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group; Based on the third method, selecting L query sentences (L is a natural number of 1 or more) from each of the at least one semantic group; and determining the K questions and the L questions as recommended questions.

In addition, according to some embodiments of the present disclosure for solving the problems described above, a computing device for generating recommended queries is disclosed. The device includes: a memory storing a plurality of candidate queries; and a processor that generates a recommended query using the plurality of candidate queries, wherein the processor selects N (N) corresponding to user input from among the plurality of candidate queries based on the first method. Filter candidate queries (N is a natural number greater than or equal to 1), classify the N candidate queries into at least one semantic group, and based on the second method, K number (K is select a number of queries (a natural number of 1 or more), and based on the third method, select L questions (L is a natural number of 1 or more) from each of the at least one semantic group, and select the K questions and the L number of questions. The query can be determined from the above recommended query.

Additionally, according to some embodiments of the present disclosure for solving the problems described above, a computer program stored in a computer-readable storage medium is disclosed. The computer program includes instructions for causing a processor of the computing device to perform the following steps, which include: Based on the first method, select N (N is filtering candidate queries (a natural number of 1 or more); Classifying the N candidate queries into at least one semantic group; Based on the second method, selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group; Based on the third method, selecting L query sentences (L is a natural number of 1 or more) from each of the at least one semantic group; and determining the K questions and the L questions as recommended questions.

The technical solutions obtainable from the present disclosure are not limited to the solutions mentioned above, and other solutions not mentioned above will become clear to those skilled in the art from the description below. It will be understandable.

The present disclosure can inform users in advance of work areas that a business chatbot can handle through a method for generating recommended queries.

The effects that can be obtained from the present disclosure are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. .

1 is a block diagram of a computing device according to some embodiments of the present disclosure.
Figure 2 is a flowchart illustrating an example of a method for determining a recommended query according to some embodiments of the present disclosure.
FIG. 3 is a diagram illustrating an example of filtered candidate queries according to some embodiments of the present disclosure.
Figure 4 is a diagram for explaining an example of a semantic group according to some embodiments of the present disclosure.
FIG. 5 is a diagram illustrating an example in which recommended questions are presented to a user according to some embodiments of the present disclosure.
FIG. 6 is a diagram illustrating an example of a method for calculating weights of candidate queries according to some embodiments of the present disclosure.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the disclosure. However, it is clear that these embodiments may be practiced without these specific descriptions.

As used herein, “embodiments,” “examples,” “aspects,” “examples,” etc. may not be construed to mean that any aspect or design described is better or advantageous over other aspects or designs. .

The term “or” is intended to mean an inclusive “or” and not an exclusive “or.” That is, unless otherwise specified or clear from context, “X utilizes A or B” is intended to mean one of the natural implicit substitutions.

Additionally, the terms “comprise” and/or “comprising” should be understood to mean that the corresponding feature and/or element is present. However, the terms “comprise” and/or “comprising” should be understood as not excluding the presence or addition of one or more other features, elements and/or groups thereof. Additionally, unless otherwise specified or the context is clear to indicate a singular form, the singular terms herein and in the claims should generally be construed to mean “one or more.”

1 is a block diagram of a computing device according to some embodiments of the present disclosure.

According to some embodiments of the present disclosure, the computing device 100 may be a user terminal or a server for providing chatbot functions to users. For example, the computing device 100 of the present disclosure may be a chatbot system or a portion thereof. Here, the user terminal may be a terminal in which a program for providing a chatbot function is stored in the memory 120, or a program for providing a separate chatbot function is not installed in the memory 120 and may be installed on a server that provides the chatbot function. It may be a terminal that connects and receives recommended questions from the server.

Referring to FIG. 1 , computing device 100 may include a processor 110 and memory 120 . However, the above-described components are not essential for implementing the computing device 100, so the computing device 100 may have more or less components than the components listed above.

