JP4743766B2 - Impression determination system, advertisement article generation system, impression determination method, advertisement article generation method, impression determination program, and advertisement article generation program - Google Patents

Description

  The present invention relates to an impression determination system, an advertisement article generation system, an impression determination method, an advertisement article generation method, an impression determination program, and an advertisement article generation program used in an affiliate system.

  In recent years, advertising business using the Internet has attracted attention. In such a field, a technique for displaying an advertisement corresponding to a phrase that a user has executed a search on a search site on the Internet has been proposed (for example, see Patent Document 1).

The advertisement providing device described in Patent Document 1 has a search database that stores advertisement information and corresponding words in association with each other and stores corresponding words and related words related to the corresponding words in association with each other. is doing. Then, the advertisement providing device searches the phrase that has received the input as a related phrase, obtains a corresponding phrase corresponding to the related phrase, further acquires the advertisement information in association with the corresponding phrase, and provides the advertisement information. . The advertisement providing apparatus configured in this way makes it possible to provide advertisement information even when a related word / phrase is input.
JP 2002-297621 A

  As described above, a technique for displaying an advertisement corresponding to a phrase searched by a user on a search site is already known, and is related to a search phrase in that the user is interested in the search phrase. It has the effect of displaying advertisements. However, the user does not always have a good feeling for the search phrase, and may have a bad feeling. Therefore, an advertisement that matches the user's emotion is made only by displaying an advertisement related to it based only on the search phrase, and an advertisement effect is not always produced.

  The present invention has been made in view of such circumstances, and an impression determination system, an advertisement article generation system, an impression determination method, an advertisement, and an advertisement effect are enhanced by determining an impression on a keyword extracted from document information. An object is to provide an article generation method, an impression determination program, and an advertisement article generation program.

  (1) In order to achieve the above object, an impression determination system according to the present invention includes a document information acquisition unit that acquires document information of a Web site from a Web server that enables browsing of the Web site, and the document information. A word analysis unit that decomposes into words by morphological analysis; a keyword extraction unit that extracts a phrase that meets a predetermined condition as a keyword from the set of decomposed words; and the keyword in the acquired document information A related word extraction unit that extracts a word used in relation to a word as a related word, and a word that is easily used together with the specific word in association with the specific word And storing impression information for each registered word and storing the degree of positive or negative impression for the specific phrase And an impression determination unit that refers to the impression information storage unit, comprehensively calculates a degree of impression with respect to the keyword for the related word, and determines an impression with respect to the keyword obtained from the document information. It is said.

  As described above, the impression determination system of the present invention not only extracts keywords from a document but also determines impressions for the extracted keywords. Available for provision. As a result, the advertising effect can be enhanced.

  (2) Further, the advertisement article generation system of the present invention refers to the impression determination system described above, a group word storage unit that stores positive and negative word groups for a specific phrase, and the group word storage unit, respectively. A group word extraction unit that extracts a plurality of words that are likely to be used together with the keyword with the same impression as the determined impression, and an advertisement information storage unit that stores advertisement material information for each advertisement. An advertisement identifying unit that identifies a matched advertisement by matching the group word and information obtained from the advertisement material information, and an advertisement that extracts advertisement material information corresponding to the identified advertisement from the advertisement information storage unit An advertisement section information is generated using the extraction unit and the extracted advertisement material information, and the generated advertisement article information It is characterized in that and a advertisement articles generating unit that transmits to the Web server that obtains the document information.

  Thus, the advertisement article generation system of the present invention generates advertisement article information close to the impression received from the document. It is possible to generate an advertising article that matches the impression of the document with respect to the central topic. As a result, the user's willingness to purchase can be further enhanced by the advertisement, and the advertising effect can be enhanced.

  (3) The impression determination method of the present invention includes a document information acquisition step of acquiring document information of a website from a web server that enables browsing of the website, and decomposes the document information into words by morphological analysis. Used in relation to the keyword in the acquired document information, a word analyzing step for extracting a phrase that matches a predetermined condition as a keyword from the set of decomposed words, A related word extracting step of extracting from the set of decomposed words as a related word, a word that is easily used together with the specific word in association with the specific word, and stored as a registered word, and Impression information storage that memorizes the degree of positive or negative impression for the specific phrase for each registered word Refers to the, the associated word comprehensively calculating the degree of impression of keywords for, is characterized in that it comprises, and impression judgment step of judging impression of the keyword obtained from the document information.

