KR102549006B1 - System for company search using automatic correction of query vectors based on user behavior and method of the same - Google Patents

Abstract

A company search system and method using automatic correction of a query vector based on user behavior capable of recommending an appropriate company of interest in consideration of user attributes and company attributes related to the user in addition to user query keywords are disclosed. A company search system using automatic correction of a query vector based on user behavior according to an embodiment of the present invention includes: a query keyword input by a user to search for a company of interest, user attribute data related to the user, and a company to which the user belongs. a search condition setting module configured to set search conditions by extracting keywords from enterprise attribute data of the company; a search module configured to search for the company of interest based on a similarity between a search vector corresponding to the search condition and a company attribute vector, and output a search result related to the company of interest; and a search query correction module configured to generate a final search query vector by correcting the search condition according to a user's feedback on the search result.

Description

System for company search using automatic correction of query vectors based on user behavior and method of the same}

The present invention relates to a company search system and method using automatic correction of a query vector based on user behavior. A company search system and method using automatic correction of query vectors based on user behavior.

In general, a company recommendation system provides a function of filtering and displaying information that may be of interest to a specific user among companies that meet a category condition set by a user. However, the conventional company recommendation method derives results only from the query keywords entered by the user, and the number of hits and user preference of large or well-known companies within the search site are measured higher than those of small companies, resulting in relatively frequent search results. results in the displayed

Such corporate exposure methods are directly linked to corporate profits, which can cause problems that cause growth gaps and inequality among companies. In addition, there are cases in which an employee belonging to a company must sequentially find a business partner that can fulfill the role of the value chain of the project in order to succeed in the project that the company is planning to promote.

However, the conventional company search and recommendation system searches for and recommends related companies based on user information and areas of interest, and recommends companies similar to the company to which the user belongs or competitors. There is a limit to not being able to recommend related traders.

An object of the present invention is to provide a company search system and method using automatic correction of a query vector based on user behavior, capable of recommending an appropriate company of interest in consideration of user properties and company properties related to a user, in addition to user query keywords.

In addition, the present invention is to provide a company search system and method using automatic correction of user behavior-based query vectors, which can predict user feedback data through a recursive artificial neural network model and predict feedback from users without feedback. .

In addition, the present invention is to provide a company search system and method using automatic correction of a query vector based on user behavior that can be effectively used to sequentially discover customers who can fulfill the role of a value chain.

A company search system using automatic correction of a query vector based on user behavior according to an embodiment of the present invention includes: a query keyword input by a user to search for a company of interest, user attribute data related to the user, and a company to which the user belongs. a search condition setting module configured to set search conditions by extracting keywords from enterprise attribute data of the company; a search module configured to search for the company of interest based on a similarity between a search vector corresponding to the search condition and a company attribute vector, and output a search result related to the company of interest; and a search query correction module configured to generate a final search query vector by correcting the search condition according to a user's feedback on the search result.

The search query correction module may include: a user behavior pattern collection module configured to collect a feedback behavior pattern of the user for the search result; and a query vector correction module configured to generate the final search query vector by correcting the search condition by applying a weight according to the feedback behavior pattern.

The feedback behavior pattern includes the user's staying time in the search result, clicking on a link connected to the search result, adding the search result to a bookmark, scrolling in the search result, and the user's satisfaction with the search result. evaluation may be included.

The query vector correction module may be configured to learn the feedback behavior pattern as a recursive artificial neural network model, determine the weights, apply the weights, correct the search condition, and generate the final search query vector.

The search condition setting module may be configured to select one of colliding user attribute data and company attribute data according to a weight set or corrected according to the feedback behavior pattern when the user attribute data and the company attribute data collide. there is.

The user attribute data may include a company name, job title, position, and type of business, and the company attribute data may include an industry group, region, company size, and number of employees.

A company search method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention includes, by a search condition setting module, a query keyword input by a user to search for a company of interest, user attribute data related to the user, and setting a search condition by extracting a keyword from company attribute data of a company to which the user belongs. searching for the company of interest based on a similarity between a search vector corresponding to the search condition and a company attribute vector, by a search module, and outputting a search result related to the company of interest; and generating, by a search query correction module, a final search query vector by correcting the search condition according to the user's feedback on the search result.

