JP6260678B2 - Information processing apparatus, information processing method, and information processing program - Google Patents

Description

  The present invention relates to a technique for analyzing data.

In recent years, with the spread of the Internet, services such as bulletin boards on the Internet and social network services (SNS: Social Network Service) that allow users to easily upload word-of-mouth information and publish the text Is increasing. In addition, grasping such word-of-mouth information on the Internet is attracting attention from the viewpoint of corporate marketing strategies.
However, since sentences on the Internet uploaded by individual users have many omitted phrases and notations, there is a problem that it is difficult to quickly find an appropriate keyword from such sentences. As a technique for dealing with such a problem, there is a technique such as Japanese Patent Application Laid-Open No. 2011-3157 (Patent Document 1).

JP 2011-3157 A

  Patent Document 1 describes a technique for analyzing text data, specifying items that are products or services, and summarizing user's word-of-mouth information for each item. However, the accuracy of determining which item the text data to be analyzed corresponds to is not necessarily sufficient. For example, if the object of description is music or a movie, the names are very diverse, and there is no clear regularity in the character string indicating the name, so the accuracy of identifying the item that is the object of description is high In some cases it was not enough. For this reason, there are cases where the item to be described cannot be specified in the text data, or an item that is different from the item that is actually the description target may be specified.

  The present invention has been made in view of such problems, and an object thereof is to accurately specify information to be described in text data.

In order to solve the above-described problems of the prior art, the present invention searches a database and acquires a plurality of information corresponding to a search condition, and a similarity for calculating a score based on the similarity between the plurality of information There is provided an information processing apparatus comprising: a degree calculation unit; and a validity determination unit that determines validity of the search condition based on the score.
Further, since the present invention is to solve the problems of the prior art described above, it acquires a plurality of search result set searches the database, corresponding to the plurality of search conditions Ri Do a plurality of information using multiple search criteria a search unit that, said based on the similarity between a plurality of information included in the search result set, the similarity calculation unit for calculating a plurality of scores corresponding to a plurality of said search result set, the score is high There is provided an information processing apparatus comprising a validity determination unit that preferentially outputs information included in the search result set .
Further, the present invention is an information processing method executed by one or a plurality of computers to solve the above-described problems of the prior art, and a search step of searching a database and acquiring a plurality of information corresponding to a search condition A similarity calculation step for calculating a score based on the similarity between the plurality of pieces of information acquired in the search step, and the validity of the search condition based on the score calculated in the similarity calculation step. An information processing method characterized by including a validity determination step for determination.
Further, the present invention is to solve the problems of the prior art described above, an information processing method of one or more computers to perform searches the database by using a plurality of search conditions, Ri Do from multiple wherein the plurality of search conditions set of search results corresponding to the respective the search step acquires a plurality, and based on the similarity between a plurality of information included in the search result set obtained at the retrieval step, a plurality of the search A similarity calculation step for calculating a plurality of scores corresponding to each result set; and a validity determination step for preferentially outputting information included in the search result set having a high score calculated in the similarity calculation step; The information processing method characterized by including this is provided.
Further, in order to solve the above-described problems of the conventional technology, the present invention searches one or a plurality of computers for a database and acquires a plurality of information corresponding to a search condition, and the search unit acquires the information. A similarity calculation unit that calculates a score based on a similarity between a plurality of pieces of information, and a validity determination unit that determines the validity of the search condition based on the score calculated by the similarity calculation unit An information processing program is provided.
Further, since the present invention is to solve the problems of the prior art described above, one or more computers to search the database using a plurality of search conditions, corresponding to the plurality of search conditions Ri Do from multiple search unit acquires a search result set, and based on the similarity between a plurality of information contained in the acquired set of search results at the search unit, a plurality of scores corresponding to a plurality of said search result set An information processing program that functions as a validity determination unit that preferentially outputs information included in the search result set having a high score calculated by the similarity calculation unit, the similarity calculation unit to calculate provide.

  According to the present invention, information to be described in text data can be specified with high accuracy.

It is a block diagram for showing the whole structure in each embodiment. It is a figure which shows the example of article text (text data). It is a flowchart which shows operation | movement of the text information processing apparatus 1 of 1st Embodiment. It is a figure for demonstrating an example of the method of extracting a keyword from article text. It is a figure which shows an example of the data which the text data storage part 6 stores. It is a figure which shows an example of the data which item DB3 stores. It is a figure which shows an example of the data which the keyword group memory | storage part 5 stores. It is a figure which shows an example of the data which the score memory | storage part 7 stores. It is a figure which shows an example of the data which the item calculation result storage part 8 stores. It is a figure which shows an example of the data which the item ranking information storage part 9 stores. It is a flowchart which shows operation | movement of the text information processing apparatus 1 of 2nd Embodiment. It is a flowchart which shows operation | movement of the text information processing apparatus 1 of 2nd Embodiment.

Hereinafter, a text information processing apparatus, a text information processing method, and a text information processing program according to the present invention will be described with reference to the drawings. In the drawings, the same components are denoted by the same reference numerals.
In addition, the items in the following description may be contents such as voice, music, video, web pages and various articles, and may be information on financial products, real estate, persons, and the like. The items in the following description may be paid or free regardless of whether they are tangible or intangible.

<First Embodiment>
FIG. 1 is a block diagram illustrating a configuration example of the entire system including the text information processing apparatus 1 according to the first embodiment.
This system includes a text information processing apparatus 1, a text data server (blog server) 2, an item database (item data server) 3, a user terminal device 4, and the like, each of which can communicate via a network 20. It is. Note that the text information processing apparatus 1 is a server, for example.
The text data server 2 stores text data, and the item database 3 stores information about items.
In the following description, blog data is used as an example of text data processed by the text information processing apparatus 1. Blog data includes text data created by a user. For example, it includes text data (blog articles) created by a user using a social network service. Examples of social network services include Twitter (registered trademark), Facebook (registered trademark), and mixi (registered trademark).

  Further, although the text data server 2 and the item database 3 are described as different entities, some or all of them may be configured to be the same entity as the text information processing apparatus 1.

The text information processing apparatus 1 includes four processing units: a text data collection unit 10, a keyword set generation unit 11, an item specification unit 12, and a ranking information creation unit 13. These four processing units may be integrated, or may be separate from each other. In addition, a single CPU or DSP may be used, or a plurality of CPUs or DSPs may be used.
The text information processing apparatus 1 includes a keyword group storage unit 5, a text data storage unit 6, a score storage unit 7, an item calculation result storage unit 8, and an item ranking information storage unit 9. These five storage units may be integrated, or may be separate from each other. In addition, a single hard disk drive (HDD), a flash memory, or the like may be used, or a plurality of HDDs, flash memories, or the like may be used.

