JP2011003158A - Summary preparation device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically prepare a summary of a content described in a text, in the text such as a blog.SOLUTION: A summary preparation device includes a blog data storage part 15 for storing a file including the text containing a plurality of sentences, a paragraphing processing part 232 for paragraphing the text in the file by every one sentence, when receiving a summary preparation request of the text in the file stored in the blog data storage part 15, a sentence selection processing part 233 for selecting a sentence finished with a prescribed sentence end character determined in response to an attribute of the text, out of the plurality of sentences, and a network interface 11 for outputting the selected sentence as a reply to the summary preparation request.

Description

  The present invention relates to a technique for automatically generating a text summary.

  Many blog sites and websites (bulletin boards) exist on the Internet, and the opinions and impressions of consumers about various products and services called so-called word-of-mouth information are written in text. Recently, sales of products and services are influenced by this word-of-mouth information.

  For this reason, it is important for companies to understand word-of-mouth information on the Internet regarding their products and competitors' products. Therefore, for example, techniques such as Patent Documents 1 and 2 have been proposed.

JP 2004-185572 A JP 2008-262520 A

  For example, in blogs and bulletin boards, there is a need to quickly grasp what is described about a certain product and the like. Also, since information on the Internet is constantly added, I want to analyze it in real time.

  On the other hand, in the conventional Japanese text analysis processing, morphological analysis, dependency analysis, and the like are widely known, but these processings require a considerable amount of time.

  Accordingly, an object of the present invention is to automatically create a summary of the contents described therein in a text such as a blog.

  A summary creation device according to an embodiment of the present invention includes a text storage unit that stores a file including text including a plurality of sentences, and a text summary creation request in the file stored in the text storage unit. Is received, a delimiter processing unit that delimits the text in the file for each sentence, and a sentence ending with a predetermined end character determined according to the attribute of the text is selected from among a plurality of sentences delimited by the delimiter processing unit A selection processing unit that outputs the sentence selected by the selection processing unit as a response to the summary creation request.

  In a preferred embodiment, the selection processing unit may select a sentence including a predetermined number of frequent words or more in the text among a plurality of sentences delimited by the delimiter processing unit.

  In a preferred embodiment, when the text is a blog describing a product or service, the end-of-sentence characters are at least “foot”, “profit”, “mind”, “su”, “i”, “ri”. , “Ru”, “yo” and “n” may be included.

  In a preferred embodiment, when the text is a newspaper article, the end-of-sentence character may include at least “su”, “ta”, “ru”, and “dan”.

  In a preferred embodiment, when the text is a newspaper article, the selection processing unit may select a sentence having a sentence beginning with “(” and a sentence ending with “”.

  In a preferred embodiment, when the text is a newspaper article, the end-of-sentence character has one of “in this”, “just”, and “other” and the end-of-sentence is “ You may select the sentence immediately before the sentence that is one of “su”, “ta”, “ru”, and “dan”.

1 is a schematic diagram of an information providing system including an information providing apparatus according to an embodiment of the present invention. 1 is a configuration diagram of a blog analysis server 1. FIG. It is a flowchart which shows the process sequence of the blog data collection part. 4 is a schematic diagram illustrating an example of a data structure of a blog data storage unit 15. FIG. It is a block diagram of the advertisement server 4. It is a flowchart which shows the information provision process procedure in this embodiment. The example of the screen displayed on the user terminal device 3 is shown. The example of the screen displayed on the user terminal device 3 is shown. The example of the screen displayed on the user terminal device 3 is shown. The example of the screen displayed on the user terminal device 3 is shown. It is a flowchart which shows a word-of-mouth information analysis request process. It is a flowchart which shows the detailed procedure of the text analysis process of step S35. It is a flowchart which shows an item analysis request process. It is a flowchart which shows the summary production | generation process.

  Hereinafter, an information providing system including an information providing apparatus according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a configuration diagram of an information providing system according to an embodiment of the present invention. As shown in the figure, this system has a blog analysis server 1, a plurality of user terminal devices 3, an advertisement server 4, and a blog server 5, which are connected via a network 9 such as the Internet. ing.

  Here, the blog analysis server 1 and the advertisement server 4 constitute an information providing device that provides information to the user terminal device 3. The information providing device including the blog analysis server 1 and the advertisement server 4 may be realized on a single computer device or may be realized by a plurality of computer devices.

  The blog analysis server 1, the user terminal device 3, the advertisement server 4, and the blog server 5 are all configured by, for example, a general-purpose computer system, and the blog analysis server 1, user terminal device 3, advertisement server 4, and blog described below Individual components or functions in the server 5 are realized, for example, by executing a computer program. This computer program can be stored in a computer-readable recording medium.

  The blog analysis server 1 collects blog data from a plurality of blog servers 5 using a program called a crawler, and provides a blog data analysis service to the plurality of user terminal devices 3. In the present embodiment, the blog analysis server 1 collects blog data related to so-called word-of-mouth information related to various products and services (categories) from the plurality of blog servers 5. And according to the request from the user terminal device 3, the analysis result of the word-of-mouth information regarding goods and service is provided. Furthermore, the blog analysis server 1 creates and provides a summary of word-of-mouth information related to products and services in response to a request from the user terminal device 3. That is, the blog analysis server 1 functions as a text analysis device that analyzes text such as blog data and a summary creation device that summarizes the text.