  In this way, the impression determination method of the present invention not only extracts keywords from a document but also determines impressions for the extracted keywords, and therefore uses the determination result to display an advertisement that matches the impression of the keywords received from the document. Available for provision. As a result, the advertising effect can be enhanced.

  (4) Further, the advertisement article generation method of the present invention refers to an impression determined by the above-described impression determination method with reference to a group word storage unit that stores positive and negative word groups for a specific phrase. A group word extraction step for extracting a plurality of words that are likely to be used together with the keyword with the same impression as a group word, and an advertisement information storage unit that stores advertisement material information for each advertisement, and the group word and the advertisement material An advertisement identifying step for identifying a matched advertisement by matching information obtained from information, an advertisement extracting step for extracting advertisement material information corresponding to the identified advertisement from the advertisement information storage unit, and the extracted Advertising article information is generated using the advertising material information, and the generated advertising article information is converted into the document information. It is characterized in that it comprises acquired advertisement articles generating step of transmitting to the Web server has, and.

  Thus, the advertisement article generation method of the present invention generates advertisement article information close to the impression received from a document. It is possible to generate an advertising article that matches the impression of the document with respect to the central topic. As a result, the user's willingness to purchase can be further enhanced by the advertisement, and the advertising effect can be enhanced.

  (5) In addition, the impression determination program of the present invention includes a document information acquisition process for acquiring document information of a Web site from a Web server that enables browsing of the Web site, and decomposes the document information into words by morphological analysis. Used in relation to the keyword in the acquired document information, and a word extraction process that extracts a phrase that matches a predetermined condition as a keyword from the set of decomposed words. A related word extraction process for extracting a word or phrase as a related word, a word that is easily used together with the specific word in association with the specific word, and storing as a registered word; and For each registered word, refer to the impression information storage unit that stores the degree of positive or negative impression for the specific phrase. The associated word comprehensively calculating the degree of impression of keywords for, is characterized in that it comprises a determining impression determination processing an impression for the keyword obtained from the document information.

  As described above, the impression determination program according to the present invention not only extracts keywords from a document but also determines impressions for the extracted keywords. Available for provision. As a result, the advertising effect can be enhanced.

  (6) The advertisement article generation program of the present invention refers to the keyword determined by the impression determination method described above with reference to a group word storage unit that stores positive and negative word groups for a specific phrase. A group word extraction process for extracting a plurality of words that are easy to use with the keyword as a group word, and an advertisement information storage unit that stores advertisement material information for each advertisement. An advertisement specifying process for identifying a matched advertisement by matching information obtained from the advertisement material information, an advertisement extracting process for extracting advertisement material information corresponding to the specified advertisement from the advertisement information storage unit, Using the extracted advertisement material information to generate advertisement article information, the generated advertisement article information Is characterized in that it comprises, an advertising article generation processing to be sent to the Web server that obtains the document information.

  Thus, the advertisement article generation program of the present invention generates advertisement article information close to the impression received from a document. It is possible to generate an advertising article that matches the impression of the document with respect to the central topic. As a result, the user's willingness to purchase can be further enhanced by the advertisement, and the advertising effect can be enhanced.

  According to the impression determination system of the present invention, not only a keyword is extracted from a document but also an impression of the extracted keyword is determined, so that an advertisement that matches the impression of the keyword received from the document can be provided using the determination result. Be available. As a result, the advertising effect can be enhanced.

  In providing the use of the affiliate system, the present inventors pay attention to the fact that advertising that matches the text posted on the website has a higher advertising effect, and the posted text is the keyword in it. As a result, the present inventors have completed the present invention. Here, the affiliate system is that the creator of the website makes a partnership with a company, puts a link in his website, and a general user who visits the website purchases a product via the pasted link. Or when requesting materials or registering as a member, this means a system in which a reward is paid from the partner company to the operator of the website according to each result.

  The impression determination system of the present invention extracts keywords and related words that are deeply related to the keywords from the document information of the acquired website, and determines the impression of the keywords by comprehensively calculating the degree of impression of the keywords for the related words. To do. By using the determination result determined in this way, it is possible to provide an advertisement that matches the impression of the keyword that the document has, thereby enhancing the advertising effect.

  In the following description of the embodiment, the configuration of the entire affiliate system including the impression determination system and the advertisement article generation system of the present invention will be described with reference to the drawings, and then the specific contents of the advertisement article generation system and the impression determination system will be described. Will be explained. In order to facilitate understanding of the description, the same reference numerals are given to the same components in the respective drawings, and duplicate descriptions are omitted.

