JP5837447B2 - Metadata candidate generation device and metadata candidate generation method - Google Patents

Description

  The present invention relates to a metadata candidate generation device and a metadata candidate generation method for generating metadata candidates to be added to stored data stored by a user.

  Conventionally, a technique for automatically generating metadata for image data recorded by a user using a digital camera or the like is known. As a result, it is possible to search for image data using a keyword included in the metadata.

  For example, in Patent Document 1 below, a metadata generation device that receives imaging information including an imaging date and time and an imaging position of image data, and generates an initial keyword used when generating metadata for the image data based on the imaging information. Is disclosed. According to such a function, the initial keyword is extracted from the schedule data corresponding to the imaging date and time, and the initial keyword is extracted from the position information corresponding to the imaging position.

JP 2007-52581 A

  However, in the conventional metadata generation apparatus described above, an initial keyword corresponding to imaging information is extracted, and an intermediate keyword and a final keyword are extracted from document data in which the initial keyword appears frequently. In some cases, a keyword suitable for is not selected. As a result, there is a tendency that retrieval of imaging data cannot be performed efficiently.

  Therefore, the present invention has been made in view of such problems, and is a metadata candidate that can efficiently extract metadata suitable for retrieval of stored data from text information created by a user. It is an object to provide a generation device and a metadata candidate generation method.

  In order to solve the above-described problem, the metadata candidate generation device of the present invention includes a first assignment information acquisition unit that acquires generation time information related to a data generation time assigned to storage data stored by a user, and a user Time information indicating a time within a certain time range with respect to the time indicated by the generation time information, and second addition information acquisition means for acquiring time information relating to the time given to the plurality of text information created by Is specified from among the plurality of text information, and the importance is given from the temporal similarity information acquisition means for acquiring the text information and the text information acquired by the temporal similarity information acquisition means. A plurality of important words extracted and output as candidate metadata for adding the plurality of important words to the stored data.

  Alternatively, in the metadata candidate generation method of the present invention, the first assignment information acquisition unit acquires the generation time information related to the data generation time assigned to the storage data stored by the user. A second grant information acquisition step in which the second grant information acquisition means acquires time information relating to the time given to the plurality of text information created by the user; and a temporal similarity information acquisition means. The time-similar information acquisition for identifying text information to which time information indicating a time within a certain time range is given from the time indicated by the generation time information from among a plurality of text information and acquiring the text information A step and an important word extracting means extract a plurality of important words given importance from the text information acquired by the temporal similarity information acquiring means; Comprising the important word extracting step of outputting as the candidate of the metadata to be added to the number of key words in the stored data, the.

  According to such a metadata candidate generation device or metadata candidate generation method, generation time information given to storage data is acquired, and time information given to a plurality of text information is acquired. Is done. Further, text information to which time information in a certain time range is given with respect to the time indicated by the generation time information is acquired, and a plurality of important words assigned importance from these text information are output as metadata candidates. The As a result, text information related to the time close to the generation time of the stored data is widely acquired, and important words are output from the text information. Therefore, candidates as search keywords for stored data are efficiently extracted. Then, the candidate can be added as metadata to the stored data.

  The first assignment information acquisition means further acquires the generation position information related to the data generation position assigned to the stored data, and the second assignment information acquisition means further acquires the position information related to the position assigned to the text information. Spatial similarity that acquires and identifies text information to which position information indicating a position spatially related to the position indicated by the generated position information is assigned from among a plurality of text information, and acquires the text information The information acquisition means includes an important word extraction means for extracting a plurality of important words with importance from the text information acquired by the temporal similarity information acquisition means and the spatial similarity information acquisition means, and a plurality of important words Are preferably output as metadata candidates to be added to the stored data.

  With such a configuration, the generation position information given to the stored data is acquired, and the position information given to the plurality of text information is acquired. Furthermore, text information to which position information that is spatially related to the position indicated by the generated position information is obtained is acquired, and a plurality of important words that are assigned importance from these text information are output as metadata candidates. Is done. As a result, text information related to a position close to the generation position of the stored data is widely acquired, and important words are output from the text information, so that candidates as search keywords for the stored data can be made even more efficient. It can be extracted well and added as metadata to the stored data from the candidates.