  In the present embodiment, blog data will be described in particular. However, in addition to this, data on a website constructed on a web server connected to a network can be targeted.

  The user terminal device 3 is a computer device that can access the network 9 and is equipped with a web browser. Various screens to be described later are displayed using, for example, a web browser.

  The advertisement server 4 provides various information to the user terminal device 3 and information related to advertisements for products and services. For example, the advertisement server 4 outputs data of a display screen for displaying an advertisement image related to the advertisement product or service to the user terminal device 3.

  The blog server 5 stores data including various texts written by many Internet users, and a blog site based on the data can be viewed by other Internet users.

  FIG. 2 shows a configuration diagram of the blog analysis server 1.

  The blog analysis server 1 acquires blog data from the blog server 5, analyzes or summarizes the blog data, and provides the result to the user terminal device 3. Therefore, the blog analysis server 1 has the following configuration. That is, the blog analysis server 1 includes a network interface unit 11, a blog data collection unit 13, a blog data storage unit 15, a text analysis unit 17, an extracted character string storage unit 19, an item ranking processing unit 21, and a summary. And a generating unit 23.

  The network interface unit 11 communicates with other devices on the network 9 such as the user terminal device 3 and the blog server 5 via the network 9. For example, upon receiving a word-of-mouth information analysis request (blog data analysis request) including designation of a certain category from the user terminal device 3, the network interface unit 11 outputs an analysis result relating to the category. When the network interface unit 11 receives an item analysis request including designation of a character string based on the analysis result from the user terminal device 3, the network interface unit 11 outputs an analysis result regarding an item based on the designated character string. Further, when receiving a summary creation request from the user terminal device 3, the network interface unit 11 outputs a summary of text relating to the request. In the present embodiment, the category corresponds to a product or service. Details of the word-of-mouth information analysis request, the item analysis request, and the summary creation request will be described later.

  The blog data collection unit 13 collects blog data from each blog server 5 via the network 9. The blog data collection unit 13 classifies the collected blog data and stores it in the blog data storage unit 15.

  FIG. 3 is a flowchart showing a processing procedure of the blog data collection unit 13. The detailed processing of the blog data collection unit 13 will be described with reference to FIG.

  First, the blog data collection unit 13 collects blog data from a plurality of blog servers 5 via the network 9 (S11).

  The blog data collected here can be divided into predetermined blocks, for example, entries (articles). Therefore, the blog data collection unit 13 performs the following processing in a predetermined processing unit such as an entry unit. First, the blog data collection unit 13 analyzes the text in one entry and determines the category and item of the entry (S13).

  For example, the blog data collection unit 13 extracts one entry from the blog data, extracts a character string indicating a predetermined category and item from the text, and determines the category and item. Here, one or more categories and items are specified for each entry. An entry from which one or more categories and items are not extracted is an entry that is not subject to further processing.

  In the present embodiment, the type of product or service is determined as the category. The product type is a general name of a specific product such as “refrigerator”, “washing machine”, or “air conditioner”, and the service type is a specific product such as “pachinko” or “theme park”. The general name of a simple service may be used. Further, in the present embodiment, identification information that identifies an individual product or service within the product or service type is determined as an item. For example, in the case of a product, the name of the manufacturer and model (or model number) that manufactures or supplies each product. In the case of a service, the name of an individual store or facility that provides the service and the name of a specific service. You may specify an item.

  Next, the blog data collection unit 13 determines whether or not the folder of the category determined in step S13 already exists in the blog data storage unit 15 (S15). When the folder of the category does not exist (S15: No), the category-specific folder of the category is generated (S17). On the other hand, if a folder of the target category already exists in the blog data storage unit 15 (S15: Yes), step S17 is skipped.

  That is, when a new category is detected, the blog data collection unit 13 generates a category-specific folder corresponding to the category in the blog data storage unit 15. For example, the folder name of each folder may be the category name. In the present embodiment, folders with folder names such as “refrigerator”, “washing machine”, and “air conditioner” are generated (see FIG. 4).

  The blog data collection unit 13 determines whether an item-specific file corresponding to the item determined in step S13 already exists in the category folder determined in step S13 (S19). When there is no target item-specific file (S19: No), the item-specific file is generated, and the text of the entry to be processed is stored in the generated file (S21). On the other hand, when an item-specific file corresponding to the item determined in step S13 already exists (S19: Yes), the text of the target entry is additionally stored in the existing file (S23).

  For example, when a new item is detected, the blog data collection unit 13 generates a text file of the item in the corresponding category folder in the blog data storage unit 15. For example, the item name may be the file name of the item-specific file. In the present embodiment, for example, the file name may be “maker name_model name”. When the same item is described in different entries, the texts of these entries are stored in the same file.

  The processes after step S13 are repeatedly executed for all the blog data entries acquired in step S11 (S25).