Computing device 100 may include any type of computer system or computer device, such as, for example, microprocessors, mainframe computers, digital processors, portable devices, and device controllers. However, it is not limited to this.

The processor 110 of the computing device 100 typically controls the overall operation of the computing device 100. The processor 110 may run an application program stored in the memory 120 to provide a chatbot function through components included in the computing device 100, thereby providing or processing appropriate information or functions to the user.

Additionally, the processor 110 may control at least some of the components of the computing device 100 to run an application program (eg, an application related to a chatbot function) stored in the memory 120. Furthermore, the processor 110 may operate in combination with at least two or more of the components included in the computing device 100 to run the application program.

The processor 110 of the computing device 100 according to some embodiments of the present disclosure may obtain user input from an application (or program) performing a chatbot function and provide the user with a recommended query based on the user input. there is.

The computing device 100 of the present disclosure may be a user terminal executing an application that performs a chatbot function. However, it is not limited to this, and the computing device 100 of the present disclosure transmits and receives data related to recommended questions (e.g., user input data or recommended question data, etc.) with a user terminal running an application that performs a chatbot function. It could be a server that does it.

For example, when the computing device 100 of the present disclosure is a user terminal, the processor 110 may obtain user input through an interface unit (not shown) provided in the user terminal. Additionally, the processor 110 may generate a recommended query based on user input. Additionally, the processor 110 may output a recommended query through a display unit (not shown) provided in the user terminal. However, it is not limited to this.

As another example, when the computing device 100 of the present disclosure is a server, the processor 110 may receive user input information from the user terminal. Additionally, the processor 110 may generate a recommended query based on user input information. Additionally, the processor 110 may transmit recommended question information to the user terminal so that the recommended question is displayed on a display provided in the user terminal.

That is, the computing device 100 may further include a communication unit (not shown). The communication unit may include one or more modules that enable communication between the computing device 100 and another computing device or between the computing device 100 and a user terminal. Additionally, the communication unit may include one or more modules that connect the computing device 100 to one or more networks.

The network connecting communication between the computing device 100 and other computing devices or user terminals may be any wired or wireless communication network capable of transmitting and receiving data and signals of any type.

According to some embodiments of the present disclosure, memory 120 may include any type of storage medium. Specifically, the memory 120 can store programs for operating the processor 110, and can temporarily or permanently store input/output data.

The memory 120 of the computing device 100 according to some embodiments of the present disclosure may store candidate queries that the processor 110 uses when generating a recommended query.

That is, the processor 110 of the computing device 100 may generate a recommended query using the candidate query stored in the memory 120 and the user input. Additionally, the processor 110 may present a recommended query to the user.

Accordingly, the computing device 100 of the present invention can inform the user in advance of work areas that the business chatbot can handle.

A detailed description of how the computing device 100 of the present disclosure generates a recommended query and presents the recommended query to a user will be described later with reference to FIGS. 2 to 6 .

Figure 2 is a flowchart illustrating an example of a method for determining a recommended query according to some embodiments of the present disclosure.

Referring to FIG. 2 , the processor 110 of the computing device 100 may filter N candidate queries corresponding to user input from among a plurality of candidate queries based on the first method. Here, N is a natural number greater than or equal to 1, and may be a value set by the user operating the chatbot system, or a value with no separate restrictions. For example, N is not a separate limit and may be the total number of all candidate queries containing a specific word included in the user input. However, it is not limited to this.

Specifically, the processor 110 may use string matching to filter N candidate queries that have a similarity greater than a preset level to the user input from among a plurality of candidate queries.

More specifically, the processor 110 may recognize at least one character string corresponding to a user input. Additionally, the processor 110 may filter candidate queries that include at least one recognized string from among the plurality of candidate queries.

For example, if at least one string corresponding to the user input is 'insurance', the processor 110 selects a candidate query 'Tell me about the insurance subscription information page' that includes the string 'insurance' from among the plurality of candidate queries. You can filter statements. However, it is not limited to this.