  FIG. 1 is a conceptual diagram showing the configuration of the affiliate system 1. The affiliate system 1 includes a creator terminal 100, a user terminal 250, a Web server 200, an AI server 310 (advertisement article generation system), an advertisement management server 330, an advertisement distribution server 340, and a member management server 350.

  The creator terminal 100 is a general terminal used by a creator who creates a website, and is connected to the Internet N. The creator terminal 100 is used when the creator accesses the web server 200 to create a website and request membership registration as a member of the affiliate system 1. The creator terminal 100 is also used when the creator inputs tag information for advertising article placement received from the member management server 350 after member registration.

  The Web server 200 receives the information for the Web site from the creator terminal 100, accumulates the information, and puts it in a transmittable state. The information for creating the website includes document information such as a document posted on a blog. On the other hand, when the RSS feed for transmitting the RSS information is incorporated in the Web server 200 and new article information is accumulated on the Web site, the RSS information is generated. The RSS information includes information such as a headline and a summary. Note that the RSS feed is necessary when the AI server 310 acquires document information of a website using an RSS reader, but when the AI server 310 acquires document information by a general crawler. Is no longer necessary. The Web server 200 receives the advertisement article information created based on the transmitted document information from the advertisement distribution server 340 and includes the Web site information in a transmission enabled state.

  The user terminal 250 is a general terminal used by a user, is connected to the Internet N, and enables browsing of a website. Further, from the user terminal 250, the user can access an advertiser server (not shown) based on the information of the advertisement article, and can order material requests, products, etc. from the advertiser server. .

  The AI server 310 stores the member ID corresponding to the website creator obtained from the member management server 350 and the access destination information for RSS information acquisition, and acquires RSS information (document information) by the RSS reader based on the member ID. To do. Document information may be acquired by a general crawler without using an RSS reader.

  Also, the AI server 310 (advertisement article generation system) analyzes the acquired document information to determine an impression of the keyword of the document information, acquires advertisement information for generating an advertisement article based on this, and acquires the advertisement article information. Is generated. Details of the configuration of the AI server 310 will be described later.

  The advertisement management server 330 accumulates advertisement material information submitted by the advertiser, and transmits the advertisement material information when requested by the AI server 310. Details of the configuration of the advertisement management server 330 will be described later.

  The advertisement distribution server 340 receives and stores the advertisement article information generated by the AI server 310. Further, the advertisement distribution server 340 refers to the member management server 350 based on the member ID, acquires the transmission destination information of the advertisement article information, and transmits the advertisement article information to the Web server 200 from which the document information is acquired. Further, the advertisement distribution server 340 records an access log when the user terminal 250 accesses the advertiser's system from the advertisement article information generated by the AI server 310. The advertisement distribution server 340 totals the number of actions such as the number of accesses from the access log, and transmits the total data to the AI server 310.

  When there is a request for member registration as a member of the affiliate system 1 from the creator terminal 100, the member management server 350 performs member registration processing, acquires RSS information or site information of the created website, and creates The member ID of the person is generated. At the same time, the member management server 350 transmits RSS information or site information and a member management ID to the AI server 310, generates a tag for posting an advertisement article, and transmits it to the creator terminal 100. In addition, the member management server 350 acquires information on the action by the user to the advertiser from the advertisement distribution server 340 and transmits the information to a reward calculation server (not shown). The reward calculation server calculates a reward amount or reward points, and pays the creator.

  Next, details of the AI server 310 (advertisement article generation system) and the advertisement management server 330 will be described. FIG. 2 is a block diagram illustrating configurations of the AI server 310 and the advertisement management server 330.

  The AI server 310 (advertisement article generation system) includes a document information acquisition unit 311, a document information storage unit 312, a word analysis unit 313, a keyword extraction unit 315, a related word extraction unit 316, an impression determination unit 318, an impression information storage unit 317, The group word extraction unit 320, the advertisement identification unit 322, the advertisement extraction unit 324, and the advertisement article generation unit 326 are configured.

  The document information acquisition unit 311 acquires the document information of the website from the web server 200. The document information acquisition unit 311 registers an RSS feed access destination in advance. This access destination is an RSS feed access destination of a website created by a member of the affiliate system 1. The document information acquisition unit 311 sequentially collects access destination document information based on the registered RSS feed access destination information. FIG. 3 is a diagram illustrating a display example of a Web site from which document information has been acquired. In the RSS feed, the document information of the blog of December 2, 2005 that has been newly updated is set as RSS information.