  The blog data collection unit 13 executes the above-described processing related to collection and classification of blog data regularly or irregularly.

  Thereby, the collected blog data is classified according to the category and item described in each. That is, the blog data storage unit 15 stores a file containing text. A schematic diagram of an example of the data structure of the blog data storage unit 15 is shown in FIG.

  As shown in FIG. 4, the blog data storage unit 15 includes one or more category-specific folders 150 (150a to 150c). Each folder 150 stores one or more text files 152. Each text file 152 includes item-specific texts relating to different items belonging to a category corresponding to each stored folder 150.

  For example, blog data (entry) including a description about a refrigerator of a model xxx of company A is stored in a text file 152 named “A company_xxx” in the “refrigerator” folder 150a.

  Referring again to FIG. 1, when the text analysis unit 17 receives a first request (word-of-mouth information analysis request) including information indicating the category of the advertisement product or service related to the advertisement image from the user terminal device 3, the text storage unit 17 stores the text storage. Referring to the section (blog data storage section 15), a plurality of frequently occurring character strings appearing at a frequency or number of times greater than or equal to a predetermined frequency are extracted from the text described regarding the product or service of the category included in the first request. Hereinafter, the text analysis unit 17 will be described in more detail.

  The text analysis unit 17 performs text analysis of the blog data stored in the blog data storage unit 15. The text analysis unit 17 includes a specific character string processing unit 171, an erasure character string processing unit 173, an unnecessary character string processing unit 175, and a cutout processing unit 177 for performing text analysis.

  The specific character string processing unit 171 extracts a predetermined specific character string from the text in the file stored in the blog data storage unit 15, and extracts the extracted specific character string to the extracted character string storage unit 19. Store. At this time, the specific character string processing unit 171 replaces the extracted specific character string with a blank. Further, the specific character string processing unit 171 may count the number of appearances or the appearance frequency of the extracted specific character string, and store the count result in the blog data storage unit 15 together. For example, the specific character string processing unit 171 holds a specific character string dictionary in advance. The specific character string dictionary stores a plurality of specific character strings. The specific character string processing unit 171 performs matching with text and replacement with a blank in order from a specific character string having a long character string length.

  Here, the specific character string dictionary may be provided for each attribute of the target blog data (text). That is, the specific character string stored in the specific character string dictionary may be different according to the attribute of the blog data (text). For example, since “product word-of-mouth information” and “newspaper article” have different attributes, the specific character strings constituting the specific character string dictionary may be different from each other.

  The erase character string processing unit 173 blanks out erase characters including at least punctuation marks and all hiragana characters or all hiragana characters other than a predetermined excluded character in the text in the text file stored in the blog data storage unit 15. Replace.

  The inventor of the present application has found that in texts including blogs in general, many meaningful expressions related to the context described in the text are expressed in Chinese characters. Therefore, in the present embodiment, a character string composed of kanji is extracted as a character string (word, keyword) characterizing the description content of a certain text.

  The specific character string dictionary described above includes important character strings as specific character strings in relation to the attributes of the target blog data (text). This is because an important expression (character string) exists even if it is a character string including components other than kanji, so that the important expression that the specific character string processing unit 171 has set as the specific character string is used as an erasure character. The column processing unit 173 extracts the erased characters before erasing them.

  For example, the erasure character may include all hiragana characters, or all except one or more hiragana characters (excluded characters) of “no”, “ga”, “i”, and “ku”. Hiragana may be included. That is, in the latter case, any one or more of “NO”, “GA”, “I”, and “KU” is not included in the erase character. This is because these characters may play a role of combining character strings composed of kanji characters, and the kanji character strings combined with these hiragana characters may make a certain meaning. “·” And “−” may not be included in the erasure character for the same reason.

  In addition, the erasure characters may include punctuation marks, commas, periods, alphabets, and various symbols.

Next, an example of a character string included in the erased character is shown below.
Erase character = [“A”, “I”, “U”, “E”, “O”, “K”, “K”, “K”, “K”, “K”, “S”, “S” ”,“ Su ”,“ se ”,“ so ”,“ ta ”,“ chi ”,“ tsu ”,“ te ”,“ to ”,“ na ”,“ ni ”,“ nu ”,“ ne ”, “Ha”, “hi”, “fu”, “he”, “ho”, “ma”, “mi”, “mu”, “me”, “mo”, “ya”, “yu”, “yo” ",""Ra","ri","ru","re","ro","wa","wo","ga","gi","gu","ge","go","Za","Ji","Zu","Ze","Zo","Da","Da","Zu","De","Do","Ba","Bi","Bu" ”,“ Be ”,“ bo ”,“ pa ”,“ pi ”,“ pu ”,“ pe ”,“ po ”,“ nya ”,“ yu ”,“ yo ”,“ tsu ”,“ n ”, ".", ",", ".", "(", ")", "{", "}", """,""","~","]",""",""","<",">",""","""," ■ "," * ","! "," = "," * ","! ","(",")" , “/”, “[”, “+”, “¥”, “$”, “&”, “&”, “@”, “@”, “*”, “…”, “a”, “ 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n' , “O”, “p”, “q”, “r”, “s”, “t”, “u”, “v”, “w”, “x”, “y”, “z”, “ `` A '', `` B '', `` C '', `` D '', `` E '', `` F '', `` G '', `` H '', `` I '', `` J '', `` K '', `` L '', `` M '' , “N”, “O”, “P”, “Q”, “R”, “S”, “T”, “U”, “V”, “W”, “X”, “Y”, “ "Z", "!", "#", "$", "%", "&", "(", ")", "^", "=", "~", "|", "{" , “}”, “[”, “]”, “:”, “;”, “+”, “*”, “}”, “_”, “?”, “/”, “.”, “ <",">",","," \\ "," \ t "," \ b "," \ "," \ "", "\ r", "\ n", ",", ":", "(Summed brackets)", "(closed brackets)", "[", "]", """,".",","].

