KR101647087B1 - Server and method for goods providing based on natural language processing - Google Patents

Abstract

The present invention relates to a server and a method for providing goods based on natural language processing, wherein the server for providing goods based on natural language processing analyzes a review on goods, written by a user terminal, based on the natural language processing and provides the goods suitable to a user writing the review. The server for providing goods based on natural language processing includes: a communication module; a memory storing a program providing the goods for the user terminal; and a processor executing the program stored in the memory. The processor receives the review written by the user terminal from one or more review providing servers based on the communication module according to the execution of the program and calculates a score of a keyword extracted based on an emotional quotient and a weight value on the extracted keyword by extracting the keyword included in the review based on a morpheme analyzer. The processor also extracts goods information recommended for the user terminal among one or more pieces of the goods information based on the calculated score of the keyword and transmits the extracted goods information to the user terminal. The weight value is calculated based on term frequency-inverse document frequency, and the emotional quotient is calculated based on polarity information of a sentence including the keyword.

Description

TECHNICAL FIELD [0001] The present invention relates to a server and a method for providing goods based on natural language processing,

The present invention relates to a goods provision recommendation server and method based on natural language processing.

A recommendation service is a service that selectively provides information or goods that meet the user's taste or interest from excessively provided information or goods. These recommendation services are applied to electronic commerce, Internet shopping malls, and the like.

The recommendation service is a collaborative filtering based recommendation that recommends a good based on a purchase pattern for a user similar to a specific user, recommends a good of a purchase pattern similar to a purchased good, And a content based recommendation that recommends a good based on the content information of the good.

Content-based recommendation is a recommendation method based on information retrieval. It is recommended for shopping malls such as books, music, and clothing because it can recommend new goods based on metadata of purchased goods. In the case of the content-based recommendation in the shopping mall, the keyword may be extracted from reviews of the purchased user based on Opinion mining, and the extracted keyword may be used as content information.

At this time, Opinion Mining analyzes the natural language sentences included in the unstructured data document to understand the meaning and analyze the polarity. It is mainly used for analyzing user's opinions on goods or services Is a kind of text mining technique.

Korean Patent Laid-Open Publication No. 10-2014-0113771 (titled " M2M-based book recommendation method and system ") discloses a method and system for recommending a good M2M product, such as a smart phone, a tablet, machine-to-machine (UI) method and system capable of providing a personalized book recommendation service to a user of a user terminal. Specifically, the method classifies a plurality of community groups based on personal information of a plurality of mobile device users, determines a book recommendation list for each community group, and transmits the list to the user terminal.

Korean Patent Laid-Open Publication No. 10-2014-0142810 (titled invention: method and server for providing recommended product information, and mobile device) receives a connection start request including device information of a mobile device from a mobile device, Discloses a method and server for searching for a recommended product and providing it to a mobile device. Specifically, the method provides a recommendation product corresponding to the mobile device based on information of the mobile device such as a manufacturer, a mobile phone number, and model information.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems of the prior art, and it is an object of the present invention to provide a server and a method for analyzing a review on goods created through a user terminal based on natural language processing, The purpose is to provide.

It should be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

As a technical means for achieving the above technical object, a natural language processing-based goods providing server according to the first aspect of the present invention includes a communication module, a memory storing a program for providing goods to the user terminal, The processor receives a review made by the user terminal from one or more review providing servers based on the communication module, extracts the keywords included in the review based on the morpheme analyzer, Extracts the goods information to be recommended to the user terminal from among one or more goods information based on the score of the calculated keyword, and outputs the extracted goods information to the user terminal send. At this time, the weight is calculated based on the term frequency-inverse document frequency, and the emotion index is calculated based on the polarity information of the sentence including the keyword.

According to a second aspect of the present invention, there is provided a method for providing information on natural language processing in a server, the method comprising: receiving a review created by a user terminal from one or more review providing servers; Extracting keywords included in the review based on the morpheme analyzer; Calculating a score of the extracted keyword based on the weight value and the emotion index of the extracted keyword; Extracting a piece of good information to be recommended to the user terminal from among one or more pieces of good information based on the score of the calculated keyword; And transmitting the extracted goods information to the user terminal. At this time, the weight is calculated based on the term frequency-inverse document frequency, and the emotion index is calculated based on the polarity information of the sentence including the keyword.