  Moreover, it is preferable that an important word extraction means determines the importance of several important words based on the appearance frequency of several words in text information. In this way, an important word that is more suitable for a search keyword of stored data can be added to the stored data.

  Further, the important word extracting means determines the importance based on the appearance frequency of the plurality of words and the temporal coincidence with respect to the generation time information of the time information given to the text information from which the plurality of words are extracted. It is also preferable. By adopting such a configuration, since the importance level is determined by adding the temporal frequency of the stored word to the appearance frequency of the important word, more important words suitable for the search keyword are selected from a plurality of text information. It can be extracted efficiently.

  In addition, the important word extracting means includes a frequency of appearance of a plurality of words, a degree of temporal coincidence with generation time information of time information given to text information from which a plurality of words are extracted, and a position given to the text information It is also preferable to determine the importance based on the spatial coincidence with the information generation position information. By adopting such a configuration, the importance is determined by adding the temporal frequency and the spatial coincidence with the stored data to the appearance frequency of the important word, so that the importance suitable for the search keyword from a plurality of text information. Words can be extracted more efficiently.

  Furthermore, it is also preferable that the important word extracting unit calculates the importance by weighting and adding the appearance frequency and the temporal coincidence with respect to a plurality of words. In this case, since the importance of the important word is determined while weighting between the appearance frequency in the text information and the temporal coincidence with the stored data, it is possible to select an important word more suitable for the search keyword.

  Furthermore, it is preferable that the important word extracting unit calculates the importance by weighting and adding the appearance frequency, the temporal coincidence, and the spatial coincidence for a plurality of words. In this case, since the importance of the important word is determined while weighting between the appearance frequency in the text information and the temporal and spatial coincidence with the stored data, it is possible to select a more suitable important word for the search keyword. it can.

  The storage device further includes storage data specifying means for specifying a plurality of storage data having generation time information within a certain time range, and the keyword extraction means is provided for one of the plurality of storage data. It is also preferable to output the plurality of extracted important words as metadata candidates to be added to the plurality of stored data. In this way, it is possible to output important words collectively with respect to stored data that are highly related in time, and the data processing efficiency is greatly improved.

  Furthermore, it further comprises storage data specifying means for specifying a plurality of storage data having generation position information spatially related to each other, and the important word extraction means extracts one of the plurality of storage data. It is also preferable to output the plurality of important words as metadata candidates to be added to the plurality of stored data. Also in this case, important words can be output collectively for stored data that are spatially related, and the data processing efficiency is greatly improved.

  According to the present invention, metadata suitable for retrieval of stored data can be efficiently extracted from text information created by a user.

It is a schematic block diagram of the server apparatus 1 for metadata candidate generation concerning suitable one Embodiment of this invention. It is a block diagram which shows the hardware constitutions of the computer which comprises the server apparatus 1 for metadata candidate production | generation of FIG. It is a figure which shows an example of a data structure of the text data stored in the text data storage part 12 of FIG. It is a flowchart which shows the operation | movement at the time of the metadata candidate production | generation by the server apparatus 1 for metadata candidate production | generation of FIG. It is a figure which shows the example of an output of the metadata candidate by the server apparatus 1 for metadata candidate generation of FIG. It is a schematic block diagram of the mobile communication terminal 101 concerning the modification of this invention.

  Hereinafter, preferred embodiments of a metadata candidate generation apparatus and a metadata candidate generation method according to the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

  FIG. 1 is a schematic configuration diagram of a metadata candidate generation server device 1 according to a preferred embodiment of the present invention. A metadata candidate generation server apparatus 1 shown in FIG. 1 is a server apparatus that stores and manages storage data such as photo data, music data, and document data stored by a user using a mobile communication terminal 2. The metadata candidate generation server device 1 may be configured by a single server device or a server system in which a plurality of server devices operate in cooperation. The metadata candidate generation server device 1 performs data communication with the mobile communication terminal 2 via a communication network NW configured by a mobile communication network, a wired communication network, or the like adopting a mobile communication method. Is possible. The mobile communication terminal 2 is a terminal device represented by a mobile phone terminal, a smartphone, a PDA, or the like.