  The unnecessary character string processing unit 175 extracts one or more predetermined unnecessary character strings and replaces them with blanks. For example, words that have no meaning in text analysis and are frequently used on the net (for example, “tag”, “page”, “search”, etc.) may be used as unnecessary character strings.

  The cut-out processing unit 177 is a character that is sandwiched by blanks from the processed text processed by at least one of the specific character string processing unit 171, the erased character string processing unit 173, and the unnecessary character string processing unit 175. Extraction processing is performed for extracting a column and storing the extracted character string in the extracted character string storage unit 19. At this time, if the beginning (start) or end (end) of the extracted character string is a character other than Kanji, the extraction processing unit 177 sends the character string from which the character or the like has been removed to the extracted character string storage unit 19. Store. This is because hiragana, symbols, etc. that are not included in the erase character string may start or end. As will be described later, the extracted character string storage unit 19 stores the extracted character string and the number of appearances or the appearance frequency in association with each other. Therefore, when the cutout processing unit 177 extracts the same character string in duplicate, the number of appearances or the appearance frequency of the character string is counted.

  The extracted character string storage unit 19 stores a character string extracted from the text file in the blog data storage unit 15. The extracted character string storage unit 19 stores the character string and the number of appearances or the appearance frequency of each character string in association with each other.

  When one category is specified by a word-of-mouth information analysis request or the like, the text analysis unit 17 reads one or more text files 152 in the folder 150 of the specified category, and the text in the read text file 152 A specific character string replacement process by the specific character string processing unit 171, an erase character string deletion process (replacement to a blank) by the erase character string processing unit 173, and an unnecessary character string deletion process by the unnecessary character string processing unit 175 are performed. Also good. Then, the cut-out processing unit 177 performs the above-described cut-out process on the processed text subjected to these processes, and the extracted character string is stored in the extracted character string storage unit 19. For example, when “washing machine” is specified as a category by a word-of-mouth information analysis request from the user terminal device 3, the above-described processing is performed on the file 152 stored in the washing machine folder 150a, and the washing machine is related to the washing machine. Character strings that characterize the content described in the blog are extracted.

  The network interface unit 11 outputs the plurality of character strings stored in the extracted character string storage unit 19 by the above processing to the user terminal device 3 that has made the word-of-mouth information analysis request. For example, the network interface unit 11 outputs data of a display screen for displaying a plurality of frequent character strings (extracted character strings) extracted by the above process to the user terminal device 3. The plurality of character strings output from the network interface unit 11 may be a character string list sorted according to the number of appearances or the appearance frequency.

  The item ranking processing unit 21 issues a second request (item analysis request) including data indicating one selected frequent character string selected from a plurality of frequent character strings output from the network interface unit 11 from the user terminal device 3. When accepted, the frequently-selected character string for each product or service in the text describing the product or service in the category related to the first request (word-of-mouth information analysis request) with reference to the text storage unit (blog data storage unit 15) Count the frequency or number of occurrences of. For example, when receiving an item analysis request in which one character string in a character string output as a response to a word-of-mouth information analysis request is selected, the item ranking processing unit 21 is in the category targeted by the word-of-mouth information analysis request. The number of appearances or the appearance frequency of the selected character string in each file is counted. That is, the item ranking processing unit 21 counts the number of appearances or the appearance frequency of the selected character string for each item in the blog data stored in the blog data storage unit 15. The item ranking processing unit 21 generates a product or service ranking (item list in which item names are sorted) based on the counting result.

  The count result is transmitted by the network interface unit 11 to the user terminal device 3 that has requested the item analysis. For example, the network interface unit 11 may output the item name list described above. Further, the network interface unit 11 may output display screen data for displaying the ranking generated by the item ranking processing unit 21 to the user terminal device 3.

  The summary generation unit 23 generates a text summary of the blog data stored in the blog data storage unit 15. The summary generation unit 23 includes an appearance number counting unit 231, a delimiter processing unit 232, and a sentence selection processing unit 233 in order to generate a summary.

  When the summary generation unit 23 receives a third request (summary generation request) including information indicating the selected product or service selected from the ranking from the user terminal device 3, the summary generation unit 23 receives the text storage unit (blog data storage unit 15). ) To generate a text summary describing the selected product or service.