According to an embodiment of the present invention, an embodiment of the present invention can quickly and accurately extract user's preference information based on the emotion index of a keyword and a keyword included in a user review, .

In addition, since the embodiment of the present invention uses a morpheme analyzer that is updated based on machine learning, the accuracy of preference information extraction can be gradually improved.

According to an embodiment of the present invention, since the merchandise seller can integrally manage reviews of a plurality of review providing servers, it is possible to provide a high quality service to the user at a low cost, thereby enabling the satisfaction of the user and the sales volume of goods Can be improved.

1 is a service conceptual diagram of a goods providing server according to an embodiment of the present invention.
2 is a configuration diagram of a goods providing server according to an embodiment of the present invention.
3 is an exemplary diagram of a natural language processing process according to an embodiment of the present invention.
4 is a flowchart of a method for providing goods according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly explain the present invention in the drawings, parts not related to the description are omitted.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when a part is referred to as "including " an element, it does not exclude other elements unless specifically stated otherwise.

Hereinafter, a goods providing service of the goods providing server 100 according to an embodiment of the present invention will be described with reference to FIG.

1 is a service conceptual diagram of a goods providing server 100 according to an embodiment of the present invention.

The goods providing server 100 according to the embodiment of the present invention analyzes the review of the user who wrote the review and provides the recommended goods. Specifically, the user creates a review of the goods using the user terminal 110, and transmits the created review to the review providing server 120. The goods providing server 100 analyzes one or more reviews of the user uploaded to the review providing server 120 based on the natural language processing using the morpheme analyzer, and analyzes the goods provided by the analyzed review and goods metadata providing server 130 Based on the meta data of the user.

At this time, the recommended goods provided by the goods providing server 100 can be directly transmitted from the goods providing server 100 to the user terminal 110. In addition, the recommended goods provided to the user who wrote the review may be transmitted from the goods providing server 100 to the review providing server 120 and transmitted to the user terminal 110 via the review providing server 120.

In yet another embodiment of the present invention, the goods providing server 100 may be a natural language processing engine server and a goods matching server. At this time, the natural language processing engine server analyzes the user's review and extracts a keyword preferred by the user. Further, the goods matching server can extract and provide the goods to be recommended to the user based on the keyword extracted through the natural language processing engine server.

Meanwhile, the goods providing server 100 according to an embodiment of the present invention may receive a review of goods from one or more review providing servers 120. [

At this time, the review providing server 120 may be a shopping mall selling goods such as books, records, clothes, and household appliances. In addition, the review providing server 120 provides information on the prices of goods such as books, records, movies, clothes, and household appliances, information on goods such as specifications and manufacturers, and a price An information site, a portal site, and the like. In addition, the review providing server 120 may collect reviews of one or more external review providing sites and provide them to the goods providing server 100, or may provide users with information such as external web pages, blogs, and social media And may be a review collection server that collects reviews of the posted goods and provides the goods to the goods providing server 100.

The review providing server 120 may further include obtaining login information and a review collection agreement of a user who has made a review in advance to provide a review to the goods providing server 100. [ Accordingly, the review providing server 120 can provide a review to the goods providing server 100 only on the user who accepted the review gathering agreement, in accordance with the login information previously stored.

The goods metadata providing server 130 is a server for providing mate data such as name, manufacturer, and genre of goods. For example, if the goods metadata providing server 130 is a book metadata providing server, it can provide title, publisher, author, translator, genre, price and ISBN information of the book. When the goods metadata providing server 130 is a movie metadata providing server, it may provide information such as director, actor, producer, distributor, genre, and release date. At this time, the good metadata providing server 130 may be the same server as the review providing server 120. Also, the goods metadata providing server 130 may be a portal site for providing the goods metadata.

In yet another embodiment, the good metadata providing server 130 may include a good describing server, a keyword storing server, and a good keyword storing server. At this time, the goods description server stores the metadata about the goods to be analyzed in the review, the keyword storage server stores the keywords and the metadata included in the review, and the goods keyword storage server stores the metadata about the goods Keywords and metadata.