  FIG. 2 is a block diagram showing a hardware configuration of a computer constituting the metadata candidate generation server device 1 of FIG. The computer 100 constituting the metadata candidate generation server device 1 physically includes a CPU 31, a RAM 32 and a ROM 33 that are main storage devices, an auxiliary storage device 36 such as a hard disk device, an input key that is an input device, Information including an input device 35 such as a touch panel and a mouse, an output device 37 such as a display and a speaker, and a communication module 34 that manages transmission / reception of data to / from other terminal devices and server devices via the communication network NW. It is configured as a processing device. The functions realized by the metadata candidate generation server device 1 include a communication module 34 and an input device under the control of the CPU 31 by reading a predetermined program on hardware such as the CPU 31 and the RAM 32 shown in FIG. 35, and by operating the output device 37 and reading and writing data in the RAM 32 and the auxiliary storage device 36.

  Returning to FIG. 1, the metadata candidate generation server device 1 includes a storage data storage unit 11, a text data storage unit 12, and a generation time information acquisition unit (first attached information acquisition unit) as functional components. ) 13, generated position information acquisition unit (first assignment information acquisition unit) 14, time information acquisition unit (second addition information acquisition unit) 15, and position information acquisition unit (second addition information acquisition unit) 17, a temporal similarity information acquisition unit (temporal similarity information acquisition unit) 16, a spatial similarity information acquisition unit (spatial similarity information acquisition unit) 18, a morpheme analysis unit (important word extraction unit) 19, A word extracting unit (important word extracting unit) 20, an important word storage unit 21, a similar data specifying unit (stored data specifying unit) 22, and a candidate data output unit (important word extracting unit) 23 are provided.

  First, the function of each component of the metadata candidate generation server device 1 will be described in detail.

  The storage data storage unit 11 stores a plurality of storage data generated or acquired by the user using the mobile communication terminal 2. Each of the plurality of stored data is provided with generation time information related to the time when the data is stored by the user, and generation position information indicating the position of the mobile communication terminal 2 at the time of data storage. These generation time information and generation position information may be included as metadata in the storage data like EXIF (Exchangeable Image File Format) in the photo data, or may be included in the file name of the storage data. . Further, the generation time information may indicate the time when the user generated the storage data using the mobile communication terminal 2, and indicates the time when the storage data was stored in the metadata candidate generation server device 1. It may be a thing. Similarly, the generation position information may indicate a position when the user generates the storage data using the mobile communication terminal 2, and the storage data is stored in the metadata candidate generation server device 1. It may indicate the position of time.

  The text data storage unit 12 stores a plurality of text data (text information) created by the user using the mobile communication terminal 2. Text data includes schedule data indicating a user's action schedule and action history, diary data posted on a computer that provides various services such as SNS (Social Networking Service) connected to the communication network NW, and memo data. Text data such as, mail data sent by the user to the communication network NW, and the like. This text data is provided with time information related to the time related to the data and position information indicating a position related to the data. For example, the schedule data includes the date and time corresponding to the schedule included in the data, the event data includes the date and time corresponding to events such as Christmas and birthday, and the diary data includes Corresponding date / time and location are included, the mail data includes date / time and location at the time of mail transmission, and the news data includes date / time and location corresponding to the news. These time information and position information may be directly included in the text data, or may be added to the text data as metadata.

  The generation time information acquisition unit 13 reads and acquires the generation time information from the storage data of the metadata candidate output process target stored in the storage data storage unit 11. Further, the generation position information acquisition unit 14 reads out and acquires generation position information from the processing target storage data stored in the storage data storage unit 11. The generation time information acquisition unit 13 passes the acquired generation time information to the temporal similarity information acquisition unit 16, and the generation position information acquisition unit 14 passes the acquired generation time information to the spatial similarity information acquisition unit 18.

  The time information acquisition unit 15 reads and acquires time information from a plurality of text data stored in the text data storage unit 12, and the position information acquisition unit 17 stores a plurality of text data stored in the text data storage unit 12. The position information is read out and acquired. The time information acquisition unit 15 passes the acquired time information to the temporal similarity information acquisition unit 16, and the position information acquisition unit 17 passes the acquired generation time information to the spatial similarity information acquisition unit 18.