The text data collecting unit 10 acquires attribute information such as an article text (text data) such as a blog, a user identifier indicating the creator, and an article creation update date from the text data server 2 storing the text data. And stored in the text data storage unit 6. The user identifier is an identifier for identifying a user related to creation of text data or a terminal device related to creation of text data. Note that the text data storage unit 6 is not always necessary, and the text data server 2 may also serve as the text data storage unit 6.

The keyword set generation unit 11 includes an unnecessary character string processing unit 14, a keyword extraction unit 15, and a grouping processing unit 16. The keyword set generation unit 11 has a role of extracting a keyword for specifying an item from the text data acquired by the text data collection unit 10 and generating a keyword group (search key). Although details will be described later, a search is performed using this keyword group.
The unnecessary character string processing unit 14 of the keyword set generation unit 11 generates text data excluding unnecessary information that is not related to item information. Unnecessary information not related to item information is information such as document link information and meta tags, for example. The processing in the unnecessary character string processing unit 14 will be described in detail later.

The keyword extraction unit 15 of the keyword set generation unit 11 extracts keywords from the text data processed by the unnecessary character string processing unit 14.
The grouping processing unit 16 of the keyword set generation unit 11 groups one or a plurality of keywords extracted by the keyword extraction unit 15, and groups the keyword group that is a set of one or more keywords into a keyword group. Save to the storage unit 5. Even when only one keyword is included, it is referred to as a keyword group.

  The item specifying unit 12 includes a search unit 17, a similarity calculation unit 18, and a validity determination unit 19, and uses the keyword group generated by the keyword set generation unit 11 to retrieve item information from the item database 3. It has a role of searching and determining the validity of a keyword from the similarity between a plurality of item information obtained from the search result.

  The search unit 17 of the item specifying unit 12 searches the item database 3 using the keyword group generated by the keyword set generation unit 11. And when the search result set which consists of several item information is obtained, the similarity calculation part 18 of the item specific | specification part 12 calculates the similarity between several item information. Further, the similarity calculation unit 18 obtains a score related to the search result set by a calculation formula to be described later for each keyword group using the similarity between the plurality of item information, and records the score in the score storage unit 7.

  The validity determination unit 19 of the item specifying unit 12 compares the score calculated by the similarity calculation unit 18 with the threshold θ, and determines the validity of the keyword group used for the search of the item database 3. Then, an item related to the article text (text data) is specified using a search result set corresponding to the keyword group determined to be valid. The validity determination unit 19 associates the identified item (item identifier) with the blog identifier of the original text data from which the keyword group is extracted, and records it in the item calculation result storage unit 8. In addition, when there are a plurality of keyword groups determined to be appropriate, an item may be specified using a search result set corresponding to the keyword group with the highest score among them, and the plurality of keyword groups An item may be specified using all of a plurality of corresponding search result sets.

  The ranking information creation unit 13 performs ranking (ranking) based on the number of appearances of items calculated using the data in the item calculation result storage unit 8 and records the ranking in the item ranking information storage unit 9. In addition, although it is not provided with the ranking information creation part 13, the information used as the description object in text data can be specified with a sufficient precision, By providing the ranking information creation part 13, the analysis by the text information processing apparatus 1 is possible. The results can be output in a more useful format.

  The text information processing apparatus 1 is configured using a general computer including a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), an HDD (Hard Disk Drive), a network interface, and the like. May be. That is, you may make it function as the text information processing apparatus 1 by making a computer run the program which performs a process which is demonstrated later.

  Further, the text information processing apparatus 1 may be configured using a plurality of computers. For example, in order to distribute the load, a plurality of computers corresponding to a certain processing block of the text information processing apparatus 1 may be used, that is, a plurality of computers having the same processing block may be used to perform the distributed processing. Good. Further, the distributed processing may be performed in such a manner that some processing blocks of the text information processing apparatus 1 are implemented by a computer and other processing blocks are implemented by another computer.

  Specific processing of the text information processing apparatus 1 will be described in detail using an example of text data shown in FIG. 2, a flowchart shown in FIG. 3, and data configuration diagrams shown in FIGS.

  Below, the example which specifies the item information which shows the music based on the article text (text data) regarding a music, and produces the ranking information based on the specified item information is demonstrated. As described above, the items are not limited to music, but may be various contents, articles, and services.

FIG. 3 shows a processing flow of the text information processing apparatus 1.
In step S <b> 1, the text data collection unit 10 acquires text data from the text data server 2 and stores the acquired text data in the text data storage unit 6.

Specifically, the text data collection unit 10 transmits a predetermined request command to the text data server 2 to thereby include blog data including a user identifier, article text (text data), an article creation update date, and the like. Is received (acquired). The received data is stored in the text data table of the text data storage unit 6.
At this time, the text data collection unit 10 assigns one piece of identification information (blog identifier) to each article text. An example of the storage format of the text data table is shown in FIG. A blog identifier, a user identifier, an article text, and an article creation update date (for example, uploaded date and time) are stored in association with each other. For example, a blog identifier is attached to each text data transmitted to the text data server 2 by the user once uploading.

  The notation of the blog identifier in the present embodiment is a character string “BlogID” + underscore symbol (_) + number serial number in order of article creation update date, but may be user ID + number serial number, The article acquisition date and time may be a serial number. It is only necessary to uniquely identify each blog data. When the text data server 2 includes a blog identifier (or data corresponding to the blog identifier) and the text data collection unit 10 receives (acquires) the data, the text data collection unit 10 uses the blog identifier. May be omitted, and the received blog identifier may be used.

  The blog data may be read by specifying a required article creation update date range (period) with a request command and acquiring corresponding data. Similarly, a user identifier necessary for the request command may be specified to acquire only the article data of the user. Further, a search expression related to a character string may be included in the request command, and only blog data in which a specific character string pattern is included in the article text may be acquired.

(Operation of the keyword set generation unit 11)
Returning to FIG. 3, in step S <b> 2 to step S <b> 5, keyword set generation processing by the keyword set generation unit 11 is executed.

  First, in step S <b> 2, the keyword set generation unit 11 reads (acquires) text data for each blog identifier from the text data table in the text data storage unit 6. In the subsequent processing, processing is performed for each text data.