The user terminal 110 may be a terminal that purchases goods through the goods providing server 100. Also, the user terminal 110 may be a terminal that transmits a review made by the user to the goods to the review providing server 120. Specifically, the user terminal 120 may login to the review providing server 120 and write a review of the purchased or used goods by the user. In addition, the user terminal 120 may store review consent and login information in the goods providing server 100 or the review providing server 120 for review analysis and provision of goods.

At this time, the user terminal 110 includes a display module, a memory, a processor, a communication module, and the like, and collectively refers to all terminals that can use the network. For example, the user terminal 110 may be a general computing system, such as a personal computer (PC), a server computer.

The user terminal 110 may be a wireless communication device, such as a PCS (Personal Communication System), a GSM (Global System for Mobile communications), a PDC (Personal Digital Cellular), a PHS (Personal Digital Assistant), IMT (International Mobile Telecommunication) -2000, Code Division Multiple Access (CDMA) -2000, W-Code Division Multiple Access (W-CDMA), WiBro (Wireless Broadband Internet) A handheld-based wireless communication device, and the like.

The user terminal 110 may include a personal computer capable of wireless communication such as a notebook computer, a tablet PC, etc. and a smart device such as a smart phone. have.

Meanwhile, the goods providing server 100, the user terminal 110, the review providing server 120, and the goods metadata providing server 130 may be connected through a network. In this case, the network means a connection structure in which information can be exchanged between each node such as a terminal and a server. An example of such a network is a 3rd generation partnership project (3GPP) network, a long term evolution (LTE) network, a WIMAX world interoperability for microwave access networks, internet, local area network (LAN), wireless local area network (WLAN), wide area network (WAN), personal area network (PAN), bluetooth network, Broadcast networks, analog broadcast networks, digital multimedia broadcasting (DMB) networks, and the like.

The goods providing server 100 may also be a server program that runs on an independent server computer or provides the recommendation goods to the review providing server 120 or the goods metadata providing server 130, And can be serviced. The goods providing server 100 may be installed as a server program in a goods selling server different from the review providing server 120 and the goods metadata providing server 130 and may be serviced.

Next, a goods providing server 100 according to an embodiment of the present invention will be described with reference to FIG. 2 and FIG.

2 is a configuration diagram of a goods providing server 100 according to an embodiment of the present invention. 3 is an exemplary diagram of a natural language processing process according to an embodiment of the present invention.

Referring to FIG. 2, the goods providing server 100 includes a communication module 210, a memory 220, and a processor 230. In addition, the goods providing server 100 may include a goods information database 240.

The communication module 210 receives a review of the goods that the user has uploaded to the review providing server 120 through the user terminal 110. The communication module 210 receives the metadata of the goods through the goods metadata providing server 130. In addition, the communication module 210 provides a recommendable item to the user extracted through the processor 230. At this time, the communication module 210 may include both a wired communication module and a wireless communication module.

In the memory 220, a program for providing goods to the user terminal 110 is stored. Here, the memory 220 is collectively referred to as a non-volatile storage device that keeps stored information even when power is not supplied. The program stored in the memory 220 may be implemented in hardware such as software or an FPGA (Field Programmable Gate Array) or ASIC (Application Specific Integrated Circuit), and may perform predetermined roles.

The goods information database 240 may store the metadata corresponding to the goods information. At this time, the metadata corresponding to the goods information may be meta data such as the name of the goods, the manufacturer, and the genre provided by the goods metadata providing server 130, and may be additionally created, Metadata.

The processor 230 according to an exemplary embodiment of the present invention may be configured to allow a user to receive user recommendation goods from one or more review server 120 through the communication module 230 to use the user terminal 110 I receive a review. At this time, the review may be composed of one or more sentences as shown in FIG. 3 (a).

In addition, the processor 230 extracts keywords included in the review received from the review providing server 120 based on a morphological analyzer.

The morpheme analyzer according to an embodiment of the present invention is a tool for analyzing morphemes of words constituting a natural language sentence or a document. Specifically, the morpheme analyzer parses the sentences included in the review, extracts the phrases, and analyzes the morphemes of the extracted phrases through the classification analysis to extract the necessary morpheme words.

For example, referring to FIG. 3, the first sentence in the review document as shown in FIG. 3 (a), "<Going to Barcelona in search of happiness and happiness> Quot; can be analyzed as shown in FIG. 3 (b).