  FIG. 3 shows an example of the configuration of data stored in the text data storage unit 12. As shown in the figure, in one text data, ID information “Calendar_001” for identifying the data, data content “Family trip @ Hokkaido”, time information “Date / time: 2012/1/30”, Location information “Hokkaido Sapporo City…” is included, and the other text data includes ID information “Diary_001” for identifying the data, data content “Family go to Hokkaido on summer vacation…”, and time information “ Date and time: 2012/1/30 ”is included. The time information acquisition unit 15 uses a combination of ID and time information from the plurality of text data, more specifically, a combination of ID information “Calendar_001” and time information “Date / Time: 2012/1/30”, Also, a combination of ID information “diary_001” and time information “date / time: 2012/1/30” is acquired. Further, the position information acquisition unit 17 acquires a combination of ID and position information, more specifically, a combination of ID information “Calendar_001” and position information “Sapporo City in Hokkaido ...” from such text data. To do. Note that the position information included in the text data may be information indicating latitude and longitude.

  The temporal similarity information acquisition unit 16 converts the text data to which the time information indicating the time within a certain time range is added to the time indicated by the generation time information given to the storage data to be processed, as text data The text data stored in the storage unit 12 is specified. That is, a combination including time information within a predetermined time is extracted from the time indicated by the generation time information from the combination of ID information and time information acquired by the time information acquisition unit 15, and the ID included in the extracted combination Identify text data based on information. For example, in the case of the text data example shown in FIG. 3, the time range within one day from the time “2012/1/30 01:00:00” indicated by the generation time information given to the storage data to be processed “ 2012/1/29 01:00:00 ”to“ 2012/1/31 01:00:00 ”is calculated, and the time information“ 2012/1/30 ”within the time range is extracted, and the time ID information “Calendar_001” and “Diary_001” corresponding to the information is specified, and the temporal similarity information acquisition unit 16 acquires text data corresponding to the specified ID information from the text data storage unit 12, and Delivered to the morphological analyzer 19.

  The spatial similarity information acquisition unit 18 converts the text data to which the position information indicating the spatially related position is added to the time indicated by the generation position information given to the storage data to be processed. The text data stored in the storage unit 12 is specified. That is, a combination including position information within a predetermined distance range is extracted from the position indicated by the generated position information from the combination of the ID information and the position information acquired by the position information acquisition unit 17 and is included in the extracted combination. Text data is specified based on the ID information. For example, in the example of the text data shown in FIG. 3, the position information indicating the position within the distance range of Zkm from the position “latitude X degrees, longitude Y degrees” indicated by the generation position information given to the storage data to be processed. “Hokkaido Sapporo City…” is extracted as highly relevant location information, and ID information “Calendar_001” corresponding to the location information is specified. At this time, the spatial similarity information acquisition unit 18 may convert the position information indicated by the place name into latitude and longitude, or may convert the position information indicated by the latitude and longitude into the place name for comparison. . Also, the spatial similarity information acquisition unit 18 may extract the position information by determining the distance between the positions indicated by the two position information, and whether the part of the place name indicated by the two position information matches. Position information may be extracted depending on whether or not. Then, the spatial similarity information acquisition unit 18 acquires text data corresponding to the identified ID information from the text data storage unit 12 and delivers it to the morpheme analysis unit 19.

  The morpheme analysis unit 19 performs morpheme analysis on the text data passed from the temporal similarity information acquisition unit 16 and the spatial similarity information acquisition unit 18, divides the text data into words, and all the divided words Is transferred to the keyword extraction unit 20. It is preferable that the morphological analysis unit 19 discriminates the part of speech of the divided word, extracts only words determined as specific parts of speech such as nouns and verbs, and delivers them to the keyword extraction unit 20. It is also preferable to discriminate the part of speech of the divided word and extract the words determined as specific parts of speech such as particles and auxiliary verbs. For example, the morphological analysis unit 19 extracts the words “family trip” and “Hokkaido” from the text data “family trip @ Hokkaido”.