  In step S <b> 3, the unnecessary character string processing unit 14 replaces a character string (referred to as an unnecessary character string FW) that is not useful for specifying an item among characters from the beginning to the end of the text data with a predetermined delimiter K. . For example, a symbol (including a combination of a plurality of symbols) that is less likely to appear in the article text, such as “\\”, may be used as the delimiter K. Unnecessary character strings may be deleted without being replaced, or may be replaced with white space characters (for example, a space symbol, a tab symbol, etc.). This is preferable because it is useful for cutting out the. The predetermined delimiter K need not always use the same symbol, and may be changed as appropriate according to the text data. For example, the delimiter may be changed according to the language type or character type of the text data.

Here, the details of the processing performed by the unnecessary character string processing unit 14 in step S3 will be described with reference to FIGS.
FIG. 2 is a diagram showing an example of article text (text data) used for specifying item information. In the example of FIG. 2, one or more normal character strings W, a specific symbol TK, and an unnecessary character string FW are included between the beginning S and the end E of the text data. However, the specific symbol TK and the unnecessary character string FW are not always present. Further, there may be a plurality of specific symbols TK and unnecessary character strings FW. The keyword extraction unit 15 extracts the normal character string W other than the specific symbol TK and the unnecessary character string FW. This extraction method will be described later. A single character is also referred to as a character string. Further, the normal character string W is a character string that may be useful for specifying an item, for example, a character string other than the specific symbol TK and the unnecessary character string FW in the text data.

  FIG. 5 is a diagram illustrating an example of data (text data table) stored in the text data storage unit 6. As shown in FIG. 5, in the text data table, the article text, the blog identifier assigned to the article text, the user identifier indicating the user who uploaded the article text, and the article creation indicating the update date when the article text was uploaded Stored in association with the update date. As shown in the article text in FIG. 5, the various words and expressions used in various texts created by users such as blogs are very diverse.

  In general, a character string useful for specifying an item and an unnecessary character string are mixed. In the example shown in FIG. 5, “#NowPlaying” is a character string that conventionally indicates that the article is related to the reproduction of music or video content. Since this is the same character string in articles related to any item, it is not useful for specifying an item and becomes an unnecessary character string FW.

  In addition, in the text in a service (microblog service) where a relatively short article text such as Twitter is often uploaded, a URL (Uniform Resource Locator) indicating a link to another site is frequently included. Since the URL character string often does not include an item name or the like, the URL character string starting with “http: //” or the like is treated as an unnecessary character string because it is not useful for item identification. In particular, since the item name or the like is often not included in the character string of the shortened URL, it may be handled as the unnecessary character string FW only for the shortened URL.

Also, meta tags (character strings enclosed by “<” and “>”) and marks (♪), which often do not include item names, are treated as unnecessary character strings FW. These may be either half-width or full-width.
The unnecessary character string processing unit 14 refers to a database that stores a list of unnecessary character strings FW, a condition of character strings to be regarded as unnecessary character strings FW, and the like, and the above-mentioned unnecessary character strings FW are included in text data. Determine whether or not. The unnecessary character string processing unit 14 replaces it with a predetermined delimiter K.

The unnecessary character string processing unit 14 does not replace unnecessary character strings with blank characters (for example, space symbols, tab symbols, etc.), but replaces them with predetermined symbols that are less likely to be used in blog articles, etc. Keywords useful for identifying items can be extracted with high accuracy.
For example, as shown in FIG. 4A, “M1: title, M2: blank, M3: URL, M4: blank, M5: artist (last name), M6: blank, M7: artist (first name), M8: #NowPlaying If there is an article text with a pattern of “”, as shown in FIG. 4B, replacing the unnecessary character string FW M3: URL or M8: #NowPlaying with a blank character is a keyword useful for specifying an item. If there is a space between the character string M5 and the character string M7, it is difficult to determine whether the character string M5 and the character string M7 are handled as one keyword.

In other words, although it is more advantageous to specify the item by extracting the character string M5: artist (last name) and the character string M7: artist (first name) as one keyword, the unnecessary character string FW is replaced with a blank. If so, such string integration is difficult.
On the other hand, as shown in FIG. 4C, if the unnecessary character string FW M3: URL or M8: #NowPlaying is replaced with a delimiter K ("\\" in this example), a blank space Since it is only necessary to disregard the text data with this delimiter K, the character string M5 and the character string M7 can be integrated and handled as one keyword, and the item can be specified with higher accuracy. Note that, regardless of the number of characters in the unnecessary character string FW, “\\” is replaced, but each character constituting the unnecessary character string FW may be replaced with “\\”.

  Note that the excluded characters, punctuation marks, and the like described in Patent Document 1 can also be handled as unnecessary character strings FW. Examples of excluded characters described in Patent Document 1 include “NO”, “GA”, “I”, “KU”, and the like.

  Next, specific symbols will be described. There is no clear rule in the order and format for describing the song name and artist name in the text data related to the music playback that is the subject of this embodiment, but as shown in the text data of FIG. 2 and FIG. The hyphen "-" or slash "/" is often used as a symbol to separate text and artist. In the present embodiment, this symbol is referred to as a specific symbol TK. The specific symbol TK may or may not exist in the text data.

  When the unnecessary character string processing unit 14 performs processing for replacing the unnecessary character string FW with a predetermined delimiter K, the specific symbol TK may be left as it is, or may be replaced with the delimiter K as the unnecessary character string FW. Before and after the specific symbol, there is a relatively high possibility that there is a character string useful for specifying an item such as a song name or artist name. Therefore, there are cases where keyword extraction can be performed with high accuracy by using the specific symbol TK. . On the other hand, the keyword extraction process can be simplified by replacing the specific symbol TK with the delimiter K.

In addition, if the item information of the item to be described in the text data is in Japanese, the item information (for example, if it is music content, the Japanese title, the Japanese artist name, etc.) contains blank characters If the text data is Japanese, using the feature that the possibility of being generated is relatively low, all white space characters may be replaced with delimiters. Alternatively, in the case of Japanese, processing may be performed in which white space characters are deleted and character strings before and after the white space characters are connected.
The above is the details of the processing by the unnecessary character string processing unit 14.

Returning to the description of FIG. 3, in step S <b> 4, the keyword extraction unit 15 extracts keywords. With the delimiter K as a delimiter, the text area from the beginning S to the character immediately before the first delimiter K, the text area sandwiched by the delimiter K, and the character following the last delimiter K from the end of the sentence E And the character strings included in these text regions are used as keywords. Note that there are often a plurality of text regions sandwiched between the delimiters K. Further, when the specific symbol TK is used, characters included in any text area between the specific symbol TK and the delimiter K, between the specific symbol TK and the head S, or between the specific symbol TK and the sentence end E. A column may be preferentially a keyword. By performing such processing, the accuracy of keyword extraction can be further increased.
When the unnecessary character string processing unit 14 performs processing for replacing the unnecessary character string FW with a blank character, the text data is separated at the position of the blank character to extract keywords.