The generating of the final search query vector may include: collecting, by a user behavior pattern collection module, a feedback behavior pattern of the user with respect to the search result; and correcting the search condition by applying a weight according to the feedback behavior pattern by a query vector correction module.

Correcting the search condition may include determining the weight by learning the feedback behavior pattern as a recursive artificial neural network model, and correcting the search condition by applying the weight.

The setting of the search condition may include selecting one of user attribute data and company attribute data colliding with each other according to a weight set or corrected according to the feedback behavior pattern when the user attribute data and the company attribute data collide. can include

According to an embodiment of the present invention, there is provided a computer-readable recording medium on which a program for executing a company search method utilizing the automatic correction of a query vector based on user behavior is recorded.

According to an embodiment of the present invention, a company search system and method using automatic correction of a query vector based on user behavior capable of recommending an appropriate company of interest in consideration of user attributes and company attributes related to the user, in addition to user query keywords, are provided. do.

In addition, according to an embodiment of the present invention, a company search system and method using automatic correction of user behavior-based query vectors capable of predicting user feedback data through a recursive artificial neural network model and predicting user feedback without feedback. is provided.

In addition, according to an embodiment of the present invention, a company search system and method using automatic correction of a query vector based on user behavior, which can be effectively used to sequentially discover customers capable of fulfilling the role of a value chain, are provided. Provided.

1 is a block diagram of a company search system using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.
2 is a block diagram of a search condition setting module constituting an enterprise search system utilizing automatic correction of a query vector based on user behavior according to an embodiment of the present invention.
3 is a block diagram of a search query correction module constituting an enterprise search system using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.
4 is a flowchart of a company search method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.
5 is a flowchart of step S130 of FIG. 4 .
6 and 7 are conceptual diagrams illustrating a company search method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments make the disclosure of the present invention complete, and common knowledge in the art to which the present invention belongs. It is provided to completely inform the person who has the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numbers designate like elements throughout the specification.

In the present specification, when a part 'includes' a certain component, it means that it may further include other components without excluding other components unless otherwise stated. As used herein, '~ module, ~ unit' is a unit that processes at least one function or operation, and may mean, for example, software, FPGA, or hardware component. Functions provided by '~module, ~unit' may be performed separately by a plurality of components or may be integrated with other additional components. '~module, ~unit' in this specification is not necessarily limited to software or hardware, and may be configured to be in an addressable storage medium or configured to reproduce one or more processors. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

A company search system and method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention are designed to solve the inequality phenomenon caused by company size and user's inconvenience. The final search query is created by automatically adjusting the category query vector by analyzing the characteristics and user history.

A company search system and method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention first obtains a query keyword entered by a user to search for a company of interest, user attribute data, and company attribute data related to the user. Extract keywords and set search conditions.

Next, a company search system and method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention is based on a search condition generated from a query keyword, user attribute data, and keywords extracted from the company attribute data. Search for a company of interest based on the similarity between the search vector and the company attribute vector.

Subsequently, a company search system and method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention generates a final search query vector by correcting a search condition according to a user's feedback on a search result related to a company of interest. and correct the search result of the company of interest according to the final search query vector.

1 is a block diagram of a company search system utilizing automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

Referring to FIG. 1 , a company search system 100 using automatic correction of a query vector based on user behavior according to an embodiment of the present invention includes a search condition setting module 110, a search module 120, and a search query correction module 130. ), and the database 140.

The search condition setting module 110 may set search conditions by extracting keywords from a query keyword input by a user, user attribute data related to the user, and company attribute data of a company to which the user belongs to search for a company of interest.

2 is a block diagram of a search condition setting module constituting an enterprise search system utilizing automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

1 and 2, the search condition setting module 110 may include a user query input unit 112, a user property extraction unit 114, a company property extraction unit 116, and a search condition setting unit 118. can

The user query input unit 112 may be configured to receive a query keyword input by the user using the user terminal 10 to search for companies of interest.