The plurality of candidate queries described above may be stored in the memory 120 of the computing device 100. That is, the processor 110 may filter N candidate queries corresponding to user input from among the plurality of candidate queries stored in the memory 120.

Hereinafter, candidate queries filtered by the processor 110 of the computing device 100 will be described with reference to FIG. 3 .

Meanwhile, weights and semantic classifications may be set for each of the plurality of candidate queries. Here, the weight and semantic classification set for each of the plurality of candidate queries may be used by the processor 110 to determine a recommended query among the N filtered candidate queries, as will be described later.

In the present disclosure, the weight may be a value indicating accuracy for a recommendation query. For example, the weight is a rational number between 0 and 1, and the closer the weight set to a query is to 1, the more often the query is selected by the user. However, it is not limited to this.

In addition, semantic classification and semantic group may mean identification values for classifying tasks or purposes related to a set candidate query (or recommended query). For example, semantic classification and semantic group may be identifying information that indicates the purpose of the query, such as 'insurance subscription', 'login/authentication', or 'insurance coverage'. However, it is not limited to this.

The processor 110 of the computing device 100 may filter the N candidate queries and then classify the N candidate queries into at least one semantic group (S120).

Specifically, the processor 110 may classify candidate queries having the same semantic classification into the same group based on the semantic classification set for each of the N candidate queries.

Hereinafter, a description of how the processor 110 of the computing device 100 classifies candidate queries into the same group using semantic classification will be described later with reference to FIG. 4 .

Meanwhile, the processor 110 of the computing device 100 classifies candidate queries having the same semantic classification into the same group, and then, based on each of the second method and the third method, preset questions in each of at least one semantic group. You can select any number of questions. Additionally, the processor 110 may determine the selected query statement as a recommended query statement.

Specifically, the processor 110 may select K queries from each of at least one semantic group based on the second method (S130). Here, K is a natural number of 1 or more, and may be a specific value set by the user operating the chatbot system, or a value set to correspond to the number of at least one semantic group. For example, K may be set in advance as 1 or 2 multiples of at least one semantic group. That is, if the candidate questions are classified into 5 semantic groups, the number of K may be 5 (if multiple of 1) or 10 (if multiple of 2) depending on the preset multiple. However, it is not limited to this.

More specifically, processor 110 may select one or more queries from each of all semantic groups such that K queries are selected. At the same time, when selecting a query from candidate queries classified into the same semantic group, the processor 110 may select a preset number of queries in descending order of weight.

For example, assuming that the first to third queries are classified into the first semantic group, and the fourth to sixth queries are classified into the second semantic group, the processor 110 selects the highest semantic group in the first semantic group. The first query with a weight (eg, 0.5) set can be selected. Additionally, the processor 110 may select the fifth query with the highest weight (eg, 0.7) in the second semantic group. In this case, the preset number of queries that the processor 110 selects in descending order of weight from among candidate queries classified into the same semantic group may be 1. Additionally, the number K of queries selected by the processor 110 based on the second method may be 2. However, it is not limited to this.

That is, the second method of the present disclosure may be a method of selecting several queries in order of high weight in each of at least one semantic group.

Meanwhile, the processor 110 may select L queries from each of at least one semantic group based on the third method (S140). Here, L is a natural number of 1 or more, and may be a specific value set by the user operating the chatbot system, a value set to correspond to the number of at least one semantic group, or a value set to correspond to K. For example, L may be set in advance to be a multiple of 1 or 2 of at least one semantic group. That is, if the candidate query is classified into 5 semantic groups, the number of L may be 5 (if multiple of 1) or 10 (if multiple of 2) depending on the preset multiple. For another example, L may be preset to be a multiple of 1 or 2 of K. That is, when the number of K is 3, the number of L may be 3 (if multiple of 1) or 6 (if multiple of 2) depending on the preset multiple. However, it is not limited to this.