Note that the character type (kanji, hiragana, katakana, alphabet, numbers, etc.) of the text area may be determined to determine whether or not to include a blank character in the keyword. For example, when the text area is mainly composed of alphabetic character types, the character string before and after the blank is extracted as one keyword without performing the process of connecting the character strings before and after the blank. For example, in the example shown in FIG. 4C, “M5 artist (last name), M6 blank, M7 artist (name)” is extracted as one keyword.
On the other hand, in the case of mainly composed of kanji, hiragana, and katakana, the process of connecting the character strings before and after the blank is performed, and then the character strings before and after are extracted as one keyword. For example, in the example shown in FIG. 4C, “M5 artist (last name), M7 artist (name)” is extracted as one keyword.

  It should be noted that it is preferable that the beginning S and end E of the keyword are not blank characters. When a blank character is not included in the keyword, it is preferable to extract a character string other than the blank closest to the specific symbol as a keyword.

  Further, only a character string having a specific length may be used as a keyword. For example, a keyword may be extracted by providing a criterion (condition) such as using a character string of 5 to 15 characters as a keyword. In this case, the length condition of the character string used as the keyword may be changed according to the character type. For example, in a character string using alphabets, the character string of one word tends to increase. Therefore, when the total length of non-blank characters and blank characters is 7 to 20 characters, the keyword is used. Such a condition may be set.

  In the case of a character string containing many kanji characters, the length of the character string as a keyword is set shorter than in the case of other character types, and a condition such as 2 to 10 characters is set. Good. When the specific symbol TK is used, the keyword extraction condition may be changed between a text region adjacent to the specific symbol and a text region located away from the specific symbol. For example, in a text area adjacent to a specific symbol, the length condition of the character string as a keyword is relaxed (for example, 3 to 20 characters), and in a text area located away from the specific symbol, the character string Processing such as strict length conditions (for example, 6 to 12 characters) may be performed.

  As described above, for example, J (J ≧ 1) keywords are extracted from one article text in step S4.

In the next step S5, the grouping processing unit 16 creates a keyword group for each article text using the keyword for each article text created in step S4.
When the number of keywords is one (J = 1), one keyword group is created. When there are a plurality of keywords (J ≧ 2), a plurality of keyword groups are basically created. The number of keywords included in one keyword group is an arbitrary number of 1 or more.
Here, a method for creating a keyword group will be described by taking four keywords K1, K2, K3, and K4 extracted from the text data shown in FIG. 2 as an example.

First, the case where a keyword group is created so that one keyword group includes one keyword will be described. Such a case is also referred to as a keyword group.
The grouping processing unit 16 sets the keywords K1, K2, K3, and K4 as separate keyword groups. Then, each created keyword group is assigned a keyword group identifier that can be identified, and stored in the keyword group storage unit 5 in a format as shown in FIG. FIG. 7 is an example of a search keyword group table based on the text data shown in FIG.

  The grouping processing unit 16 assigns keyword group identifiers Gr001-001, Gr001-002, Gr001-003, and Gr001-004 to the keywords K1, K2, K3, and K4, respectively. In this example, the part before the hyphen “-” of the keyword group identifier is a character string determined by the blog identifier, and “Gr001” corresponds to “BlogID_001”. Alternatively, the blog identifier may be directly used in the first half of the keyword group identifier, such as “BlogID_001-001”. The portion after the hyphen “-” of the keyword group identifier is a serial number, but it may be a serial number in the order of creation, or may be an article acquisition date + a serial number. Then, the grouping processing unit 16 stores the keyword group identifier, the blog identifier, and the keywords included in the keyword group in association with each other.

  Next, the case where a keyword group is created so that two keywords are included in one keyword group will be described.

The grouping processing unit 16 is a combination of selecting two keywords from four keywords, “K1 and K2”, “K1 and K3”, “K1 and K4”, “K2 and K3”, “K2 and K4”, “ Six keyword groups “K3 and K4” are created.
In the example illustrated in FIG. 7, the grouping processing unit 16 includes “K1 and K2”, “K1 and K3”, “K1 and K4”, “K2 and K3”, “K2 and K4”, and “K3 and K4”. Keyword group identifiers Gr001-005, Gr001-006, Gr001-007, Gr001-008, Gr001-009, and Gr001-010 are assigned respectively.
When a plurality of keywords exist in one keyword group, each keyword may be stored as one character string concatenated with a blank character, or each keyword may be stored in a format that can be read out separately.

  As in the present embodiment, when an item is a song, two character strings of a song name and an artist name are often useful for specifying the item. Therefore, in the case of a keyword group including two keywords, it is often possible to accurately specify information to be described. However, the number of keyword groups is larger and the processing amount is larger than when a keyword group is created so that one keyword group includes one keyword.

  It should be noted that the grouping processing unit 16 has a keyword group including a first number (one in the example of FIG. 7) of keywords and a second number (two in the example of FIG. 7) larger than the first number. If both the keyword group including the keyword is created, the information to be described can be specified with high accuracy while suppressing the processing amount as will be described later. Although not shown in FIG. 7, a column holding priority or priority may be added to the search keyword group table to give priority (priority) to each keyword group. In step S6 described later, the item database 3 may be searched according to the priority (priority order). That is, after performing a search using the keyword group having the highest priority, the search may be performed using the keyword group having the second highest priority.

  As a method of assigning priority (priority order), it is possible to use the degree to which each keyword satisfies the keyword criteria (conditions) relating to the length of the character string and the character type. In addition, a process for increasing the priority of a keyword extracted from a character string near the specific symbol TK may be performed.

(Operation of the item specifying unit 12)
Returning to FIG. 3, in step S <b> 6, the search unit 17 of the item specifying unit 12 sequentially reads out the keyword groups from the keyword group table stored in the keyword group storage unit 5 and creates a search expression for each keyword group. Then, a search request is transmitted to the item database 3.

The item database 3 in the present embodiment stores an item table configured as shown in FIG. When the item database 3 receives the search request, the item database 3 searches the item table, and when at least one of the title column and the artist column matches the condition (search formula) specified in the search request, information on the item is stored. (Title, artist name, etc.) is transmitted to the text information processing apparatus 1. The item identifier may be included in the information transmitted to the text information processing apparatus 1.
Further, if a search model such as a vector space model is used, even if a search keyword is not included in the item information, the item information can be used as a search output. The search unit 17 of the item specifying unit 12 acquires a list of item information based on the search formula included in the search request.