上記式中、n_i,jは、テキストデータｊにおける出現回数、|D|は総テキストデータ数、

は、単語ｉを含むテキストデータ数を示す。
また、重要語抽出部２０は、該当単語の抽出元のテキストデータに付与された時刻情報の示す時刻と、処理対象の記憶データに付与された生成時刻情報の示す時刻との間の時間差を、各単語の記憶データに対する時間的近接度を示す重要度パラメータＳ_ＴＩＭＥとして決定する。また、重要語抽出部２０は、該当単語の抽出元のテキストデータに付与された位置情報の示す位置と、処理対象の記憶データに付与された生成位置情報の示す位置との間の距離を、各単語の記憶データに対する空間的近接度を示す重要度パラメータＳ_{ＳＰＡＣＥ}として決定する。さらに、重要語抽出部２０は、計算した３つの重要度パラメータＳ_{ＴＦ−ＩＤＦ}，Ｓ_ＴＩＭＥ，Ｓ_{ＳＰＡＣＥ}を、下記式；
重要度パラメータＳ＝αＳ^’ _{ＴＦ−ＩＤＦ}＋βＳ^’ _ＴＩＭＥ＋γＳ^’ _{ＳＰＡＣＥ}
を用いて重み付け加算をすることにより、各単語の記憶データに対する重要度を示す重要度パラメータＳを計算して各単語に付加する。ここで、α＋β＋γ＝１であり、Ｓ^’ _{ＴＦ−ＩＤＦ}は、全ての単語の値のうちの上位Ｎ値の総和でＳ_{ＴＦ−ＩＤＦ}を正規化した値である。また、Ｓ^’ _ＴＩＭＥ，Ｓ^’ _{ＳＰＡＣＥ}は、それぞれ、下記式；
Ｓ^’ _ＴＩＭＥ＝（ΣＳ_ＴＩＭＥ−Ｓ_ＴＩＭＥ）／ΣＳ_ＴＩＭＥ，
Ｓ^’ _{ＳＰＡＣＥ}＝（ΣＳ_{ＳＰＡＣＥ}−Ｓ_{ＳＰＡＣＥ}）／ΣＳ_{ＳＰＡＣＥ}
によってＳ_ＴＩＭＥ，Ｓ_{ＳＰＡＣＥ}を正規化した値である（ΣＳ_ＴＩＭＥ，ΣＳ_ＴＩＭＥは、全ての単語の値の総和を示し、ｎ個のＳ_ＴＩＭＥの総和ΣＳ_ＴＩＭＥ＝０の場合には、Ｓ^’ _ＴＩＭＥ＝１／ｎとし、ｎ個のＳ_{ＳＰＡＣＥ}の総和ΣＳ_{ＳＰＡＣＥ}＝０の場合には、Ｓ^’ _{ＳＰＡＣＥ}＝１／ｎとする）。そして、重要語抽出部２０は、重要度パラメータＳを付加した各単語を、メタデータ候補の出力処理対象の記憶データ毎に重要語格納部２１に格納する。 The important word extraction unit 20 determines the importance of each word extracted by the morphological analysis unit 19 based on the appearance frequency of each word, the temporal coincidence with respect to the generation time information, and the spatial coincidence with respect to the generation position information. To do. That is, the important word extraction unit 20 calculates a TF (Term Frequency) value tf from the appearance frequency in the text data from which the corresponding word is extracted, the number of all text data stored in the text data storage unit 12, and An IDF (Inverse Document Frequency) value idf is calculated from the number of text data including the corresponding word, and a TF-IDF value tfidf obtained by multiplying these values is calculated as an importance parameter S _TF-IDF (see the following formula). ).

In the above formula, n _{i, j} is the number of appearances in text data j, | D | is the total number of text data,