Specifically, the processor 110 may select one or more queries from each of all semantic groups so that L queries are selected. At the same time, the processor 110 may randomly select a preset number of queries from among the candidate queries classified into the same semantic group, excluding the query sentences selected based on the second method described above.

For example, queries 1 to 3 are classified into the first semantic group, queries 4 to 6 are classified into the second semantic group, and queries 1 and 5 are classified in a second way. Assuming that it is selected, the processor 110 may randomly select at least one of the second and third queries, which are the remaining queries of the first semantic group. Additionally, the processor 110 may randomly select at least one of the remaining queries, the fourth query and the sixth query, from the second semantic group. In this case, the preset number of queries that the processor 110 randomly selects from among the remaining candidate queries classified into the same semantic group may be 1. Additionally, the number L, which is the number of queries selected by the processor 110 based on the third method, may be 2. However, it is not limited to this.

That is, the third method of the present disclosure may be a method of randomly selecting several queries from among the remaining queries for each of at least one semantic group that was not selected in the second method. For example, the third method may be stratified sampling. However, it is not limited to this.

The processor 110 of the computing device 100 according to some embodiments of the present disclosure may generate a recommendation query using two methods, as described above. In this case, it is possible to prevent a candidate query with a low weight from being determined as a recommended query.

Specifically, the weight of the present disclosure is calculated based on the number of recommendations and the number of selections. When queries are selected and recommended using only these weights, queries whose weights were initially set low cannot be recommended. To prevent this, the processor 110 of the present disclosure selects several questions using weights, randomly selects some questions from among the questions with relatively low weights, and presents various recommended questions to the user.

Accordingly, the processor 110 may provide various recommended queries to the user and update the weights of the recommended queries. In addition, because the processor 110 determines some recommended queries using weights that are updated from time to time, recommendations are made by reflecting the trend of queries that may vary depending on the time of chatbot use (e.g., year-end, beginning of the year, morning, evening, etc.) A query can be presented to the user.

Hereinafter, a description of a method by which the processor 110 of the computing device 100 selects a preset number of queries from each of at least one semantic group based on each of the second method and the third method will be provided in FIG. 4, which will be described later. For an easier explanation, refer to .

Meanwhile, the processor 110 of the computing device 100 may select K questions, select L questions, and then determine the K questions and L questions as recommended questions (S150).

Additionally, the processor 110 may present a recommended query to the user.

Specifically, when the processor 110 determines a recommended query, it can sort the recommended queries so that K queries are presented to the user with priority over L queries.

More specifically, the processor 110 may determine which of the K queries is presented preferentially based on the weight.

For example, the processor 110 may sort recommended queries among the K queries in order of highest weight so that queries with high weights are presented preferentially to the user. However, it is not limited to this.

The weight of the present disclosure may be calculated by dividing the first number of times each of the plurality of candidate questions is selected by the user when each is determined to be a recommended question by the second number of times each of the plurality of candidate questions is determined to be a recommended question. Here, the first number of times may be updated when each of the plurality of candidate questions is determined to be a recommended question, and the second number of times may be updated when the recommended question is selected by the user.

In other words, the weight may be higher when a recommended query is presented to the user, the more frequently it is selected by the user.

Accordingly, the processor 110 of the present disclosure, which sorts recommended queries so that K queries selected based on weights are presented preferentially to the user, can improve user convenience and user experience.

In addition, the processor 110 of the present disclosure sorts the recommended queries so that the recommended queries with a high weight among the K recommended queries are presented preferentially to the user, and recommends the accuracy of the recommended queries (the query desired by the user). (recommendation accuracy) can be increased.

Hereinafter, an example in which a recommended query is presented to the user will be described later with reference to FIG. 5, and an example in which weights are updated will be described later with reference to FIG. 6.

FIG. 3 is a diagram illustrating an example of filtered candidate queries according to some embodiments of the present disclosure. Figure 4 is a diagram for explaining an example of a semantic group according to some embodiments of the present disclosure.