  When collecting information by RSS, the RSS feed can transmit document information including title information and summary information created when the website is updated in advance. Increases the efficiency of collection. However, the present invention is not limited to using RSS, and can also be applied to general crawling. The document information acquisition unit 311 includes, for example, a memory such as a CPU and a RAM, or an external storage device such as a hard disk.

  The document information storage unit 312 stores the document information acquired by the document information acquisition unit 311. The stored document information is used for updating information in the impression information storage unit 317. The document information storage unit 312 is configured by a memory such as a RAM or an external storage device such as a hard disk.

  The word analysis unit 313 decomposes the acquired document information into words by morphological analysis. By performing the morphological analysis, the sentence is divided into columns of morphemes (minimum units having meaning in the language) and analyzed from reading, original form, part-of-speech type, utilization type, utilization form, and the like. When morphological analysis is performed, a grammar and a dictionary (a word list associated with information such as part of speech) prepared in advance in the word analysis unit 313 are referred to. The word analysis unit 313 is configured by a memory such as a CPU and a RAM or an external storage device such as a hard disk.

  The keyword extraction unit 315 extracts a phrase that meets a predetermined condition as a keyword from the decomposed word set. The predetermined conditions include, for example, (1) high occurrence frequency of words (Term Frequency: tf), (2) appearance in only a small number of documents in the document set (Document Frequency: df), (3) words There are combinations of verbs and adjectives related to. Among these, for df, an idf obtained by taking the logarithm of the reciprocal and normalized by the total number of documents may be used. The keyword extraction unit 315 is configured by a CPU, for example.

  The related word extraction unit 316 extracts words used in association with the keyword from the set of words decomposed as related words in the acquired document information. Whether or not they are related is determined by dependency analysis, for example, and it is determined that the modifier and the modifier are related to each other. The related word extraction unit 316 is configured by a CPU, for example.

  The impression information storage unit 317 stores, as a registered word, a word that is likely to be used together with a specific word in a document in association with a specific word, and for each registered word, a positive or negative impression value (impression) Of the degree). FIG. 4 shows an example of a table that stores positive or negative impression values for specific words for each registered word. For example, registered words such as “tea”, “green tea”, “bitterness”, “blend”, and “fragrance” are stored for a specific word “coffee”. The impression value for “coffee” is stored in association with each other. For example, in the example illustrated in FIG. 4, the impression value of “tea” for “instant” is “−1”, and the impression value of “blend” for “instant” is “2”. The impression information storage unit 317 is configured by a memory such as a RAM or an external storage device such as a hard disk.

  The impression determination unit 318 refers to the impression information storage unit 317 and comprehensively calculates the impression value (degree of impression) for the keyword for the related word, and determines the impression for the keyword obtained from the document information. As a result, not only the keyword is extracted from the document but also the impression of the extracted keyword is determined, so that it can be used to provide an advertisement that matches the impression received from the document using the determination result. As a result, the advertising effect can be enhanced.

  FIG. 5 is a diagram illustrating a result of totaling impression values for keywords. In the example shown in FIG. 5, for the keyword “coffee”, the related word “tea” has an impression value of “−1”, a related value in the document of “4”, and the number of occurrences in the document of “coffee”. When they are “1”, their product is “−4”. The related value in the document indicates the related depth determined by the proximity of the used related word to the keyword, the subject predicate relationship, and the like. For this related value, for example, a numerical value can be obtained by dependency analysis.

  Similarly, in FIG. 5, the product of the numerical values of the related word “bitter” is “−12”, and the product of the numerical values of the related word “blend” is “2”. Therefore, the sum of the products of these numerical values is “−14”, which is a negative numerical value. The impression determination unit 318 determines an impression for the keyword obtained from the document information by determining whether or not the value of the generally calculated impression value is 0 or more. When the numerical value is 0 or more, the determination result is positive, and when the numerical value is negative, the determination result is negative. Therefore, in the example illustrated in FIG. 5, the impression determination unit 318 determines that the impression of the keyword “coffee” obtained from the document information is negative. The impression determination unit 318 is configured by a CPU, for example. Note that the above-described determination method is merely an example, and does not limit combinations of parameters such as related values and the number of appearances.