Indicates the number of text data including the word i.
In addition, the important word extraction unit 20 calculates a time difference between the time indicated by the time information given to the text data from which the word is extracted and the time indicated by the generation time information given to the storage data to be processed. The importance parameter S _TIME indicating the temporal proximity to the stored data of each word is determined. Further, the important word extraction unit 20 calculates the distance between the position indicated by the position information given to the text data from which the word is extracted and the position indicated by the generated position information given to the storage data to be processed. The importance parameter S _SPACE indicating the spatial proximity of each word to the stored data is determined. Furthermore, the keyword extraction unit 20 calculates the calculated three importance parameters S _TF-IDF , S _TIME , S _SPACE by the following formula:
Importance parameter S = αS ^′ _TF−IDF + βS ^′ _TIME + γS ^′ _SPACE
Is added to each word by calculating the importance parameter S indicating the importance of each word with respect to the stored data. Here, α + β + γ = 1, and S ^′ _TF-IDF is a value obtained by normalizing S _TF-IDF with the sum of the upper N values of all word values. Also, S ^{_'TIME, S'} _SPACE, respectively, the following formulas;
S ^′ _TIME = (ΣS _TIME −S _TIME ) / ΣS _TIME ,
S ^′ _SPACE = (ΣS _SPACE −S _SPACE ) / ΣS _SPACE
By a _{S TIME,} the value obtained by normalizing the _{S SPACE} (ΣS _TIME, ΣS _TIME represents the sum of all the words of the value, if the sum [sigma] s _TIME = 0 of n _{S TIME} is S _^'TIME = a 1 / n, in the case of the sum [sigma] s _SPACE = 0 of n _{S SPACE} is a _{^{S 'SPACE = 1 / n)}} . Then, the important word extraction unit 20 stores each word to which the importance level parameter S is added in the important word storage unit 21 for each storage data of the metadata candidate output processing target.

  The similar data specifying unit 22 specifies storage data similar to the storage data that is the output processing target of the metadata candidate by referring to the storage data storage unit 11. Specifically, a plurality of pieces of storage data to which generation time information indicating a time within a certain time range from a time indicated by generation time information added to the storage data to be processed is classified. For example, when the generation time information “2012/1/30 01:00:00” is given to the storage data to be processed, the range of 1 hour before and after that time “2012/1/30 00:00 : 00 ”to“ 2012/1/30 02:00:00 ”The storage data to which the generation time information is assigned is specified. Further, the similar data specifying unit 22 specifies and classifies a plurality of storage data to which generation position information indicating a spatially related position is assigned from the position indicated by the generation position information assigned to the processing target storage data. To do. For example, when the generation position information “latitude X degrees, longitude Y degrees” is given to the storage data to be processed, the storage data to which the generation position information in the range of 10 km is attached to the position, The storage data to which the position where the place name specified from the generated position information is duplicated is specified. The similar data specifying unit 22 classifies the storage data similar to each other by specifying the generation time bias and the generation position bias by using a clustering technique such as self-organization or the K-means method. Also good. Then, the similar data specifying unit 22 outputs information for specifying a plurality of storage data similar to the storage data to be processed to the candidate data output unit 23.

  In response to a transmission request from the mobile communication terminal 2, the candidate data output unit 23 reads a plurality of words from the important word storage unit 21 in order of importance, and uses them as metadata candidates to be added to the storage data to be processed. Output. At this time, the candidate data output unit 23 converts the plurality of words output for one storage data to be processed into a plurality of storage data classified as similar to the storage data by the similar data specifying unit 22. Output as data candidates at the same time. Further, when the candidate data output unit 23 receives an input from the mobile communication terminal 2 for selecting a word to be reflected in the metadata from the metadata candidates, the candidate data output unit 23 stores the stored data so that the word is reflected in the metadata. The corresponding stored data in the unit 11 is updated. Here, when the candidate data output unit 23 outputs metadata candidates to the storage data to be processed, the candidate data output unit 23 outputs the metadata candidates for each type of text data (for example, schedule data, diary data). , And for each mail data), metadata candidates may be ranked and output, or candidates extracted from all types of text data and ranked may be output collectively.

  Hereinafter, the operation of the metadata candidate generation server device 1 will be described with reference to FIG. 4, and the metadata candidate generation method in the metadata candidate generation server device 1 will be described in detail. FIG. 3 is a flowchart showing an operation at the time of metadata candidate generation by the metadata candidate generation server apparatus 1.