In addition, the morpheme analyzer may include a grammar dictionary database for expressing connection relationships between word-based morphemes and words in order to analyze the syntax and morpheme of each word. At this time, the morpheme analyzer can update the grammar dictionary database based on the machine learning in order to quickly and accurately analyze the morpheme of the review, the analysis of the words frequently used in the review, and the morpheme analysis.

For example, the morpheme analyzer calculates the score according to the frequency of the grammar used, and calculates the grammar dictionary database based on the calculated score and the machine learning so that the higher- Can be automatically corrected.

The processor 230 may perform preprocessing to process stop words in the analyzed word if the morpheme of the word is analyzed by the morpheme analyzer. At this time, an abbreviation refers to a word that appears at a high frequency in a document or a sentence but has no special meaning. For example, in a review, an idiom can be a pronoun such as "I", "it", "me", and an investigation such as "silver", "silver", "is" In addition, abbreviations in the review can be words that occur at high frequency in the field related to goods. For example, in a review of a book, words such as "book" and "author" can be abbreviated.

The processor 230 extracts words necessary for providing goods based on the morpheme for the preprocessed word as keywords. For example, in FIG. 3 (b), the processor 230 extracts the words "Ogi", "Barcelona", "architect", "Oh Young-wook", "Spain" .

In addition, the processor 230 may extract a word including consecutive nouns among the words extracted through the morpheme analyzer as one keyword. For example, the processor 230 may extract a single keyword if the extracted word is a compound noun or a contiguous noun with spaces.

In addition, the processor 230 may extract the name, manufacturer, etc. of the goods among the analyzed words based on the metadata stored in the goods information database 240 as one keyword. For example, in FIG. 3 (b), if the word "Ogata, leaving happily for Barcelona" is present in the goods information database 240 as the name of the goods, the processor 230 reads " Quot; leave "can be extracted as one keyword.

Meanwhile, the processor 230 may extract keywords from the metadata stored in the form of a natural language sentence or phrase in the goods information database 240 through the same process as the process of extracting the keywords included in the review, and utilize the keywords as goods information. Specifically, the processor 230 may extract a keyword based on the morpheme analyzer with respect to the metadata stored in the sentence or sphere stored in the goods information database 240. The processor 230 may extract the keyword included in the review by utilizing the keyword extracted from the metadata. In addition, the processor 230 may store the keyword extracted from the metadata again in the goods information database 240, and utilize the keyword as metadata.

When the keyword is extracted, the processor 230 calculates the score of the extracted keyword based on the weight for the extracted keyword and the emotion index.

At this time, the weight of the keyword extracted from the review is calculated based on the term frequency-inverse document frequency (TF-IDF). Word frequency - Inverse document frequency weighting is a representative word weight measurement method of the vector space model that is mainly used in the field of information retrieval. For example, the word frequency - inverse document frequency can be calculated as shown in equation (1).

는 가중치를 구하는 단어,

는 단어

의 가중치이며,

는 가중치를 구하고자 하는 리뷰에 포함된

의 빈도수,

는 전체 리뷰 중

를 포함하는 리뷰의 빈도수이다. At this time, in Equation (1)

Is a word for obtaining a weight,

Word

&Lt; / RTI &gt;

Is included in the review for which we want to calculate weights.

Frequency,

In the full review

Is the frequency of reviews that include.

는 4가 될 수 있다. 만약 전체 리뷰 중 "바로셀로나"라는 키워드를 포함하는 리뷰가 40건이라면

는 40이 되며, 가중치

는 0.1이 될 수 있다. Referring to FIG. 3 (a), in the case of the keyword "barcelona", the keyword appears four times in FIG. 3 (a)

Can be four. If there are 40 reviews in the entire review that contain the keyword "Barcelona"

Is 40, and the weight

Can be 0.1.

In addition, the emotion index is calculated based on the polarity information of the sentence including the keyword. The processor 230 may determine based on the morpheme included in the sentence that the sentence includes a meaning such as positive, negative, and neutral. The processor 230 may set the emotion index to a positive number if it is affirmative based on the discriminated polarity information, and may calculate the emotion index by setting the emotion index to a negative value if it is negative.