  The group word accumulation unit 319 accumulates positive and negative word groups for specific words. FIG. 6 is a conceptual diagram showing an example of a word group stored in the group word storage unit 319. As shown in FIG. 6, the word group is accumulated in a weighted undirected structure. The numerical values described on the network in the figure indicate the relationship between each other, and the smaller the value, the higher the relationship. In general, data stored close to each other in the data structure is highly likely to be related. However, actual relevance is determined by weighting between data. The group word storage unit 319 is configured by a memory such as a RAM or an external storage device such as a hard disk.

  The group word extraction unit 320 refers to the group word storage unit 319 and extracts a plurality of words that can be used together with the keyword with the same impression as the determined impression as group words. For example, when a negative determination result is obtained, the group word storage unit 319 stores a negative word group of “coffee” for the keyword “coffee”, “tea” as a data structure close to “coffee”, When there are “green tea”, “juice”, “cake” and “water”, these are first extracted. Furthermore, the relevance including weighting is calculated for these, and when it is determined that the relevance of “tea”, “green tea”, and “cake” is high, these are extracted as group words.

  The advertisement specifying unit 322 refers to the advertisement information storage unit 335 and specifies a matched advertisement by matching the group word and information obtained from the advertisement material information. That is, identification is performed by searching for sentences close to the group word from sentences such as advertisement introduction sentences, advertisement sentences, and advertisement-related keywords. For example, the advertisement material information of the product name “XX tea” matching “tea”, the advertisement of the product name “□□ green tea” matching “green tea”, and the like are specified. The advertisement specifying unit 322 is configured by a CPU, for example. In the above matching, information on the advertiser's website can be used in addition to data prepared as advertisement information. Further, information search using a vector space model or the like as a matching method, random extraction of matching information, or the like can be used.

  The advertisement extraction unit 324 extracts advertisement material information corresponding to the specified advertisement from the advertisement information storage unit 335. For example, when text data “A tasteful XX tea and a comfortable afternoon time ...” is stored in the advertisement information storage unit 335 as the advertisement material information of “XX tea”, the text data is Extract. The advertisement extraction unit 324 is configured by a CPU, for example. The advertisement material information is not limited to text data, and may be accompanied by image data.

  The advertisement article generation unit 326 generates advertisement article information using the extracted advertisement material information, and transmits the generated advertisement article information to the Web server 200 that acquired the document information. Thereby, advertisement article information close to the impression received from the document is generated, and it is possible to generate an advertisement article that matches the impression of the document with respect to the central topic. As a result, the user's willingness to purchase can be further enhanced by the advertisement, and the advertising effect can be enhanced. For generating an advertisement article, at least extracted advertisement material information may be used, and a keyword or a group word may be used as necessary. When generating the advertisement article information, the advertisement article generation unit 326 generates a catch phrase using a keyword or a group word as necessary, and includes this in the advertisement article information. Depending on the setting, there is a case where no catch phrase is generated. In this manner, the advertisement article generation unit 326 can generate a catch phrase reflecting the impression that the document information has for the keyword.

  For example, when it is determined that the document information has a negative impression with respect to the keyword “coffee”, the advertisement article generation unit 326 generates a catch phrase such as “best selection to be sent to you who dislike coffee”. By transmitting the advertisement article information generated in this way to the Web server 200, the Web server 200 can post the advertisement article information generated in the information on the website. FIG. 7 is a diagram illustrating a display example of a website on which the generated advertisement article information is posted. The advertisement article generation unit 326 is configured by a CPU, for example.

The advertisement management server 330 includes an advertisement information reception unit 332 and an advertisement information storage unit 335. The advertisement information receiving unit 332 receives the advertisement processing performed by the advertiser server and the advertisement material information submitted from the advertiser server. Advertising information reception part 332 is constituted by CPU, for example. The advertisement information accumulation unit 335 accumulates the received advertisement material information, receives the request from the advertisement extraction unit 324 of the AI server 310, and transmits the advertisement information to the AI server 310. The advertisement information storage unit 335 is configured by a memory such as a RAM or an external storage device such as a hard disk.
In the above example, the AI server 310 performs the function of generating an advertisement article, but the present invention is not limited to such a form and can be applied. That is, the advertisement article generation system may be configured by a plurality of servers, and the advertisement article generation system may include the function of the advertisement management server 330 in addition to the function of advertisement article generation.

  Next, a characteristic operation of the AI server 310 configured as described above will be described. As a premise, the creator must be registered as a member of the affiliate system 1. At the time of member registration, the creator terminal 100 accesses the member management server 350 and receives an advertisement placement tag. The member management server 350 receives the access destination information of the RSS feed of the creator's website and issues a member ID. The AI server 310 (advertisement article generation system) receives the RSS feed access destination information and the member ID from the member management server 350 and stores them.