According to some embodiments of the present disclosure, a plurality of candidate queries may be stored in the memory 120 of the computing device 100. Additionally, the processor 110 of the computing device 100 may filter N candidate queries corresponding to user input from among the plurality of candidate queries.

For example, when the user inputs the text 'insurance', the processor 110 of the computing device 100 may filter candidate queries containing the string 'insurance'.

Referring to FIG. 3, examples of filtered N candidate queries are shown in the form of a table 300 for easy understanding.

As shown, when the user inputs the text 'insurance', the processor 110 of the computing device 100 generates candidate queries including the string 'insurance', such as 'Tell me about the insurance subscription information page' and 'Insurance'. You can filter candidate questions such as ‘I want to sign up’, ‘Tell me about insurance coverage’, etc.

Additionally, a semantic classification 320 and a weight 330 may be set for each of the N filtered candidate queries 310 (here, 8).

These semantic classifications 320 and weights 330 can be used by the processor 110 to determine a recommended query among the filtered N candidate queries.

First, the processor 110 may classify candidate queries having the same semantic classification into the same group using the semantic classification 320 set for each of the candidate queries 310.

To be described in more detail with reference to FIG. 4, the processor 110 divides the candidate questions 310 shown in FIG. 3 into a first semantic group according to the semantic classification 320 set for each of the candidate questions 310. It can be classified into (410), second semantic group (420), and third semantic group (430). Here, each of the first semantic group 410, the second semantic group 420, and the third semantic group 430 is shown in the form of a table to facilitate understanding of the present disclosure, but is not limited thereto.

For example, the first semantic group 410 is a semantic group related to 'insurance subscription', the second semantic group 420 is a semantic group related to 'login/authentication', and the third semantic group 430 is ' It may be a group of meanings related to ‘insurance coverage’. However, it is not limited to this.

Next, the processor 110 classifies each of the candidate queries 310 by semantic group and then, based on each of the second method and the third method, generates a preset number of queries in each of at least one semantic group. You can choose.

For example, the processor 110 uses a second method (e.g., a method of selecting a query statement with a high weight within each semantic group) to select the highest weight of 0.5 in the first semantic group 410. The first query 411 with is set can be selected. Additionally, the processor 110 may select the second query 421 with the highest weight of 0.6 in the second semantic group 420 using the second method. Additionally, the processor 110 may use the second method to select the third query 431 with the highest weight of 0.7 in the third semantic group 430.

Additionally, the processor 110 may select at least one query from among the remaining queries 412 not selected from the first semantic group 410 using a third method (eg, stratified extraction method). Additionally, the processor 110 may select at least one query from the remaining queries 422 in the second semantic group 420 using the third method. Additionally, the processor 110 may select at least one query from among the remaining queries 432 of the third semantic group 430 using the third method.

Since the processor 110 of the present disclosure determines recommended queries using two methods (the second method and the third method), the weight set for the queries is low and therefore no recommendations are made for some queries at all. It can be prevented.

In this case, the processor 110 may provide various recommended questions to the user and update the weights of the recommended questions. In addition, because the processor 110 determines some recommended questions using weights that are updated from time to time, recommended questions can be presented to the user by reflecting the trend of inquiries that may vary depending on the time of chatbot use or the user's situation. there is.

Meanwhile, the query selected by the processor 110 as described above may be presented to the user as a recommended query.

Hereinafter, an example of how a recommended query is presented to a user will be described later with reference to FIG. 5 .

FIG. 5 is a diagram illustrating an example in which recommended questions are presented to a user according to some embodiments of the present disclosure.

According to some embodiments of the present disclosure, the processor 110 of the computing device 100 may filter N candidate queries corresponding to user input from among a plurality of candidate queries. Additionally, the processor 110 may classify N candidate queries into at least one semantic group. Additionally, the processor 110 may select K queries from each of at least one semantic group. Additionally, the processor 110 may select L queries from each of at least one semantic group. Additionally, the processor 110 may determine K queries and L queries as recommended queries.