Data (item information list) that can be acquired from the item database 3 in correspondence with one search expression (one search) is referred to as a search result set (search result list). When there is an item that matches the search formula, the search result set includes one or more item information. The item information acquired by the search is also simply referred to as “search result”.
Further, the item database 3 according to the present embodiment interprets that a plurality of keywords are combined with the AND condition when the AND or OR condition is not specified in the search formula and a plurality of keywords are specified. Further, the item database 3 may transmit search results with priorities when there are a plurality of items that match the search formula. For example, the item with the highest priority may be the first search result, the item with the second highest priority may be the second search result, and the order of the search results may be determined in the same manner.

Here, the priority order may be calculated using the degree of matching between the search expression and the item information, or may be calculated using the item popularity. For example, the number of times to be output as a search result may be counted for each item, and the number of times may be set as a popularity level, and processing for increasing the priority of items having a high popularity level may be performed. Further, the popularity may be calculated using information such as the number of times the item is used and the sales amount of the item, which can be acquired from the outside. Furthermore, the text information processing apparatus 1 calculates the popularity for each item based on ranking information to be described later with respect to the item database 3, and periodically provides this information to the item database 3. It may be used to determine priority.
Thus, although the search part 17 and the item database 3 cooperate and perform a search process, either may carry out independently.

The search unit 17 uses one keyword group for each search. When the keyword group includes a plurality of keywords, a search expression is created so as to satisfy an AND condition using them. A set of keywords used in one search expression is called a search key. In the present embodiment, the keyword group corresponds to the search key. If the search expression does not include an AND or OR condition and the search expression is composed of only one or more keywords, it can be said that the search expression and the search key are equivalent.
For example, when a search is performed with a keyword group including only one keyword, item information (here, title and artist name) including the keyword in at least one of the title and artist is output.

  As shown in FIG. 7, when a search is made with a keyword group having only the keyword “song” with the keyword group identifier Gr001-001, a search result including “song” in the title and artist name is output. For example, titles including “Love Song / Z Yama Toro”, “Graduation Song / Y Band”, “Song Song / C & A”, “Spring Song / A Band”, “Summer Song / A Band”, etc. A list of artist names is output as a search result.

When two keywords (K1, K2) are included in the keyword group, a search expression such as (K1 AND K2) is created. That is, information indicating an item in which the keyword K1 is included in at least one of the title and the artist and the keyword K2 is included in at least one of the title and the artist is output.
For example, as shown in FIG. 7, when searching for a keyword group including “song” and “A band” of the keyword group identifier Gr001-006, “song” is included in the title and artist name, A search result including “A band” in the artist name is output. For example, a list of titles and artist names “Haruka / A band” and “Natsuka / A band” is output.

  In the next step S7, the item specifying unit 12 normalizes each item information included in the search result set. This normalization is performed as a process corresponding to this because the item database 3 may return substantially the same item as different search results. When the item is a song, even if the item is substantially the same, a plurality of patterns of song name notation may be used.

  For example, in the item database 3 for one song “title A / artist B”, “title A (version C) / artist B”, “title A / artist B (featuring X)”, “title A / artist B with X” May return multiple search results. In particular, when an item data table is created based on music information created and provided by many users, such a phenomenon is likely to occur. By normalizing the item information, it is possible to combine the above music notation variations into one. Specifically, for each item information (title and artist name) character string included in the search result set, a predetermined character string is deleted, or a character string is converted to create a normalized character string. To do. For example, a character string enclosed in parentheses (a character string surrounded by “(” and “)”) may be deleted.

  Alternatively, a character string frequently used to supplement the artist name such as “featuring” or “with” may be registered in advance, and the character string and subsequent characters may be deleted from the artist name as a search result. Also, conversion processing of character types such as half-width katakana to full-width katakana, full-width alphabet to half-width alphabet, and full-width number to half-width number may be performed. The normalization process does not necessarily have to be performed, but by performing such a search result normalization process, the text data and the item can be associated with each other more accurately.

  Next, in step S8, the item specifying unit 12 calculates the similarity between each item information included in the search result set using the normalized item information created in step S7, and calculates the average value of the calculation results. Is calculated as a score. Then, the calculated score is stored in the score column of the search result score table shown in FIG. 8 in association with the keyword group identifier. As a result of the search, if the corresponding item is not found (when the search result set is an empty set), the score for the keyword group is not stored.

  Next, a score calculation method will be described. For example, when three items of information of (1) “Spring Song / A Band”, (2) “A Song / A Band”, and (3) “Summer Song / A Band” are output as a search result set. (1) “Haruka / A band” and (2) “Song / A band” similarity, (1) “Haruka / A band” and (3) “Natsuka / A band” The similarity of (2) “A song / A band” and (3) “Natsuka / A band” is calculated. Then, an average value of these three similarities may be calculated as a score. Thus, if the similarity is calculated for all combinations of item information included in the search result set, the score can be calculated with high accuracy, but the processing amount increases.

  Also, select one reference item (reference search result) from the items in the search result set, calculate the similarity between the reference item and other items in the search result set, and score the average value of them May be calculated as For example, (1) “Haruka / A band” is a reference item, (1) “Haruka / A band” and (2) “A song / A band” similarity, (1) “Haruka / A band” Two similarities of “A band” and (3) “Natsuka / A band” similarity may be calculated, and the average value thereof may be used as a score. In this way, compared to the case where the similarity is calculated for all combinations of item information included in the search result set, the accuracy of the score is reduced, but the processing amount is small. When there is a lot of item information included in the search result set, a method using a reference item is desirable.

  When only two item information items are included in the search result set, the similarity between the two item information items may be used as a score as it is. If only one item information is included in the search result set, the similarity and score calculation processing may not be performed, and the item information in the search result set may be associated with the blog identifier.

Various methods can be used for the similarity calculation. For example, morphological analysis processing is performed on N normalized search results (N ≧ 2), and words are extracted. At this time, processing such as extracting a specific part of speech such as a noun or an adjective or excluding a particle or an auxiliary verb may be performed. When a total of M types of words (1 to M words) can be extracted, the search results (item information) correspond to the rows of the matrix and the words correspond to the columns. ) Is created as a matrix element. Alternatively, the matrix element may be “1” when a certain word appears in a certain search result and “0” when it does not appear.
In the following, the elements of the occurrence matrix are represented as dij (i = 1 to N, j = 1 to M). i represents the i-th row of the matrix, and j represents the j-th column of the matrix.