  Based on such a situation, the AI server 310 (advertising article generation system) operates. FIG. 8 is a flowchart showing a characteristic operation of the AI server 310. First, the AI server 310 acquires the document information of the Web site from the Web server 200 by using the document information acquisition unit 311 (step S1). Then, the word analysis unit 313 decomposes the acquired document information into words by morphological analysis (step S2). Next, the AI server 310 uses the keyword extraction unit 315 to extract, as keywords, phrases that meet a predetermined condition from the decomposed word set (step S3). Details of the keyword extraction process will be described later. Subsequently, the AI server 310 uses the related word extraction unit 316 to extract words (related words) used in association with the keyword from the decomposed word set in the acquired document information. Whether or not they are related is determined by dependency analysis, for example.

  Next, the AI server 310 refers to the impression information storage unit 317 by the impression determination unit 318, comprehensively calculates the impression value for the keyword over all related words, and determines the impression for the keyword obtained from the document information. (Step S5). Details of the impression determination process will be described later. The AI server 310 uses the group word extraction unit 320 to refer to the group word storage unit 319 and extract a group word (step S6). A group word is a plurality of words that are easy to use with keywords with the same impression as the determined impression. Details of the group word extraction processing will be described later.

  Next, the AI server 310 uses the advertisement identification unit 322 to refer to the advertisement information storage unit 335 and identifies a matched advertisement by matching the group word and information obtained from the advertisement material information (step S7). Then, the advertisement material information corresponding to the specified advertisement is extracted from the advertisement information storage unit 335 by the advertisement extraction unit 324 (step S8). Finally, the AI server 310 uses the advertisement article generation unit 326 to generate advertisement article information using the extracted advertisement material information, and transmits the generated advertisement article information to the Web server 200 that acquired the document information. (Step S9). Details of the advertisement article generation processing will be described later. By transmitting the advertisement article information generated in this manner to the Web server 200, the Web server 200 can post the advertisement article information generated on the Web site. The advertisement article information is transmitted through the advertisement distribution server 340.

  In this way, when the user accesses the advertiser site from the advertisement article information posted on the website, the advertisement distribution server 340 records the access log and transmits the access log to the AI server 310. The AI server 310 can increase the relevance of words and the reliability of the impression value based on the fed back information. In addition, when the user accesses the advertiser site from the advertisement article information and performs an action that leads to the advertisement result, the reward calculation server (not shown) calculates the reward amount, and the creator (user member) receives the reward. Will be paid.

  Of the operations of the AI server 310, the keyword extraction operation will be described in detail below. FIG. 9 is a flowchart showing the keyword extraction process. First, the keyword extraction unit 315 determines whether one of the words (minimum unit) decomposed by the word analysis unit 313 is a part of speech that is the center of the topic (step T1). If it is not the part of speech that is the center of the topic, the process proceeds to step T5. If the word is a part of speech that is the center of the topic, it is determined whether or not it is necessary to process the word to create a phrase that can be a keyword (step T2). When it is not necessary to process, it progresses to step T4. If the word needs to be processed, the word is processed into a phrase (step T3). For example, words are combined and processed into phrases with nouns + nouns, or processed into phrases with nouns + adverbs + nouns.

  Subsequently, the processed words or words determined not to be processed are selected (step T4). Then, it is determined whether or not the processing up to the above selection has been completed for all the decomposed words (step T5). If all the words have not been sorted, the process returns to step T1. When all the words have been selected, important words are extracted from the selected words (step T6). Phrases that are important by selecting a local weighting method, a global weighting method, a method of giving weight when there is a verb or adjective related to the phrase, or a combination thereof (predetermined conditions) among the selected words (Keyword) is extracted, and the keyword extraction process ends. The local weighting method is a method that considers the high frequency of words in the document, and the global weighting method is a method that considers that only a small number of documents in the document set appear. .

  Next, of the operations of the AI server 310, the impression determination operation will be described in detail below. FIG. 10 is a flowchart showing impression determination processing. First, the impression determination unit 318 refers to the impression information storage unit 317 and determines whether or not the extracted related word is a registered word (step P1). Here, the registered word is a phrase stored in the impression information storage unit 317 as a phrase that can be easily used with a specific phrase. As a result of the determination, if the extracted related word is not a registered word, the process proceeds to step P4. If the extracted related word is a registered word, the impression value of the registered word is read from the impression information storage unit 317 (step P2). Then, the impression value of the registered word is multiplied by a weight and added (step P3). The weighting includes, for example, a related value indicating the related depth of the related word used in the document with respect to the keyword, the number of appearances of the related word, and the like.