For example, the word " I want to see "in the phrase " I would like to go to Barcelona once again" included in FIG. 3 (a) is an expression of affirmation. Therefore, the processor 230 may determine that the polarity information of the sentence containing the keyword "barcelona" is affirmative and calculate the emotion index of "barcelona "

The processor 230 calculates the score of the keyword based on the calculated weight and the emotion index when the weight and emotion index of the extracted keyword are calculated. For example, the processor 230 may calculate 0.2, which is the product of the two values, as a score for the extracted keyword, if the calculated weight of the extracted keyword " barcelona " is 0.1 and the emotion index is 1.

The processor 230 may further extract a keyword based on an ontology dictionary included in the goods information database 240. [

At this time, the ontology dictionary is a dictionary which defines a conceptual classification of keywords and a relation according to causal relationships. For example, if the word "traveler" is extracted as the keyword and the upper word of the "traveler" is "travel" in the ontology dictionary, the processor 230 can extract "travel" At this time, the score of "travel" can be the maximum value of the score of the sub-words of "travel" in the ontology dictionary.

The processor 230 extracts a keyword, calculates a score of each keyword, extracts a keyword having a preset threshold value or more, or extracts a predetermined number of keywords as a user's favorite keyword to be provided with a recommended product.

Referring to FIG. 3, (c) in FIG. 3 is a result of extracting a keyword from the morphological analysis result and calculating a score. If the preset threshold value is 0.1, the keywords preferred by the user to receive the recommended goods may be "Barcelona" and "Travel ".

The processor 230 may extract the goods information to be recommended to the user among the goods information as one or more recommendable objects based on the scores of the extracted preferred keywords and preferred keywords.

At this time, the goods information to be recommended may be information included in the list of goods purchased by the user or not included in the used history among the goods included in the goods information database 240. The recommended product information may be information included in the list of goods provided by the review providing server 120 or the third goods selling server such as a new product list, but is not limited thereto.

The processor 230 may rank the one or more goods to be provided to the user based on the degree of matching between the extracted preferred keyword and the goods information. For example, if the information included in the first item to be referenced is "Travel", "Barcelona" and "Architecture", and the information contained in the second item is "Barcelona" and " The user's favorite keywords are matched with two, and the second good is matched with one user's preferred keyword. Thus, the processor 230 may rank in order of first matching goods and second goods.

In addition, the processor 230 can rank the goods to be provided to the user based on the degree of matching between the extracted preferred keyword and the goods information and the score of each extracted preferred keyword. In the above example, the first goods match two user-preferred keywords. Therefore, referring to FIG. 3 (c), since the score of two matching keywords is 0.3436 and 0.1250, the preference of the first good can be 0.2343, which is the average of the two values. In addition, since the second favorite matches one keyword that the user prefers, and the score of the matching keyword is 0.3436, the preference of the second good may be 0.3436. Thus, the processor 230 may rank in order of preference second goods and first goods.

After ranking the goods to be provided to the user in this way, the processor 230 can extract the goods information to be recommended to the user among the one or more goods information to be recommended, according to a predetermined number or a predetermined score. For example, if the predetermined number is n, the processor 230 may provide the user with the highest ranked n items of the one or more goods to be presented to the user. If n is 2, then the processor 230 may provide the user with two more ranked items. Further, if the predetermined score is x or more, the user can be provided with the goods whose preference of the calculated goods is x or more. If the predetermined score is 0.3, it is possible to provide the user with a preference of 0.3 or more.

On the other hand, the processor 230 can extract the goods to be provided to the user based on the extracted keyword and analysis of association with one or more goods information. At this time, the association can be extracted through association analysis. At this time, association analysis is a typical analysis algorithm of data mining, and it is a method to find hidden rules for purchasing or using goods based on the concurrence existing in users' purchase history of goods.

The processor 230 may derive an association hidden in the keyword based on the association analysis and extract the goods to be provided to the user through the association. In addition, the processor 230 may generate a network between the keywords using the correlation coefficient, support, and confidence derived from the association analysis as the connection strength, and provide the generated network to the user . For example, the processor 230 may generate a network of keywords and provide the user with a visualization result in the form of a radial graph.