The user terminal 10 is a terminal operated or used by a user, and may include a business terminal of a company or a portable terminal of a user.

The user terminal 10 may communicate with the enterprise search system through a wired/wireless communication interface.

The user terminal 10 may include, for example, a smart phone, a smart pad, a desktop PC, a notebook computer, a laptop PC, and the like, but is not limited thereto.

The user terminal 10 may include an input interface (eg, a keyboard, mouse, touch pad, etc.) for inputting a query keyword for searching for a company of interest.

In addition, the user terminal 10 may include a display unit that displays search results of companies of interest on a display screen.

The user attribute extractor 114 may extract user attribute data related to a user (eg, a person in charge of each company) from the database 140 . In an embodiment, the user attribute data may include user attribute information such as a user's profile, interest (preferred) company category, and the like, company name, job title, position, type of business, and the like.

The database 140 may store user attribute information such as a user's profile, interest (preferred) company category, and the like for each user of a company. The user attribute information may be information input by the user using the user terminal 10 .

The company attribute extractor 116 may extract company attribute data from the database 140 . The company attribute data is company attribute data of a company to which the user belongs, and may include, for example, industry group, region, company size, number of employees, and the like.

The database 140 may store not only user attribute information, but also company attribute information such as industry group, region, company size, and number of employees for each company to which the users belong.

The search condition setting unit 118 may set search conditions from query keywords input by the user, user attribute data and company attribute data extracted from the database 140 .

When the user attribute data and company attribute data collide, the search condition setting module 110 may select one of colliding user attribute data and company attribute data according to a weight set or corrected according to a feedback behavior pattern.

For example, in case of conflict between user attribute data and enterprise attribute data, the enterprise attribute data may be selected or attribute data matching the user's feedback behavior pattern may be selected.

The search module 120 may search for a company of interest based on a similarity between a search vector corresponding to a search condition set by the search condition setting unit 118 and a company attribute vector, and output a search result related to the company of interest.

The search module 120 may calculate the similarity between the search vector and the company attribute vector by, for example, cosine similarity, and provide search results in order of high similarity.

The search query correction module 130 may generate a final search query vector by correcting search conditions according to the user's feedback on the search results output by the search module 120 and displayed on the user terminal 10 .

3 is a block diagram of a search query correction module constituting an enterprise search system using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

Referring to FIGS. 1 and 3 , the search query correction module 130 may include a user behavior pattern collection module 132 and a query vector correction module 134 .

The user behavior pattern collection module 132 may collect feedback behavior patterns of the user with respect to search results output by the search module 120 and displayed on the user terminal 10 .

The feedback behavior pattern may include a user's staying time in a search result, clicking on a link within a search result, adding a bookmark to a search result, scrolling in a search result, and evaluating a user's satisfaction with a search result.

The query vector correction module 134 may generate a final search query vector by correcting search conditions by applying weights according to the feedback behavior patterns of the user collected by the user behavior pattern collection module 132 .

The query vector correction module 134 may determine a weight by learning the feedback behavior pattern as a Recurrent Neural Network (RNN) model, correct a search condition by applying the weight, and generate a final search query vector. .

The collected results can be learned through the RNN learning model together with the user's feedback (positive or negative) results.

Accordingly, even if the user rejects the feedback in the future based on the data learned according to the user's behavior pattern, the RNN learning model predicts positive and negative feedback (satisfaction or dissatisfaction) for the behavior of the user who did not give feedback. Probability prediction) can be performed, and information on recommended companies such as business partners can be provided by applying weights when adjusting the user's search query vector based on the predicted results of the feedback.

On the user terminal 10, similar companies matching the query searched by the user may be sequentially displayed in the form of a list in order of highest similarity, and the similarity rate of each company may be displayed on the list.

Conventional search methods derive results only with a set query keyword, but according to an embodiment of the present invention, a more effective search is performed by using the basic characteristics of a user and the attributes of a company connected to a user as query correction data. results can be drawn.