  Subsequently, it is determined whether or not the processing up to the above addition has been completed for all related words (step P4). If there is a related word that has not been processed, the process returns to step P1. When the processing has been completed for all the related words, the above addition values are summed for all the related words, and it is determined whether or not the sum of the impression values is 0 or more (step P5). If the total impression value is negative as a result of the determination, it is determined that the impression for the keyword of the document information is negative (step P6). On the other hand, if the total impression value is 0 or more, it is determined that the impression of the keyword of the document information is positive (step P7). Then, the impression determination process ends.

Next, the group word extraction operation among the operations of the AI server 310 will be described in detail below. FIG. 11 is a flowchart showing group word extraction processing. First, the group word extraction unit 320 determines whether or not the keyword impression is determined to be positive (step Q1). If it is determined to be negative, a phrase close to the data structure is extracted from the negative word group (step Q2). On the other hand, if it is determined to be positive, words that are close to the data structure are extracted from the positive word group (step Q3). Next, the group word extraction unit 320 extracts a word having particularly high relevance as a group word by calculating the relevance from the weight of the relevance of the extracted word with the keyword (step Q4).
Next, among the operations of the AI server 310, the operation of generating an advertisement article will be described in detail below. FIG. 12 is a flowchart showing the advertisement article generation process. The advertisement article generation unit 326 determines whether a catch phrase is necessary (step R1). Whether a catch phrase is necessary or not depends on the setting. For example, if the advertiser does not want it, it is possible to set not to generate a catch phrase. If it is determined that the catch phrase is not necessary, the process proceeds to step R5.

  On the other hand, if it is determined that the catch phrase is necessary, it is determined whether or not the impression of the keyword of the document information is determined to be positive (step R2). If the impression is determined to be negative, a negative catch phrase is generated (step R3). The catch phrase is generated by combining nouns, verbs, adjectives, templates, fixed phrases, and the like. The negative catch phrase is, for example, “XX tea rather than coffee”, “coffee is good but □□ beverage □□ green tea”, and the like. On the other hand, if the impression is determined to be positive, a catch phrase for positive is generated (step R4). The positive catch phrase is, for example, “△△” for coffee or “xx for delicious coffee”. Finally, advertisement article information is generated by synthesizing the catch phrase and the advertisement material information (step R5). In this way, the advertisement article generation process is completed.

  The characteristic operation of the present invention as described above is performed by executing a program. This program can be obtained in a state of being recorded on a recording medium such as a CD-ROM or a DVD. In addition, such a program is transmitted by a computer as a transmission device via a transmission medium such as a communication network including a public telephone line, a dedicated telephone line, a cable TV line, a wireless communication line, etc. constituting the network. It can also be obtained by receiving the received signal.

  In the above embodiment, an impression is determined using an impression value that can take a real number range. However, an impression is determined based on the magnitude of both using a positive value and a negative value that take positive values. May be.

It is a conceptual diagram which shows the structure of an affiliate system. It is a block diagram which shows the structure of AI server (advertisement article production | generation system) and an advertisement management server. It is a figure which shows the example of a display of the Web site from which document information was acquired. The example of the table which shows the impression value with respect to a specific word for every registration word. It is a figure which shows the result which totaled the impression value with respect to a keyword. It is a conceptual diagram which shows an example of the word group accumulate | stored in a group word accumulation | storage part. It is a figure which shows the example of a display of the website which published the produced | generated advertisement article information. It is a flowchart which shows the characteristic operation | movement of AI server (advertisement article production | generation system). It is a flowchart which shows a keyword extraction process. It is a flowchart which shows an impression determination process. It is a flowchart which shows the extraction process of a group word. It is a flowchart which shows the production | generation process of an advertisement article.