  Here, the similarity may be calculated for all N combinations, but in order to simplify the process, one of the N rows of the occurrence matrix is selected as the reference search result (reference item), and the reference The similarity between the search result and other search results (other rows) is calculated. The reference search result may be selected randomly using a random number, but in the present embodiment, the search result on the first line (item information output first by the item database 3) is used as the reference search result.

  In the present embodiment, the cosine similarity is used for calculating the similarity as shown in the following equation (1). If the reference search result is the k-th row, the similarity Sik between the reference search result and the i-th search result (i-th row) is obtained by the equation shown in Equation 1. However, i = 1 to N, i ≠ k, and j = 1 to M.

  In the present embodiment, the cosine similarity is used, but the similarity calculation formula is not limited to this. For example, the similarity may be calculated using a known Jaccard coefficient, Simpson coefficient, Pearson product moment correlation coefficient, or the like. Further, instead of extracting words using morphological analysis, the similarity may be calculated by comparing the search results in units of characters. For example, it may be determined whether or not the p-th character from the top matches each of two normalized search results, and the similarity may be calculated using that. Also, a scale generally used as the similarity of character strings, such as Levenshtein distance, may be calculated.

  Then, an average value of similarities obtained for one search result set is calculated and used as a score. For example, when N search results (N ≧ 3) are obtained, (N−1) similarities between the reference search result and the other (N−1) search results are calculated. . What is necessary is just to calculate the average value of these (N-1) similarities. Here, the average value of the similarity is calculated as the score, but the minimum value, the average value, the median value, the mode value, the quartile value, and the like of the similarity may be calculated and used as the score. A larger score means that a plurality of search results are similar. In addition, among a plurality of similarities calculated from one search result set, the number of similarities that are equal to or greater than a predetermined value is counted, and the number is included in the search result set for which the similarity is calculated. The number N or a value divided by the number of similarities calculated from the search result set may be used as the score.

  Common words used in blog posts and words used in song titles often overlap, making it difficult to distinguish between them by creating rules in advance. For this reason, a general word unrelated to the item may be included in the extracted keyword.

  If the keyword is a general word and the item database 3 is searched using the keyword, there is a high possibility that a search result related to a plurality of pieces of music is returned instead of one piece of music. For example, since there are many songs that include a general word such as “love” in the song name, if the item database 3 is searched using “love” as a search key, it is very likely that search results relating to a plurality of songs will be obtained. . In such a case, various search results are obtained, the similarity between the search results is low, and the score is also low.

  On the other hand, when a keyword is a phrase unique to a certain piece of music or a phrase that is not frequently used, even if there are a plurality of search results, it is substantially related to one piece of music. Many. In this case, the similarity between the search results is high, and the score is also high. Therefore, by calculating the score by the method described above, it is possible to accurately determine whether one item has been identified by the keyword (keyword group) used for the search.

  In the next step S9, the validity determination unit 19 of the item specifying unit 12 determines whether the score is greater than a predetermined threshold value θ. The value of θ may be set using search results collected in advance on a trial basis, or the set value may be changed according to the situation. If the score is equal to or greater than θ, the validity determination unit 19 determines that the keyword group is associated with item identification, moves to step S10, returns true, and sets the item associated with the blog article from the search result set. A candidate item that is a candidate is selected, and the item identifier of the candidate item is registered in the “item identifier of candidate item” column of the search result score table shown in FIG. If the score is smaller than θ, it is determined that the keyword group is not linked to item identification, and the process moves to step S11 to return false.

Here, FIG. 8 shows a search result score table indicating the scores of the keyword groups having the keyword group identifiers Gr001-001 to Gr001-010. The search result score table is stored in the score storage unit 7.
Assuming that the threshold θ is “0.4”, in the example of FIG. 8, the three keyword groups Gr001-006, Gr001-008, and Gr001-010 have scores equal to or higher than the threshold θ. A keyword group whose score is equal to or greater than a threshold is associated with one item identifier in the search result set. Note that the threshold θ may be changed according to the number of keywords included in the keyword group. In this case, a larger threshold value (threshold value that is less likely to be true) may be used as the number of keywords increases. As a method for selecting one item (candidate item) from the search result set, the following method can be used.

  The first method is a method of selecting the first (first) item (the item acquired first by the search unit 17) that the item database 3 outputs as a search result. This method can be used when the item database 3 outputs search results with priorities. The text information processing apparatus 1 stores information on the order of the acquired search results.

  The second method is a method of calculating a similarity between a keyword group (search key) and each search result based on the keyword group and selecting a search result (item) having the highest similarity. For example, for the keyword group Gr001-010 including “A song” and “A band”, the keywords “A song” and “A band”, and “A song / A band” and “A song” as search results. Calculate the similarity to “single ver./A band” and “A song / A band withT”. The similarity here may be calculated using a method of a type in which two character strings are compared in units of one character. In this example, since the search result “A song / A band” has the highest similarity, the validity determination unit 19 determines this “A song / A band” as a candidate item, and the items shown in FIG. While referring to the table, A001 which is an item identifier of “A song / A band” is specified and registered as a candidate item identifier corresponding to the keyword group (Gr001-010) of the search result score table. Instead of the similarity, a difference (degree of difference) or a distance between the keyword group and each search result may be calculated.

  The third method is a method of selecting an item having the smallest difference between the item information normalized in step S7 and the item information before normalization. For example, the item database 3 outputs (1) “A song / A band”, (2) “A song single ver./A band”, and (3) “A song / A band withT”. When all of these normalization results are “A song / A band”, the character string does not change before and after normalization. (1) “A song / A band” is selected.

  The fourth method is a method of selecting an item having the highest ranking information created in the past using ranking information described later. This is because the more items that have appeared in blog articles in the past, the more likely they will appear in new blog articles.

  In the next step S12, the validity determination unit 19 determines whether the validity determination has been performed for all the keyword groups. If there is a keyword group that has not been validated yet, the validity determination unit 19 returns to step S9, and the next keyword Compare group scores to thresholds. If it is determined in step S12 that the validity of all keyword groups has been determined, the process proceeds to the next step S13. In step S12, it is not determined whether the validity determination has been performed for all the keyword groups, but when there is one keyword group for which the result of the validity determination is true, the process proceeds to the next step S13. You may move. By doing so, the calculation load is reduced.

  In the next step S13, the validity determination unit 19 of the item specifying unit 12 sets the item identifier and blog identifier for which the validity determination result is true based on the search result score table as shown in FIG. It registers in the item calculation result table with which the calculation result memory | storage part 8 is provided. As described above, the validity determination unit 19 specifies the item identifier for which the result of the validity determination is true as item information corresponding to the blog identifier.