In a state in which the primary search result page is displayed on the screen of the user terminal 10 based on the conditions set by the user to search for business partners, the user among the companies listed on the screen of the user terminal 10 for specific companies When corresponding company information is confirmed by selection (clicking), the search query correction module 130 may extract common attributes of the companies clicked by the user and generate a user preference attribute vector.

In addition, the search query correction module 130 automatically corrects the default value set by the user so that companies satisfying the corresponding attributes are exposed to the user terminal 10 more than the initial search results, and the user's preferred attribute vector and the company A search result recommended to the user may be corrected according to the similarity between attribute vectors and provided to the screen of the user terminal 10 .

According to an embodiment of the present invention, feedback from a user without feedback can be predicted by predicting user feedback data through an RNN model, and it can be applied to produce various results even in a search with limited input keywords.

In the embodiment of the present invention, in providing value to customers, companies can sequentially fulfill the role of the value chain, which means the linkage of a series of activities, functions, and processes related to value creation directly or indirectly. can be used effectively for excavation.

In addition, according to an embodiment of the present invention, it is possible to automatically display companies close to user-set conditions by analyzing user behavioral attribute data, and through user behavior patterns, information about the most important conditions for B2B transactions by industry group is determined. information can be collected.

4 is a flowchart of a company search method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

Referring to FIGS. 1, 2 and 4 , the search condition setting module 110 includes a query keyword input by a user to search for a company of interest, user attribute data related to the user, and company attribute data of a company to which the user belongs. Search conditions can be set by extracting keywords from (S110).

More specifically, the user query input unit 112 may receive a query keyword input by the user using the user terminal 10 to search for companies of interest.

The user attribute extractor 114 may extract user attribute data related to the user from the database 140 (eg, company name, job title, position, type of industry, etc.).

The company attribute extractor 116 may extract company attribute data (eg, industry group, region, company size, number of employees, etc.) related to the company to which the user belongs from the database 140 .

The search condition setting unit 118 may set search conditions from query keywords input by the user, user attribute data and company attribute data extracted from the database 140 .

When the user attribute data and company attribute data collide, the search condition setting module 110 may select one of colliding user attribute data and company attribute data according to a weight set or corrected according to a feedback behavior pattern.

The search module 120 may search for a company of interest based on a similarity between a search vector corresponding to a search condition set by the search condition setting unit 118 and a company attribute vector, and output a search result related to the company of interest ( S120).

The search module 120 may calculate the similarity between the search vector and the company attribute vector by, for example, cosine similarity, and provide search results in order of high similarity.

The search query correction module 130 may generate a final search query vector by correcting search conditions according to the user's feedback on the search results output by the search module 120 and displayed on the user terminal 10 (S130). ).

5 is a flowchart of step S130 of FIG. 4 . 6 and 7 are conceptual diagrams illustrating a company search method using automatic correction of a query vector based on user behavior according to an embodiment of the present invention.

Referring to FIGS. 1, 3 to 7 , the user behavior pattern collection module 132 collects feedback behavior patterns of the user for the search results output by the search module 120 and displayed on the user terminal 10 . It can (S132).

The feedback behavior pattern may include a user's staying time in a search result, clicking on a link within a search result, adding a bookmark to a search result, scrolling in a search result, and evaluating a user's satisfaction with a search result.

The query vector correction module 134 may generate a final search query vector by correcting search conditions by applying weights according to the feedback behavior patterns of the user collected by the user behavior pattern collection module 132 (S134).

The query vector correction module 134 may learn the feedback behavior pattern as a recursive neural network (RNN) model to determine weights, apply the weights to correct search conditions, and generate a final search query vector. .

The collected results can be learned through the RNN learning model together with the user's feedback (positive or negative) results. Accordingly, even if the user rejects the feedback in the future based on the learned data, the RNN learning model predicts positive and negative feedback (predicting the probability of each satisfaction and dissatisfaction) even for the behavior of a user who did not give feedback. Based on the prediction result of this feedback, it can be applied as a weight when adjusting the user's search query vector.

Conventional search methods derive results only with a set query keyword, but according to an embodiment of the present invention, a more effective search is performed by using the basic characteristics of a user and the attributes of a company connected to a user as query correction data. results can be drawn.