  The appearance count unit 231 counts the number of appearances of each text word (character string) in the file stored in the blog data storage unit 15. For example, the appearance count counting unit 231 performs a counting process on the text before the delimiter processing unit 232 performs the delimiter process. This counting result is used in the sentence selection processing unit 233. Note that the appearance count unit 231 may perform the same processing as the text analysis unit 17 regarding the extraction of the character string and the count of the appearance frequency or the appearance frequency.

  When the delimiter processing unit 232 receives a text summary creation request in a file stored in the blog data storage unit 15, the delimiter processing unit 232 performs delimiter processing for delimiting the text in the file for each sentence. For example, the text in the file corresponding to the item specified in the summary creation request from the user terminal device 3 is read, and a plurality of sentences in the text are detected and separated, for example.

  The sentence selection processing unit 233 performs a selection process of selecting a sentence ending with a predetermined end-of-sentence character determined according to the text attribute from among the plurality of sentences delimited by the delimiter processing unit 232. Here, there may be a plurality of types of predetermined sentence end characters. Each type of sentence end character may be a single character or a plurality of characters. For example, when the text is a blog that describes a product or service, the end-of-sentence characters are at least “foot”, “profit”, “mind”, “su”, “i”, “ri”, “ru”, “yo” And “n” may be included. Alternatively, when the text is a newspaper article, the end-of-sentence characters may include at least “su”, “ta”, “ru”, and “dan”. In addition, when the text is a newspaper article, the sentence selection processing unit 233 may select a sentence having a sentence head of “(” and a sentence end of “)”. Further, when the text is a newspaper article, the sentence selection processing unit 233 has the sentence head of any one of “inside”, “just”, and “other” and the sentence end is “s”, The sentence immediately before the sentence that is one of “TA”, “RU”, and “DA” may be selected. This is because in the case of a newspaper article, there is a sentence that quotes the previous sentence, and if there is no previous sentence, the sentence may be unclear. In the case of a newspaper article, if a character string enclosed in brackets (“,”) includes a punctuation mark, the above processing may be performed after the punctuation point is replaced with a punctuation mark.

  The inventor of the present application has found that sentence endings that can be text summaries are different depending on the text attributes as described above. The sentence selection process using the above sentence end character is based on this knowledge.

  In addition, the sentence selection processing unit 233 may select a sentence including a frequent word whose number of appearances in the text counted by the appearance number counting unit 231 is a predetermined number or more among the plurality of sentences delimited by the delimiter processing unit 232. Good. For example, in a text in a target file, a sentence including a character string that appears more than a predetermined number of times is selected regardless of the sentence end character. For example, a character string that appears 3 times or more when the text is less than 3000 characters and 4 or more times when the text is less than 4000 characters may be used as a frequent word. Hereinafter, the threshold value of the frequent word may be increased by one character every 1000 characters.

  The network interface unit 11 outputs data of a display screen for displaying the summary generated by the above process to the user terminal device 3.

  FIG. 5 is a configuration diagram of the advertisement server 4.

  The advertisement server 4 includes a network interface unit 41, a web server 43, a web page data storage unit 45, and an advertisement data storage unit 47, as shown in FIG.

  The web page data storage unit 45 stores web page data including objects such as images and scripts to be pasted on the web page. The web page data is described by a script such as HTML (Hyper Text Markup Language) or XML (extensible Markup Language).

  The advertisement data storage unit 47 stores a plurality of advertisement data entities related to a plurality of products or services. The advertisement data is described in, for example, HTML (Hyper Text Markup Language) or XML (extensible Markup Language). The advertisement data may include image data, for example. The advertisement data includes identification information regarding the category and item of each product or service. For example, each advertisement data entity includes identification information of a category such as “refrigerator” and “washing machine”, and item identification information by manufacturer name and model name (or model number).

  The network interface unit 41 communicates with other devices connected to the network 9 such as the user terminal device 3 via the network 9. For example, the network interface unit 41 receives an HTTP request from the user terminal device 3, and transmits web page data, advertisement data, and the like as a response to the HTTP request.

  The web server 43 receives a request from the user terminal device 3 and performs a process corresponding thereto. For example, the web server 43 acquires web page data from the web page data storage unit 45 based on an HTTP (Hyper Text Transfer Protocol) request from the user terminal device 3, and the advertisement target product or advertisement product 47 Get advertising image data for the service. The web server 43 outputs display screen data for displaying an advertisement image related to one or more advertisement products or services to the user terminal device 3 via the network interface unit 41. Here, the advertising product or service may be, for example, one or more products or services among the products or services for which the blog analysis server 1 has collected blog data.

  Next, an example of a processing procedure in the information providing system having the above configuration will be described using a flowchart.

  First, as a premise for performing the following processing, the blog data collection unit 13 of the blog analysis server 1 performs the processing shown in FIG.

  FIG. 6 shows an information provision processing procedure in the present embodiment. This information provision processing procedure will be described together with screen transitions in the user terminal device 3 shown in FIGS.