In the example shown in FIG. 8, there are a plurality (three) of keyword groups having a score equal to or higher than the threshold value, and item identifiers (candidate item identifiers) are associated with each of them. An item identifier of a high keyword group may be adopted and registered in the item calculation result table, or item identifiers of all keyword groups having a score equal to or higher than a threshold value may be registered. This is because a plurality of items may be described in one text data. However, when emphasizing the accuracy of specifying an item, it is better to adopt only the item identifier of the keyword group with the highest score. A plurality of keyword groups may be selected in descending order of score, and a plurality of item identifiers corresponding to them may be registered in the item calculation result table. In addition, even when the score calculated in step S8 is less than the threshold, the candidate item is registered in the search result score table, and the item identifier of the candidate item corresponding to the keyword group with the highest score is obtained as the item calculation result. You may register in the table.

When the item identifier of the keyword group with the highest score is adopted, the blog identifier BlogID_001 and the item identifier A001 are output to the item calculation result table in the example shown in FIGS.
As described above, an item identifier that is a description target can be associated with a blog identifier with high accuracy. In the above description, one candidate item is selected from one search result set and registered in the search result score table. However, a plurality of candidate items are selected and registered from one search set. You may do it.

You may make it provide the display control part 21 which displays on the display part the information which shows the item specified as mentioned above with the information (for example, blog identifier or blog title) which shows the corresponding blog article itself or a blog article. . For example, by displaying a blog article together with an item name, it is possible to immediately identify word-of-mouth information related to the item. The display unit may be a display unit (not shown) included in the text information processing apparatus 1 or a display unit (not shown) included in the terminal device 4.
Also, displaying the item name and a plurality of blog articles associated with the item on the same screen is useful because a plurality of word-of-mouth information related to the item can be viewed at a time.

(Operation of the ranking information creation unit 13)
Returning to FIG. 3, processing performed by the ranking information creation unit 13 will be described.
In step S14, the ranking information creation unit 13 is registered in the item calculation result table with reference to the item calculation result table shown in FIG. 9, the text data table shown in FIG. 5, and the item table shown in FIG. The item information (title, artist, etc.), user identifier, and article creation update date corresponding to the combination (blog identifier, item identifier) are extracted.

  In step S15, the ranking information creation unit 13 counts the number of occurrences of the item identifier in the item calculation result table, and sorts the items (item identifier, the number of appearances) in the descending order (item identifier, appearance count). ) Is created and stored in the item ranking information storage unit 9. When a blog article about a certain item is written a predetermined number of times or more, a process of making the number of appearances of the item less than the original number of appearances according to a predetermined rule may be added.

  In step S16, the ranking information creation unit 13 uses the data created in step S14 and counts the number of types of user identifiers (number of appearances of different user identifiers) for each item identifier registered in the item calculation result table. To do. That is, it counts how many users' blogs describe an item. Then, a combination list (second list) sorted in descending order of appearance count (descending order) is created and stored in the item ranking information storage unit 9.

  In step S17, the ranking information creation unit 13 creates a ranking table in the format shown in FIG. 10 using the first list created in step S15 and the second list created in step S16. The ranking table is stored in the ranking information storage unit 9. The ranking table is a table in which the rank, the item identifier, and the number of appearances of the item identifier are associated with each other, and can be created by various methods.

  Specifically, according to the first list, items are ranked in descending order of the number of appearances of items. Next, when there are items having the same number of appearances of items, the items are ranked in descending order according to the second list according to the second list. In other words, the items may be sorted in order of increasing number of appearances of item identifiers as the first priority item and the number of types of user identifiers as the second priority item, and the ranks may be given. The number of types of user identifiers may be sorted as the first priority item, and the number of appearances of the item identifier may be sorted as the second priority item, and the ranking may be given.

  The ranking table creation method described above is merely an example, and rankings can be created by various methods. For example, the total score may be calculated based on the number of appearances of list 1 and the number of types of user identifiers of list 2, and the rank may be given in the order of the total score. This total score may be registered in the ranking table. Moreover, you may make it perform a statistical process based on the various numerical value which concerns on the specified item. For example, set multiple counting periods, compare the number of occurrences of items in each counting period, calculate the rate of increase / decrease in the number of appearances, etc. May be given.

  The display control unit 21 may display the ranking created as described above on the display unit. Moreover, you may display the blog article linked | related with the item contained in the ranking with the ranking, and the information of the user who wrote the blog article. In addition, about a display part, the display part (not shown) which the text information processing apparatus 1 has may be sufficient, and the display part (not shown) which the terminal device 4 has may be sufficient.

  As described above, the text information processing apparatus according to the present embodiment can accurately extract items that are products or services from text data such as a blog.

  Moreover, according to the text information processing apparatus of this embodiment, the extracted item information can be statistically processed.

  For example, within a predetermined period (for example, one week, one day, one hour, etc.), a song that is the object of description by a microblog service or the like is extracted, and the number of articles and the number of users for each song are counted, By ranking the songs according to the numbers, it can be used for marketing as statistical data of market trends. In addition, by presenting such information to the user, it can be expected to increase the user's willingness to purchase.

Second Embodiment
Next, another embodiment of the processing in the text information processing apparatus 1 will be described using the flowcharts of FIGS. 11 and 12.
In the first embodiment, either a search by a keyword group including a first number of keywords and a search by a keyword group including a second number of keywords greater than the first number are performed, or either However, in this embodiment, the number of keywords included in the keyword group is increased according to whether or not the item has been identified. The information can be specified with high accuracy.

  Note that the processing is basically the same as that of the first embodiment except for step S5a, step S12a, step S12b in the flowchart of FIG. 11 and step S5b, step S12c, and step S12d in the flowchart of FIG. Therefore, description of the same processing as in the first embodiment is omitted.

  In the present embodiment, in step S5a of FIG. 11, the grouping processing unit 16 creates a keyword group including the first number of keywords for each article text. For example, like the keyword group identifiers Gr001-001 to Gr001-004 in FIG. 7, the grouping processing unit 16 creates a keyword group including one keyword.

About steps S6-S11, the validity of a search result is judged by the method similar to 1st Embodiment.
In the next step S12a, the validity determination unit 19 determines whether the validity determination has been performed for all the keyword groups including the first number of keywords, and if there is a keyword group that has not yet been validated, the validity determination unit 19 performs step S9. Returning to, the score of the next keyword group is compared with the threshold value. If it is determined in step S12a that the validity of all keyword groups including the first number of keywords has been determined, the process proceeds to the next step S12b.