  First, an input for selecting a metadata candidate output processing target is received from the user of the mobile communication terminal 2 (step S101). Then, the generation time information given to the storage data to be processed is read from the storage data storage unit 11 by the generation time information acquisition unit 13, and at the same time, a plurality of texts are read from the text data storage unit 12 by the time information acquisition unit 15. The time information attached to the data is read (step S102). The temporal similarity information acquisition unit 16 identifies a plurality of text data to which time information within a certain time range is given with respect to the time indicated by the generation time information, and the text data is stored in the text data storage unit 12. (Step S103). Furthermore, the generated position information acquisition unit 14 reads the generated position information given to the storage data to be processed from the stored data storage unit 11, and at the same time, the position information acquisition unit 17 reads a plurality of texts from the text data storage unit 12. The position information given to the data is read (step S104). The spatial similarity information acquisition unit 18 identifies a plurality of text data to which position information highly related to the position indicated by the generation position information is assigned, and the text data is stored in the text data storage unit 12. (Step S105).

  Thereafter, the morphological analysis unit 19 divides the plurality of text data into words (step S106). Next, the important word extraction unit 20 calculates the importance parameter S for the plurality of divided words, and the plurality of words ranked based on the importance parameter S are processed storage data. Every time it is stored in the keyword storage unit 21 (step S107). Further, the similar data specifying unit 22 classifies a plurality of stored data that are similar to each other spatially or temporally (step S108). Finally, in response to a transmission request from the mobile communication terminal 2, words as metadata candidates are ranked and output for the storage data to be processed and the storage data classified as similar to the storage data. (Step S109).

FIG. 5 shows an example of the output screen D ₁ displayed on the mobile communication terminal 2 in response to the output of metadata candidates from the metadata candidate generation server device 1. As shown in the figure, a photograph data G ₁ is a storage data to be processed, photograph data G ₂ similar temporally or spatially with respect to picture data G _1, G _3, the metadata candidates for the G ₄ The words “Hokkaido”, “Family trip”, “Summer vacation”, “Barbecue”, “Campfire”, “Kuma no figurine” are displayed, and the user selects “Hokkaido”, “ "Family trip", "Summer vacation", "Barbecue", "Campfire" are selected as words to be reflected as metadata.

  According to the metadata candidate generation server device 1 and the metadata candidate generation method in the metadata candidate generation server device 1 described above, the generation time information and the generation position information given to the storage data are acquired. At the same time, time information and position information given to the plurality of text data are acquired. Further, text data to which time information in a certain time range is assigned to the time indicated by the generation time information is acquired, and position information that is spatially related to the position indicated by the generation position information is assigned. Text data is acquired, and a plurality of important words given importance from these text information are output as metadata candidates. As a result, text information related to the time and position close to the generation time and generation position of the stored data is widely acquired, and important words are output from the text information. Candidates can be efficiently extracted and added to the stored data as metadata from the candidates.

  In addition, since the importance of the plurality of words is determined based on the appearance frequency of the plurality of words in the text data, it is possible to add an important word more suitable for the search keyword of the storage data to the storage data. Furthermore, since the importance is determined by adding the temporal and spatial coincidence to the stored data to the occurrence frequency of the important word, the important word suitable for the search keyword is more efficiently selected from a plurality of text information. Can be extracted.

  In addition, metadata candidates can be output collectively for a plurality of stored data having generation time information within a certain time range, and the data processing efficiency is remarkably improved. Furthermore, metadata candidates can be output collectively for a plurality of stored data having generated position information that is spatially related to each other, further improving data processing efficiency.

  In addition, this invention is not limited to embodiment mentioned above.

  For example, some or all of the components shown in FIG. 1 may be included in the terminal device used by the user. For example, as shown in FIG. 6, all the components may be provided in the mobile communication terminal 101 used by a user who wants to add metadata to stored data.

  DESCRIPTION OF SYMBOLS 1 ... Metadata candidate production | generation server apparatus, 101,2 ... Mobile communication terminal, 13 ... Generation | occurrence | production time information acquisition part (1st provision information acquisition means), 14 ... Generation | occurrence | production position information acquisition part (1st provision information acquisition means) ), 15... Time information acquisition unit (second attached information acquisition unit), 16... Temporal similarity information acquisition unit (temporal similarity information acquisition unit), 17... Position information acquisition unit (second addition information acquisition unit) 18 ... Spatial similarity information acquisition unit (spatial similarity information acquisition unit), 19 ... Morphological analysis unit (important word extraction unit), 20 ... Important word extraction unit (important word extraction unit), 22 ... Similar data specifying unit ( Storage data specifying means), 23... Candidate data output section (important word extracting means).