  First, the advertisement server 4 acquires the web page data stored in the web page data storage unit 45 and the advertisement data stored in the advertisement data storage unit 47 and transmits them to the user terminal device 3 (S400). This data transmission may be performed, for example, triggered by a request from the user terminal device 3.

  The user terminal device 3 receives the web page data and advertisement data sent from the advertisement server 4 and displays a web page including an advertisement image using a web browser (S421).

  FIG. 7 shows an example of this web page 500. As shown in the figure, the web page 500 includes a page content display area 501 and an advertisement image display area 502. Furthermore, in this embodiment, the user selection area 503 is provided in the web page 500, and when a user who wants to know the word-of-mouth information for the advertisement in the advertisement image display area 502 selects the user selection area 503, the following description will be given. The analysis result of word-of-mouth information is provided according to the procedure.

  When the web page 500 is displayed, for example, when the user receives a predetermined operation such as selecting the user selection area 503, the user terminal device 3 sends a word-of-mouth information analysis request to the blog analysis server 1. (S423). The word-of-mouth information analysis request includes information indicating the category of the advertisement product or service related to the advertisement displayed in the advertisement image display area 502. When the category can be specified by the identification information (item identification information) of the advertising product or service, the identification information of the advertising product or service may be included in the word-of-mouth information analysis request.

  When receiving the word-of-mouth information analysis request, the blog analysis server 1 executes blog analysis processing (S441). By this blog analysis process, frequent character strings appearing at a predetermined number of times or frequency are extracted from the blog data (word-of-mouth information) described for the advertising products or services and the products or services belonging to the same category as these. Details of the blog analysis process will be described later. The blog analysis server 1 generates a word-of-mouth information analysis result screen that displays the frequent character strings extracted by the blog analysis processing, and transmits the screen data to the user terminal device 3 (S443).

  The user terminal device 3 displays a word-of-mouth information analysis result screen including the frequent character string transmitted from the blog analysis server 1 (S425).

  FIG. 8 shows an example of the word-of-mouth information analysis result screen 600. As shown in the figure, the word-of-mouth information analysis result screen 600 includes a category display 601 indicating a category for which word-of-mouth information has been analyzed, and a frequent character string list 602 that is the result of word-of-mouth information analysis. In the frequent character string list 602, character strings are sorted by the number of appearances or the frequency.

  Thereby, the user can know what keywords are frequently used for the product or service in the same category as the advertisement product or service in the word-of-mouth information described in the blog. In other words, it is possible to know what a consumer is highly interested in regarding an advertisement product or service by using this keyword. For example, in the example of FIG. 8, it can be seen that the consumer is interested in “quiet”, “satisfied”, “design”, “function”, and the like among the characteristics of the refrigerator.

  When the word-of-mouth information analysis result screen 600 is displayed, for example, when the user selects one frequent character string from the frequent character string list 602, the user terminal device 3 accepts it (S427). Based on this selection, the user terminal device 3 transmits an item analysis request to the blog analysis server 1. This item analysis request includes information indicating the category displayed in the category display 601 and the selected character string.

  The blog analysis server 1 receives an item analysis request from the user terminal device 3 and performs an item analysis process (S445). In the item analysis, in the word-of-mouth information of each item in the category included in the item analysis request, the number or frequency of appearance of the selected character string for each item is counted, and the item ranking is generated according to the result. Details of the item analysis will be described later. The blog analysis server 1 generates an item ranking display screen based on the item analysis result, and transmits this screen data to the user terminal device 3 (S447).

  The user terminal device 3 displays an item analysis result screen including the item ranking transmitted from the blog analysis server 1 (S429).

  FIG. 9 shows an example of the item analysis result screen 700. As shown in the figure, an item analysis result screen 700 includes a category display 701 indicating a category subjected to item analysis, a display 702 indicating a selected character string used for item analysis, and an item ranking which is an item analysis result. 703.

  Thereby, the user can know a product or service in which many keywords selected by the user are included in the word-of-mouth information described in the blog regarding the product or service in the same category as the advertisement product or service. For example, in the example of FIG. 9, it is possible to know what manufacturers and models have a lot of word-of-mouth information related to “design” of “refrigerator”.

  When the item analysis result screen 700 is displayed, for example, when the user selects one item from the item ranking 703, the user terminal device 3 accepts it (S431). Based on this selection, the user terminal device 3 transmits to the blog analysis server 1 a request for generating a summary of word-of-mouth information related to the selected item. The summary generation request includes information indicating the selected item.

  Upon receiving this summary generation request, the blog analysis server 1 performs summary generation processing (S449). In this summary generation process, a summary of word-of-mouth information about items included in the summary generation request is generated. Details of the summary generation process will be described later. The blog analysis server 1 generates a summary display screen for displaying the generated summary, and sends this screen data to the user terminal device 3 (S451).

  The user terminal device 3 receives this screen data and displays a summary display screen.

  FIG. 10 shows an example of the summary display screen 900. The summary display screen 900 includes a summary display area 901 generated by the summary generation process, and a frequent word display area 902 in the word-of-mouth information subjected to the summary generation.