Explanation of symbols

1 Affiliate System 100 Creator Terminal 200 Web Server 250 User Terminal 310 AI Server (Advertisement Article Generation System)
311 Document information acquisition unit 312 Document information storage unit 313 Word analysis unit 315 Keyword extraction unit 316 Related word extraction unit 317 Impression information storage unit 318 Impression determination unit 319 Group word storage unit 320 Group word extraction unit 322 Advertisement specification unit 324 Advertisement extraction unit 326 Advertising article generation unit 330 Advertising management server 332 Advertising information reception unit 335 Advertising information storage unit 340 Advertising distribution server 350 Member management server

Claims (3)

A document information acquisition unit that acquires document information of a website from a web server that enables browsing of the website;
A word analysis unit that decomposes the document information into words by morphological analysis;
A keyword extraction unit that extracts, as a keyword, a phrase that meets a predetermined condition from the set of decomposed words;
In the acquired document information, a related word extraction unit that extracts a phrase used in association with the keyword as a related word from the set of decomposed words;
An impression information storage unit that stores words that are likely to be used together with the specific word as a registered word in association with the specific word, and that stores a positive or negative impression degree for the specific word for each registered word; ,
An impression determination unit that refers to the impression information storage unit, comprehensively calculates the degree of impression with respect to the keyword for the related word, and determines an impression with respect to the keyword obtained from the document information;
A group word accumulation unit for accumulating positive and negative word groups for a specific phrase,
Referring to the group word storage unit, a group word extraction unit that extracts, as a group word, a plurality of words that are easily used with the keyword in the same impression as the determined impression;
For each advertisement, refer to an advertisement information storage unit that stores advertisement material information, an advertisement specifying unit that specifies a matched advertisement by matching the information obtained from the group word and the advertisement material information,
An advertisement extraction unit that extracts advertisement material information corresponding to the identified advertisement from the advertisement information storage unit;
An advertisement article generation unit that generates advertisement article information using the extracted advertisement material information and transmits the generated advertisement article information to the Web server that acquired the document information. Advertising article generation system. An advertisement article generation method performed by an advertisement article generation system including a computer,
A document information acquisition step of acquiring document information of a Web site from a Web server that enables browsing of the Web site;
A word analysis step of decomposing the document information into words by morphological analysis;
A keyword extraction step of extracting a phrase that meets a predetermined condition as a keyword from the set of decomposed words;
In the acquired document information, a related word extracting step of extracting from the set of decomposed words as a related word a phrase used in association with the keyword;
An impression information storage unit that stores words that are easily used together with the specific word as a registered word in association with the specific word, and stores a positive or negative impression degree for the specific word for each registered word Referencing, comprehensively calculating the degree of impression of the keyword with respect to the related word, and an impression determination step of determining an impression of the keyword obtained from the document information;
With reference to a group word storage unit that stores positive and negative word groups for a specific phrase, a plurality of phrases that are easy to use with the keyword with the same impression as the determined impression are extracted as group words. A group word extraction step;
For each advertisement, referring to an advertisement information storage unit that stores advertisement material information, an advertisement specifying step for specifying a matched advertisement by matching the information obtained from the group word and the advertisement material information;
An advertisement extraction step of extracting advertisement material information corresponding to the specified advertisement from the advertisement information storage unit;
An advertisement article generating step of generating advertisement article information using the extracted advertisement material information and transmitting the generated advertisement article information to the Web server that acquired the document information. An advertisement article generation method characterized by being performed by an article generation system. An advertisement article generation program to be executed by an advertisement article generation system provided with a computer,
Document information acquisition processing for acquiring document information of a Web site from a Web server that enables browsing of the Web site;
A word analysis process for decomposing the document information into words by morphological analysis;
A keyword extraction process for extracting, as a keyword, a phrase that meets a predetermined condition from the set of decomposed words;
In the acquired document information, a related word extraction process for extracting a word / phrase used in association with the keyword as a related word from the set of decomposed words,
An impression information storage unit that stores words that are easily used together with the specific word as a registered word in association with the specific word, and stores a positive or negative impression degree for the specific word for each registered word Referring to the related word, comprehensively calculating the degree of impression of the keyword, and impression determination processing for determining the impression of the keyword obtained from the document information;
Referring to a group word storage unit that stores positive and negative word groups for a specific word and phrase, a plurality of words and phrases that are easy to use with the keyword with the same impression as the determined keyword. Group word extraction processing to be extracted as
For each advertisement, referring to an advertisement information storage unit that stores advertisement material information, an advertisement specifying process for specifying a matched advertisement by matching the information obtained from the group word and the advertisement material information;
An advertisement extraction process for extracting advertisement material information corresponding to the specified advertisement from the advertisement information storage unit;
An advertisement article generation process that generates advertisement article information using the extracted advertisement material information, and transmits the generated advertisement article information to the Web server that acquired the document information. An advertisement article generation program which is executed by an article generation system.

Images