Further, through the association analysis, the processor 230 calculates the correlation coefficient of "(Barcelona, travel)" to 0.8, the correlation coefficient of "(Barcelona, 0.1. &Lt; / RTI &gt; At this time, the correlation coefficient has a strong positive correlation with the closer to 1, and a stronger negative correlation with the closer to -1. In addition, the correlation coefficient between the two keywords decreases as the correlation coefficient approaches zero. Therefore, in the example, "Barcelona" and "travel" have a strong positive correlation, so the two keywords have a high probability of coincidence. In addition, the strong correlation between "Barcelona" and "picture" shows a strong negative correlation, so that the probability of two keywords occurring at the same time is low. And "Barcelona" and "Architecture" have little correlation with each other.

When the goods information to be provided to the user is extracted in this manner, the processor 230 transmits the goods information extracted through the communication module 210 to the user terminal 110. At this time, the processor 230 can directly transmit the extracted goods information through the communication module 210 to the user terminal 110. The processor 230 transfers the extracted goods information to the review providing server 120 corresponding to the extracted goods information among the one or more review providing servers 120 through the communication module 210, The received review providing server 120 can transmit the extracted goods information to the user terminal 110. [

Next, referring to FIG. 4, a description will be made of a method for providing goods information based on natural language processing in the goods providing server 100 according to an embodiment of the present invention.

In step S400, the goods providing server 100 receives a review of the goods purchased by the user created by the user terminal 110 from the one or more review providing servers 120 or the used goods.

Then, the goods providing server 100 extracts keywords included in the review based on the morpheme analyzer (S410). At this time, the morpheme analyzer may include a grammar dictionary database that is updated based on the machine learning in order to quickly and accurately analyze the processing of the words frequently included in the review, according to the morphological analysis result.

In addition, if the product providing server 100 includes a consecutive noun among the words extracted through the morpheme analyzer, the product providing server 100 may extract the consecutive nouns as one keyword. In addition, the goods providing server 100 may extract words included in the goods information metadata stored in the goods information database 240 among the extracted words as keywords. The goods providing server 100 can further extract the keyword based on the ontology dictionary defining the concept classification and association of the keyword.

The goods providing server 100 calculates the score of the extracted keyword based on the weight and the emotion index of the extracted keyword (S420). At this time, the weight is calculated based on the term frequency-inverse document frequency. In addition, the emotion index is calculated based on polarity information such as positive, negative, and neutral of the sentence including the keyword.

The goods providing server 100 calculates the score for the keyword, and then extracts the goods information to be recommended to the user terminal 110 from among one or more goods information based on the calculated score of the keyword (S430). At this time, the goods providing server 100 can extract the goods to be provided to the user based on the degree of matching between the extracted keyword and the goods information.

Then, the goods providing server 100 transmits the extracted goods information to the user terminal 110 (S440). At this time, the goods providing server 100 can deliver the extracted goods information to the user terminal 110 directly. In addition, the goods providing server 100 may transmit the goods information extracted by the review providing server 120 corresponding to the extracted goods information to the user terminal 110.

The method of providing goods can be performed in a stand-alone manner on an independent server computer. In addition, the goods providing method may be implemented as an application programming interface (API), and may be installed and serviced as a program for providing recommendation goods to the user to the review providing server 120 or the goods metadata providing server 130.

The natural language processing-based goods providing server 100 and method according to an embodiment of the present invention can quickly and accurately extract the user's preference information based on the emotion index of the keywords and keywords included in the user review, . In addition, since the goods providing server 100 and the method use a morpheme analyzer that is updated based on machine learning, the accuracy of preference information extraction can be gradually improved.

The goods providing server 100 and the method can provide a good quality service to the user at a low cost since the goods seller can collectively manage reviews of the plurality of review providing servers 120, Satisfaction and merchandise sales volume can be improved.

One embodiment of the present invention may also be embodied in the form of a recording medium including instructions executable by a computer, such as program modules, being executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium can include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes any information delivery media, including computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transport mechanism.