Hereinafter, a description of how the processor 110 of the computing device 100 determines (or generates) a recommended query will be omitted since it has been described in detail with reference to FIGS. 2 to 4, and the method of presenting the recommended query to the user will be described below. Examples will be described in detail.

According to some embodiments of the present disclosure, recommended queries may be presented to the user in a sorted state.

Referring to Figure 5, an example of a recommended query is presented to the user is shown.

For example, a user may enter the user input 511 of 'insurance' into a predetermined input window 510 to use the chatbot function.

In this case, the processor 110 of the computing device 100 may present a recommended query 520 related to 'insurance', which is the user input 511, to the user.

Meanwhile, the processor 110 may present the recommended query 520 to the user in an aligned state, as described above.

Specifically, the processor 110 may arrange the K questions 521 selected through the second method at the top so that the K questions 521 are presented preferentially to the user. Additionally, the processor 110 may place the L number of questions 522 selected through the third method under the K number of questions 521.

More specifically, the processor 110 may arrange K queries 521 in order of highest weight.

For example, the weight set for the query 'Insurance coverage details inquiry' is 0.7, the weight set for the query 'How to log in to insurance services' is 0.6, and the weight set for the query 'Tell me about the insurance subscription information page' is 0.5. Assuming, the processor 110 can place 'Insurance coverage details inquiry', which is a query with the highest weight, at the top. Additionally, the processor 110 may place 'How to log in to insurance service', which is a query with the second highest weight, immediately below the top (i.e., in the second position). Additionally, the processor 110 may place the query with the third highest weight, 'Tell me about the insurance subscription guide page', in the third position. However, it is not limited to this.

According to some additional embodiments of the present disclosure, the processor 110 of the computing device 100 may place L queries 522 with relatively low weights between K queries 521 with relatively high weights. It may be possible.

For example, the processor 110 places the first query statement, which is one of the K query statements, at the top, places the second query statement, which is one of the L query statements, in the second position, and places the first query statement, which is one of the K query statements, at the second position. The other third query statement can be placed in the third position, and the fourth query statement, one of the L queries, can be placed in the fourth position. That is, the processor 110 may arrange some queries included in the K queries and some queries included in the L queries alternately.

Meanwhile, according to the present disclosure, the weight can be calculated based on the number of times each question is selected as a recommended question and the number of times each question is selected by the user. In this case, if queries are selected based only on weights, some queries with relatively high weights may be exposed to the user preferentially, thereby increasing the probability of being selected by the user. In addition, some queries with low weights may have a lower number of exposures to the user, lowering the probability of being selected by the user. In other words, only queries with high weights can have their weights continuously increased, and the weights of specific queries can be continuously lowered. However, according to at least one of the above-described embodiments, it is possible to prevent the weight from being continuously lowered because there is a possibility that some queries with low weights may also be selected. However, it is not limited to this.

Meanwhile, the processor 110 of the present disclosure provides recommended questions by arranging frequently selected questions to the user, thereby increasing user convenience. When a recommendation query is presented to a user through the recommendation query generation method of the present disclosure, the user may determine that the recommendation accuracy is high.

Hereinafter, a description of the method for calculating the weight of the present disclosure will be described later with reference to FIG. 6.

FIG. 6 is a diagram illustrating an example of a method for calculating weights of candidate queries according to some embodiments of the present disclosure.

According to some embodiments of the present disclosure, the weight of the present disclosure may be set to each of a plurality of candidate queries. These weights of the present disclosure may be used when the processor 110 of the computing device 100 determines a recommended query.

Additionally, the weights set for each of the plurality of candidate queries may be calculated and updated by the processor 110 of the computing device 100.