  Thereby, the user can see the summary (extract) of the content of the blog regarding the product or service designated by the user. That is, the user does not need to read the entire blog regarding the product or service designated by the user, and can extract and read the important part.

  FIG. 11 is a flowchart showing word-of-mouth information analysis request (blog data analysis request) processing.

  First, the network interface unit 11 receives a word-of-mouth information analysis request including a category of advertisement product or service from the user terminal device 3 (S31).

  The text analysis unit 17 reads all the files stored in the folder of the category related to the word-of-mouth information analysis request among the folders generated in the blog data storage unit 15 (S33).

  Then, the text analysis unit 17 performs a text analysis process on the text of the file read here (S35). The character string extracted by the text analysis process is stored in the extracted character string storage unit 19. The detailed processing procedure of this text analysis processing will be described later.

  The network interface unit 11 uses the data of the word-of-mouth information analysis result screen 600 that displays the extracted character string list obtained by sorting the character strings extracted in step S35 according to the number of appearances, and the user terminal device that has transmitted the word-of-mouth information analysis request 3 (S37).

  Thereby, the user can know in real time the character string (keyword) frequently used in the description about the category of the advertising product or service from the blog data including the word-of-mouth information. Note that many of the character strings extracted here are nouns.

  Next, FIG. 12 is a flowchart showing a detailed procedure of the text analysis process in step S35 of FIG.

  First, the specific character string processing unit 171 extracts one or more predetermined specific character strings from the read text, and counts the number of appearances of each. Then, the specific character string processing unit 171 stores the specific character string extracted here and the number of appearances thereof in the extracted character string storage unit 19 (S51). The specific character string processing unit 171 further replaces the specific character string extracted here with a blank in the read text (S53).

  Next, the erased character string processing unit 173 replaces one or more predetermined erase characters with a blank in the text in which the specific character string is replaced with a blank (S55).

  Next, the unnecessary character string processing unit 175 replaces one or more predetermined unnecessary character strings with blanks in the text in which the specific character string and the erased character are replaced with blanks (S57).

  Then, the cut-out processing unit 177 separates the character string delimited by the blank from the text in which the specific character string, the erased character, and the unnecessary character string are replaced with the blank (S59). The cutout processing unit 177 removes any character other than kanji at the beginning and end of the character string separated in step S59 (S61). Then, the number of appearances of each character string obtained here is counted, and each character string and the number of appearances are stored in the extracted character string storage unit 19 (S63).

  11 and 12, a character string extracted from the blog data related to the category specified by the user is output to the user terminal device 3 as a response to the word-of-mouth information analysis request (blog analysis request). The item analysis request process described below is a process related to a request including a selected character string selected from the extracted character strings output to the user terminal device 3.

  Next, FIG. 13 is a flowchart showing item analysis request processing.

  First, the network interface unit 11 receives from the user terminal device 3 an item analysis request including a character string selected by the user from the extracted character string list and a category for word-of-mouth information analysis (S71).

  The item ranking processing unit 21 reads the text file 152 included in the folder 150 of the category related to the item analysis request (S73). Then, the item ranking processing unit 21 counts the number of appearances of the selected character string selected by the user from the text read here for each file (S75).

  The item ranking processing unit 21 sorts the item names corresponding to the respective files in accordance with the counting result, generates data of the item analysis result screen 700 displaying the item ranking, and uses this as the transmission source of the item analysis request To the user terminal device 3 (S77).

  Thereby, the user can know in which item the selected character string is frequently used.

  FIG. 14 is a flowchart showing a summary sentence creation processing procedure.

  First, the network interface unit 11 receives a summary creation request including an item (product or service) selected by the user from the user terminal device 3 (S81).

  The summary generation unit 23 reads the file of the item related to the summary creation request among the folders generated in the blog data storage unit 15 (S83).

  The appearance count counting unit 231 divides the read text into words, and counts the number of appearances of each word (S84).

  The delimiter processing unit 232 replaces a predetermined symbol included in the read text (S85). For example, the symbols to be replaced include, for example, a blank “”, an asterisk “*”, a colon “:”, a punctuation mark “,”, a dot “•”, a circle “●” “○”, a double circle “◎”, a square “■”, “□”, black brackets, etc. may be used, and either half-width symbols or full-width symbols may be used. In this embodiment, the symbol to be replaced is replaced with a punctuation mark “.”.

  The delimiter processing unit 232 delimits the text with the replaced symbols for each sentence (S87). That is, the delimiter processing unit 232 detects the punctuation points in the text, and divides the detected punctuation points.

  The sentence selection processing unit 233 identifies one sentence to be processed among the sentences delimited at step S87 (S89). Then, it is determined whether or not the frequently used word is included in the target sentence (S91). As described above, the threshold value of the number of appearances that is a reference for frequent words may vary according to the text length.

  When the target sentence does not contain a frequent word (S91: No), it is determined whether or not the sentence end of the target sentence is a specific sentence end predetermined for each text attribute (S93). Examples of sentence endings by text attribute are as described above.

  When a frequent word is included (S91: Yes) and when the sentence end of the target sentence is a specific sentence end by text attribute (S93: Yes), the target sentence is selected as a sentence to be included in the summary ( S95).