  In the next step S12b, it is determined whether or not there is a keyword group that has become true. If there is a keyword group that has become true, the process proceeds to step S13, and a valid keyword group and item are output. On the other hand, if there is no true keyword group in step S12b, the process proceeds to step S5b shown in the flowchart of FIG.

  In step S5b of FIG. 12, the grouping processing unit 16 creates a keyword group including a second number of keywords larger than the first number for each article text. For example, like the keyword group identifiers Gr001-005 to Gr001-010 in FIG. 7, the grouping processing unit 16 creates a keyword group including two keywords. Since the process for creating a keyword group has a smaller system load than the search process, a keyword group including the second number of keywords may be created in advance.

Then, about steps S6-S11, the validity of a search result is judged by the method similar to 1st Embodiment.
In the next step S12c, the validity determination unit 19 determines whether the validity determination has been performed for all the keyword groups including the second number of keywords. If there is a keyword group that has not yet been validated, the validity determination unit 19 performs step S9. Returning to, the score of the next keyword group is compared with the threshold value. If it is determined in step S12c that the validity of all keyword groups including the second number of keywords has been determined, the process proceeds to next step S12d.

In the next step S12d, it is determined whether or not there is a keyword group that has become true. If there is a keyword group that has become true, the process proceeds to step S13 in FIG. 11 to output a valid keyword group and item. On the other hand, if there is no true keyword group in step S12b, the process proceeds to step S18, and the validity determination unit 19 determines that the article text does not describe an item.
Note that the grouping processing unit 16 creates a keyword group including a third number of keywords larger than the second number for each article text without terminating the processing in step S18, and performs similar processing. You may continue. The number of keyword groups including a certain number of keywords may be appropriately determined according to, for example, the type of item to be specified.

  As described above, depending on whether or not an item has been identified, the number of keywords included in the keyword group is increased so that the information to be described can be accurately managed while reducing the amount of processing. Can be identified well

  The above-described embodiments of the present invention are illustrative examples, and are not limited to the above-described embodiments. The present invention can also be applied to texts other than blogs, for example, data such as questionnaires. Moreover, although the example which processes using the blog article about music was shown, it cannot be overemphasized that it can process similarly about the article of other fields other than music.

  The present invention includes a program for causing a computer to realize the functions of the respective units. These programs may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

  The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention. For example, you may combine each embodiment, a modification, etc. Further, the functions of the text information processing apparatus 1 may be realized by making a part of the configuration of the text information processing apparatus 1 separate and communicating with the separate configuration via a network or the like.

1 Text information processing device (server)
2 Text data server 3 Item database 4 Terminal device 5 Keyword group storage unit 6 Text data storage unit
7 score storage unit 8 item calculation result storage unit 9 item ranking information storage unit 10 text data collection unit 11 keyword set generation unit 12 item specification unit 13 ranking information creation unit 14 unnecessary character string processing unit 15 keyword extraction unit 16 grouping processing unit 17 Search unit 18 Similarity calculation unit 19 Validity determination unit 20 Network

Claims (13)

A search unit that searches the database and obtains a plurality of information corresponding to the search condition;
A similarity calculation unit that calculates a score based on the similarity between the plurality of pieces of information;
A validity determination unit that determines the validity of the search condition based on the score;
An information processing apparatus comprising: The validity determination unit determines that the search condition is valid when the score is equal to or greater than a predetermined value.
The information processing apparatus according to claim 1. The search condition includes one or more keywords, and the validity determination unit changes a criterion for determining validity according to the number of keywords.
The information processing apparatus according to claim 1, wherein the information processing apparatus is an information processing apparatus. When the validity determination unit determines that the search condition is not valid, the search unit searches the database using a search condition including a number of keywords different from the keyword included in the search condition.
The information processing apparatus according to claim 1, wherein the information processing apparatus is an information processing apparatus. A search unit which searches the database by using a plurality of search conditions, acquires a plurality of search result sets corresponding to each Do Ri said plurality of search conditions from a plurality of information,
Said based on the similarity between a plurality of information included in the search result set, the similarity calculation unit for calculating a plurality of scores corresponding to a plurality of said search result set,
A validity determination unit that preferentially outputs information included in the search result set having a high score;
An information processing apparatus comprising: The validity determination unit identifies a score having the highest value from a plurality of scores calculated by the similarity calculation unit, and outputs at least one information included in the search result set corresponding to the identified score To
The information processing apparatus according to claim 5. The validity determination unit outputs at least one information included in the search result set when the score is equal to or greater than a predetermined value, and is included in the search result set when the score is less than a predetermined value. Output no information,
The information processing apparatus according to claim 5. The search condition includes one or more keywords, and the validity determination unit changes the predetermined value according to the number of keywords included in the search condition.
The information processing apparatus according to claim 7. The similarity calculation unit calculates a plurality of similarities corresponding to a combination of two pieces of information included in the search result set, an average value, a median value, a mode value, a quartile value of the plurality of similarities , Or, the minimum value is the score,
The information processing apparatus according to claim 5 , wherein the information processing apparatus is an information processing apparatus. An information processing method executed by one or more computers,
A search step of searching the database and obtaining a plurality of information corresponding to the search condition;
A similarity calculation step of calculating a score based on the similarity between the plurality of pieces of information acquired in the search step;
A validity determination step of determining the validity of the search condition based on the score calculated in the similarity calculation step;
An information processing method comprising: An information processing method executed by one or more computers,
A search step of searching the database using a plurality of search conditions, acquires a plurality of search result sets corresponding to each Do Ri said plurality of search conditions from a plurality of information,
A similarity calculation step of calculating a plurality of scores are based on the similarity between a plurality of information, corresponding to a plurality of said search result set that is included in the acquired set of search results in the search step,
A validity determination step for preferentially outputting information included in the search result set having a high score calculated in the similarity calculation step;
An information processing method comprising: One or more computers,
A search unit that searches a database and acquires a plurality of information corresponding to the search condition,
A similarity calculation unit that calculates a score based on the similarity between the plurality of pieces of information acquired in the search unit;
A validity determination unit that determines the validity of the search condition based on the score calculated by the similarity calculation unit;
An information processing program that functions as a computer program. One or more computers,
Search unit searches the database by using a plurality of search conditions, acquires a plurality of search result sets corresponding to each Do Ri said plurality of search conditions from a plurality of information,
Said based on the similarity between a plurality of information contained in the acquired set of search results in the search unit, the similarity calculation unit for calculating a plurality of scores corresponding to a plurality of said search result set,
A validity determination unit that preferentially outputs information included in the search result set having a high score calculated by the similarity calculation unit;
An information processing program that functions as a computer program.

Images