Referring to FIG. 6, an example of weights set for each of a plurality of candidate questions is shown in the form of a table 600 to facilitate understanding.

As shown, a weight 640 may be set for each of the plurality of candidate questions 610. Specifically, the number of recommendations 620 and the number of selections 630 may be mapped to each of the plurality of candidate questions 610. And, the weight 640 may be the number of selections 630 divided by the number of recommendations 620.

Here, the number of recommendations 620 refers to the number of times a specific query is determined as a recommended query by the processor 110 of the computing device 100. And, the number of selections 630 refers to the number of times a specific query is selected by the user when it is provided to the user as a recommended query.

The processor 110 of the computing device 100 of the present disclosure may determine a recommended query and update the weight of each of the recommended queries whenever a specific recommended query is selected by the user.

Specifically, the processor 110 of the computing device 100 may update the recommendation count 620 of each of the plurality of candidate queries when each of the plurality of candidate queries is determined to be a recommended query. Additionally, the processor 110 may update the selection count 630 when a specific recommended query is selected by the user. The processor 110 may update the weight when at least one of the recommendation count 620 or the selection count 630 is updated.

As described above, the processor 110 of the computing device 100 of the present disclosure may determine some recommended queries using weights that are updated from time to time. Accordingly, the processor 110 may present recommended queries to the user by reflecting query trends that may vary depending on the time of chatbot use or the user's situation.

Although the present disclosure has been described generally in the context of computer-executable instructions that can be executed on one or more computers, those skilled in the art will recognize that the disclosure can be combined with other program modules and/or implemented as a combination of hardware and software. You will know.

Computers typically include a variety of computer-readable media. Media accessible by a computer includes volatile and non-volatile media, transitory and non-transitory media, removable and non-removable media. By way of example, and not limitation, computer-readable media may include computer-readable storage media.

Computer-readable storage media refers to volatile and non-volatile media, transient and non-transitory media, removable and non-removable, implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Includes media. Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage. This includes, but is not limited to, a device, or any other medium that can be accessed by a computer and used to store desired information.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits and algorithm steps described in connection with the embodiments disclosed herein may be used in electronic hardware, (for convenience) It will be understood that the implementation may be implemented by various forms of program or design code (referred to herein as “software”) or a combination of both. To clearly illustrate this interoperability of hardware and software, various illustrative components, blocks, modules, circuits and steps have been described above generally with respect to their functionality. Whether this functionality is implemented as hardware or software depends on the specific application and design constraints imposed on the overall system. A person skilled in the art of this disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of this disclosure.

The various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term “article of manufacture” includes a computer program or media accessible from any computer-readable device. For example, computer-readable storage media include magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash. Includes, but is not limited to, memory devices (e.g., EEPROM, cards, sticks, key drives, etc.). The term “machine-readable media” includes, but is not limited to, wireless channels and various other media capable of storing, retaining, and/or transmitting instruction(s) and/or data.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Thus, the present disclosure is not limited to the embodiments presented herein but is to be interpreted in the broadest scope consistent with the principles and novel features presented herein.

Claims (12)