In addition, according to an embodiment of the present invention, feedback from a user without feedback can be predicted by predicting user's feedback data through an RNN model, and it can be applied so that various results can be obtained even in a search with limited keywords that can be input. .

The embodiments described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA) array), programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions.

A processing device may run an operating system and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art know that a processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It will be understood that it can include

For example, a processing device may include a plurality of processors or a processor and a controller. Also, other processing configurations are possible, such as a parallel processor. Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device.

Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. Computer readable media may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CDROMs and DVDs, and ROMs, RAMs, and flash memories. hardware devices specially configured to store and execute program instructions, such as; Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or the components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved. Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

10: user terminal
100: Enterprise search system using automatic correction of query vectors based on user behavior
110: search condition setting module
112: user query input unit
114: user attribute extraction unit
116: enterprise attribute extraction unit
118: search condition setting unit
120: search module
130: search query correction module
132: user behavior pattern collection module
134: query vector correction module
140: database

Claims (10)

a search condition setting module configured to set search conditions by extracting keywords from query keywords input by a user, user attribute data related to the user, and enterprise attribute data of a company to which the user belongs to search for a company of interest;
a search module configured to search for the company of interest based on a similarity between a search vector corresponding to the search condition and a company attribute vector, and output a search result related to the company of interest; and
A search query correction module configured to generate a final search query vector by correcting the search condition according to a user's feedback on the search result; According to claim 1,
The search query correction module:
a user behavior pattern collection module configured to collect a feedback behavior pattern of the user for the search result; and
A query vector correction module configured to generate the final search query vector by correcting the search condition by applying a weight according to the feedback behavior pattern; According to claim 2,
The feedback behavior pattern includes the user's staying time in the search result, clicking on a link connected to the search result, adding the search result to a bookmark, scrolling in the search result, and the user's satisfaction with the search result. include evaluation;
The query vector correction module is configured to learn the feedback behavior pattern as a recursive artificial neural network model, determine the weight, apply the weight to correct the search condition, and generate the final search query vector. Enterprise search system utilizing automatic correction of query vectors. According to claim 2,
Wherein the search condition setting module selects one of colliding user attribute data and company attribute data according to a weight set or corrected according to the feedback behavior pattern when the user attribute data and the company attribute data collide, A corporate search system using automatic correction of query vectors based on user behavior. According to claim 4,
The user attribute data includes a company name, job title, position, and industry type, and the company attribute data includes an industry group, region, company size, and number of employees. extracting keywords from query keywords input by a user, user attribute data related to the user, and corporate attribute data of a company to which the user belongs, and setting search conditions, by a search condition setting module; ;
searching for the company of interest based on a similarity between a search vector corresponding to the search condition and a company attribute vector, by a search module, and outputting a search result related to the company of interest; and
A company search method using automatic correction of a query vector based on user behavior, comprising: generating a final search query vector by correcting the search condition according to a user's feedback on the search result, by a search query correction module. According to claim 6,
The step of generating the final search query vector is:
collecting, by a user behavior pattern collection module, a feedback behavior pattern of the user with respect to the search result; and
Correcting the search condition by applying a weight according to the feedback behavior pattern by a query vector correction module; According to claim 7,
The feedback behavior pattern includes the user's staying time in the search result, clicking on a link connected to the search result, adding the search result to a bookmark, scrolling in the search result, and the user's satisfaction with the search result. include evaluation;
The correcting of the search condition includes determining the weight by learning the feedback behavior pattern as a recursive artificial neural network model, and correcting the search condition by applying the weight, and automatically correcting the query vector based on user behavior. Company search method using . According to claim 7,
The setting of the search condition may include selecting one of user attribute data and company attribute data colliding with each other according to a weight set or corrected according to the feedback behavior pattern when the user attribute data and the company attribute data collide. including,
The user attribute data includes a company name, job title, position, and industry type, and the company attribute data includes an industry group, region, company size, and number of employees. A computer-readable recording medium in which a program for executing the company search method using the automatic correction of a query vector based on user behavior according to any one of claims 6 to 9 is recorded thereon.

Images