  When the sentence end of the target sentence is not a specific sentence end by text attribute (S93: No), the process for the target sentence is terminated, and it is determined whether or not the processing of all the sentences included in the read text is terminated. (S97). If the processing of the whole sentence has not been completed, the process returns to step S89 to repeat the processing (S97: No). When the processing for the whole sentence is completed (S97: Yes), the blog analysis server 1 selects the one selected in step S95. The above sentence is output as summary to the user terminal device 3 that is the source of the summary creation request (S99).

  In the above-described processing, either one of step S91 and step S93 may be omitted. That is, the sentence selected by the sentence selection processing unit 233 may be only a sentence including a frequent word or only a sentence having a specific sentence ending. Further, step S85 may be omitted.

  According to the embodiment of the present invention described above, it is possible to quickly extract a character string that characterizes the contents described in text such as a blog.

  Further, according to the embodiment of the present invention, it is possible to statistically process a character string extracted from text such as a blog.

  Moreover, according to the embodiment of the present invention, the advertising effect can be enhanced by using the word-of-mouth information.

  Moreover, according to the embodiment of the present invention, it is possible to provide word-of-mouth information related to a product or service related to an advertisement.

  Further, according to the embodiment of the present invention, it is possible to select and provide information effective for the user from the word-of-mouth information related to the product or service related to the advertisement.

  For example, when an Internet user sees an advertisement, the user can efficiently know word-of-mouth information about a product or service related to the advertisement. Thereby, the user can make use of the decision as to whether or not to purchase the advertising product or service. In addition, if the merchant of the product can objectively show that there is a lot of positive word-of-mouth information about the product related to the advertisement, it can give the user a strong motivation to purchase and increase the advertising effect. Can do.

  The above-described embodiments of the present invention are examples for explaining the present invention, and are not intended to limit the scope of the present invention only to those embodiments. Those skilled in the art can implement the present invention in various other modes without departing from the gist of the present invention.

  For example, in the embodiment described above, text processing for blog data has been described, but the present invention can also be applied to text other than blogs.

DESCRIPTION OF SYMBOLS 1 Blog analysis server 3 User terminal device 4 Advertising server 5 Blog server 11 Network interface part 13 Blog data collection part 15 Blog data storage part 17 Text analysis part 19 Extracted character string storage part 21 Item ranking processing part 23 Summary generation part 41 Network interface Unit 43 Web server 45 Page data storage unit 47 Advertising data storage unit 171 Specific character string processing unit 173 Erase character string processing unit 175 Unnecessary character string processing unit 177 Cutout processing unit 231 Appearance count unit 232 Separation processing unit 233 Sentence selection processing unit 500 Web page 600 Word-of-mouth information analysis result screen 700 Item analysis result screen 900 Summary display screen

Claims (8)

A text storage unit for storing a file including text including a plurality of sentences;
Upon receiving a text summary creation request in a file stored in the text storage unit, a delimiter processing unit that divides the text in the file into sentences,
A selection processing unit that selects a sentence ending with a predetermined end-of-sentence character determined according to the attribute of the text among a plurality of sentences separated by the separation processing unit;
An abstract creation device comprising: output means for outputting a sentence selected by the selection processing unit as a response to the summary creation request.   The summary creation device according to claim 1, wherein the selection processing unit selects a sentence including a predetermined frequent word or more in the text among a plurality of sentences delimited by the delimiter processing unit.   When the text is a blog describing a product or service, the end-of-sentence characters are at least “foot”, “profit”, “mind”, “su”, “i”, “ri”, “ru”, “ru” The summary creation device according to claim 1, comprising “yo” and “n”.   3. The summary creation device according to claim 1, wherein when the text is a newspaper article, the end-of-sentence character includes at least “su”, “ta”, “ru”, and “dan”.   5. The summary creation device according to claim 4, wherein when the text is a newspaper article, the selection processing unit selects a sentence having a sentence beginning with “(” and a sentence ending with “”.   When the text is a newspaper article, the end-of-sentence character has one of “in this”, “just”, and “other” at the beginning of the sentence, and “s”, “ta” at the end of the sentence. 6. The summary creation device according to claim 4, wherein a sentence immediately before a sentence that is any one of “RU” and “DA” is selected. A computer-generated text summarization method,
Storing a file including text including a plurality of sentences in a text storage unit;
Receiving a text summary creation request in a file stored in the text storage unit, dividing the text in the file into sentences;
Selecting a sentence ending with a predetermined end-of-sentence character determined according to an attribute of the text among the plurality of separated sentences;
Outputting the selected sentence as a response to the summary creation request. A computer program for creating a text summary,
Storing a file including text including a plurality of sentences in a text storage unit;
Receiving a text summary creation request in a file stored in the text storage unit, dividing the text in the file into sentences;
Selecting a sentence ending with a predetermined end-of-sentence character determined according to an attribute of the text among the plurality of separated sentences;
A computer program for causing a computer to execute the step of outputting the selected sentence as a response to the summary creation request;

Images