In a method of generating a recommendation query performed by a processor of a computing device,
Based on the first method, filtering N candidate questions (N is a natural number of 1 or more) corresponding to a user input from among a plurality of candidate questions;
Classifying the N candidate queries into at least one semantic group;
Based on the second method, selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group;
Based on the third method, selecting L query sentences (L is a natural number of 1 or more) from each of the at least one semantic group; and
determining the K questions and the L questions as recommended questions;
Including,
A weight is set for each of the plurality of candidate questions,
The weight is,
Calculated using the value of comparing the first number of times that each of the plurality of candidate questions is selected by the user when each of the plurality of candidate questions is determined to be the recommended question and the second number of times that each of the plurality of candidate questions is determined to be the recommended question. felled,
How to create a recommended query.
According to claim 1,
Based on the first method, the step of filtering N candidate questions (N is a natural number of 1 or more) corresponding to the user input from among the plurality of candidate questions,
Using string matching, filtering the N candidate queries that have a similarity of more than a preset level to the user input from among the plurality of candidate queries;
Including,
How to create a recommended query.
According to claim 1,
A semantic classification is further set for each of the plurality of candidate queries,
How to create a recommended query.
According to claim 3,
The step of classifying the N candidate queries into at least one semantic group is:
Classifying candidate queries having the same semantic classification into the same group based on the semantic classification set for each of the N candidate queries;
Including,
How to create a recommended query.
According to claim 3,
Based on the second method, the step of selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group includes:
One or more queries are selected from each of all semantic groups so that the K queries are selected, but when selecting a query from candidate queries classified into the same semantic group, a preset number of queries are selected in order of highest weight. steps;
Including,
How to create a recommended query.
According to claim 1,
Based on the third method, the step of selecting L questions (L is a natural number of 1 or more) from each of the at least one semantic group includes:
One or more queries are selected from each of all semantic groups so that the L number of queries are selected, and among the candidate queries classified into the same semantic group, the query sentences selected based on the second method are excluded from the remaining queries. Randomly selecting a preset number of questions;
Including,
How to create a recommended query.
According to claim 1,
The compared values above are,
Calculated by dividing the first number of times each of the plurality of candidate questions is selected by the user when each of the plurality of candidate questions is determined to be the recommended question by the second number of times each of the plurality of candidate questions is determined to be the recommended question,
How to create a recommended query.
According to claim 1,
The first number of times is,
Each of the plurality of candidate questions is updated when determined as the recommended question,
The second number of times is,
Updated when the recommended query is selected by the user,
How to create a recommended query.
According to claim 1,
Sorting the recommended questions so that the K questions are presented to the user with priority over the L questions;
Containing more,
How to create a recommended query.
According to clause 9,
The step of sorting the recommended questions so that the K questions are presented to the user with priority over the L questions is:
determining a preferentially presented question among the K questions based on a weight;
Including,
How to create a recommended query.
In a computing device that generates a recommendation query,
a memory storing a plurality of candidate queries; and
a processor that generates a recommended query statement using the plurality of candidate query statements;
Including,
The processor,
Based on the first method, filter N candidate queries (N is a natural number of 1 or more) corresponding to user input from among the plurality of candidate queries,
Classify the N candidate queries into at least one semantic group,
Based on the second method, select K questions (K is a natural number of 1 or more) from each of the at least one semantic group,
Based on the third method, select L questions (L is a natural number of 1 or more) from each of the at least one semantic group,
Determine the K questions and the L questions as the recommended questions,
A weight is set for each of the plurality of candidate questions,
The weight is,
Calculated using the value of comparing the first number of times that each of the plurality of candidate questions is selected by the user when each of the plurality of candidate questions is determined to be the recommended question and the second number of times that each of the plurality of candidate questions is determined to be the recommended question. felled,
A computing device that generates recommendation queries.
A computer program stored on a computer-readable storage medium, comprising:
The computer program includes instructions for causing a processor of a computing device to perform the following steps, which include:
Based on the first method, filtering N candidate questions (N is a natural number of 1 or more) corresponding to a user input from among a plurality of candidate questions;
Classifying the N candidate queries into at least one semantic group;
Based on the second method, selecting K questions (K is a natural number of 1 or more) from each of the at least one semantic group;
Based on the third method, selecting L query sentences (L is a natural number of 1 or more) from each of the at least one semantic group; and
determining the K questions and the L questions as recommended questions;
Including,
A weight is set for each of the plurality of candidate questions,
The weight is,
Calculated using the value of comparing the first number of times that each of the plurality of candidate questions is selected by the user when each of the plurality of candidate questions is determined to be the recommended question and the second number of times that each of the plurality of candidate questions is determined to be the recommended question. felled,
A computer program stored on a computer-readable storage medium.

Images