While the methods and systems of the present invention have been described in connection with specific embodiments, some or all of those elements or operations may be implemented using a computer system having a general purpose hardware architecture.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

110: User terminal 100: Goods providing server
120: review providing server 130: goods metadata providing server

Claims (9)

A goods providing server based on natural language processing,
At least one review providing server and a communication module for performing data communication with the user terminal,
A goods information database,
A memory for storing a program for extracting product information to be recommended to the user terminal,
And a processor for executing a program stored in the memory,
The processor extracts one or more keywords included in the review received from the review providing server based on the morpheme analyzer,
Analyzing the polarity information of the review based on the end of the review, calculating an emotion index for the review using the analyzed polarity information,
Calculating a score for the extracted one or more keywords based on the weight for the extracted one or more keywords and the calculated emotion index,
Selecting one or more goods information matching the extracted one or more keywords among the one or more goods information stored in the goods information database based on the extracted relation of one or more keywords, Calculates a score for the selected good,
Extracting the goods information to be recommended to the user terminal in descending order of the score of the calculated goods information,
Generating a network between the keywords based on the association relation of the extracted one or more keywords,
Transmitting the network between the generated keywords and the extracted goods information to the user terminal,
Wherein the review is generated by the user terminal and then stored in the review providing server,
Wherein the weight of the extracted one or more keywords is calculated by using a weight for each keyword included in the extracted one or more keywords,
The weight for each of the keywords is calculated based on the number of keywords included in the review and the number of reviews including the keyword to calculate the weight among the one or more reviews included in the review providing server ,
Wherein the association relation for the extracted one or more keywords is preliminarily generated on the basis of an association analysis for one or more goods information stored in the goods information database,
Wherein the network between the keywords is generated based on a connection strength between the at least one keyword,
Wherein the connection strength between the at least one keyword is one of a correlation coefficient, a support degree and a reliability among the one or more keywords calculated based on the association analysis,
The polarity information is any one of positive, negative and neutral,
The emotion index for the review is set to a preset positive value if the end of the review is positive and to a negative value if the end of the review is negative and to 0 if the end of the review is neutral A goods offering server, The method according to claim 1,
Wherein the processor transmits the extracted goods information to the review providing server corresponding to the extracted goods information among the one or more review providing servers through the communication module and delivers the extracted goods information to the user terminal. The method according to claim 1,
The morpheme analyzer includes a grammar dictionary database,
Wherein the grammar dictionary database is updated based on machine learning. The method according to claim 1,
Wherein the processor extracts words containing consecutive nouns of the words extracted through the morpheme analyzer with one keyword. The method according to claim 1,
Wherein the goods information database stores metadata corresponding to the one or more goods information,
Wherein the processor selects the goods information matching the extracted keyword based on the metadata corresponding to the stored one or more goods information. 6. The method of claim 5,
Wherein the processor extracts one or more words included in the review based on the morpheme analyzer, extracts one or more keywords of the extracted one or more words based on the meta data and the ontology dictionary, The server provides a score for the goods. The method according to claim 1,
Wherein the processor calculates a score for the selected goods information based on the number of matched goods information between the extracted one or more keywords and the one or more goods information. delete A method for providing goods information based on natural language processing in a goods providing server,
Extracting one or more keywords included in the review received from at least one review providing server based on the morpheme analyzer;
Analyzing the polarity information of the review based on the end of the review, and calculating an emotion index for the review using the analyzed polarity information;
Calculating a score for the extracted one or more keywords based on the weight for the extracted one or more keywords and the calculated emotion index;
Selecting one or more goods information items matching the extracted one or more goods information items from among one or more goods information items stored in the goods information database based on the extracted relation of one or more keywords, Calculating a score for the selected good;
Extracting the goods information to be recommended to the user terminal in descending order of the score of the calculated goods information;
Generating a network between one or more keywords based on an association relation of the extracted one or more keywords; And
And transmitting the network of generated keywords and the extracted goods information to the user terminal,
Wherein the review is generated by the user terminal and then stored in the review providing server,
Wherein the weight of the extracted one or more keywords is calculated by using a weight for each keyword included in the extracted one or more keywords,
The weight for each of the keywords is calculated based on the number of keywords included in the review and the number of reviews including the keyword to calculate the weight among the one or more reviews included in the review providing server ,
Wherein the association relation for the extracted one or more keywords is preliminarily generated on the basis of an association analysis for one or more goods information stored in the goods information database,
Wherein the network between the keywords is generated based on a connection strength between the at least one keyword,
Wherein the connection strength between the at least one keyword is one of a correlation coefficient, a support degree and a reliability among the one or more keywords calculated based on the association analysis,
The polarity information is any one of positive, negative and neutral,
The emotion index for the review is set to a preset positive value if the end of the review is positive and to a negative value if the end of the review is negative and to 0 if the end of the review is neutral A method of providing goods.

Images