Claims (6)

First grant information acquisition means for acquiring generation time information related to data generation time assigned to stored data stored by a user;
A second assignment information acquisition means for acquiring time information related to the time assigned to the plurality of text information created by the user;
Time-similar information for identifying text information to which time information indicating a time within a certain time range with respect to the time indicated by the generation time information is specified from among the plurality of text information and acquiring the text information Acquisition means;
Important word extraction that extracts a plurality of important words given importance from the text information acquired by the temporal similarity information acquisition means, and outputs the plurality of important words as metadata candidates to be added to the stored data Means,
Equipped with a,
The important word extraction means includes
The appearance frequency and the time based on the appearance frequency of the plurality of words in the text information and the degree of temporal coincidence with the generation time information of the time information given to the text information from which the plurality of words are extracted Calculating and determining the importance of the plurality of important words by weighted addition of the degree of matching
A metadata candidate generation device characterized by that.   First grant information acquisition means for acquiring generation time information related to a data generation time given to storage data stored by a user and generation position information related to a data generation position given to the storage data; ,
  Second grant information acquisition means for acquiring time information related to a time given to a plurality of text information created by a user and position information related to a location given to the text information;
  Time-similar information for identifying text information to which time information indicating a time within a certain time range with respect to the time indicated by the generation time information is specified from among the plurality of text information and acquiring the text information Acquisition means;
  Spatial similarity information for specifying text information to which position information indicating a position spatially related to the position indicated by the generated position information is assigned from among the plurality of text information and acquiring the text information Acquisition means;
  Metadata that extracts a plurality of important words given importance from the text information acquired by the temporal similarity information acquisition means and the spatial similarity information acquisition means, and adds the plurality of important words to the stored data Keyword extraction means to output as a candidate for
With
  The important word extraction means includes
Frequency of appearance of a plurality of words in the text information, degree of temporal coincidence of the time information given to the text information from which the plurality of words are extracted to the generation time information, and the position given to the text information The weight of the appearance frequency, the temporal coincidence, and the spatial coincidence of the plurality of words based on the spatial coincidence with respect to the generation position information of information, thereby calculating the importance of the plurality of important words. Calculate and determine the degree,
A metadata candidate generation device characterized by that. Further comprising storage data specifying means for specifying a plurality of the storage data having generation time information within a certain time range from each other;
The important word extracting means outputs the plurality of important words extracted for one of the plurality of stored data as metadata candidates to be added to the plurality of stored data.
The metadata candidate generation device according to claim 1 or 2 , characterized in that A storage data specifying means for specifying a plurality of the storage data having generated position information spatially related to each other;
The important word extracting means outputs the plurality of important words extracted for one of the plurality of stored data as metadata candidates to be added to the plurality of stored data.
The metadata candidate generation device according to any one of claims 1 to 3 . A first grant information acquisition step in which a first grant information acquisition means acquires generation time information related to a data generation time given to storage data stored by a user;
A second grant information acquisition step in which the second grant information acquisition means acquires time information related to the time given to the plurality of text information created by the user;
The temporal similarity information acquisition unit specifies text information to which time information indicating a time within a certain time range is given with respect to the time indicated by the generation time information from among the plurality of text information, and the text A temporally similar information acquisition step for acquiring information;
A candidate for metadata in which the important word extracting means extracts a plurality of important words given importance from the text information acquired by the temporal similarity information acquiring means, and adds the plurality of important words to the stored data Key word extraction step to output as
Equipped with a,
In the important word extraction step,
The appearance frequency and the time based on the appearance frequency of the plurality of words in the text information and the degree of temporal coincidence with the generation time information of the time information given to the text information from which the plurality of words are extracted Calculating and determining the importance of the plurality of important words by weighted addition of the degree of matching
A metadata candidate generation method characterized by:   A first assignment information acquisition unit acquires generation time information related to a data generation time assigned to storage data stored by a user and generation position information related to a data generation position assigned to the storage data. A first grant information acquisition step,
  2nd grant information in which the 2nd grant information acquisition means acquires time information about time given to a plurality of text information created by a user, and position information about a position given to the text information An acquisition step;
  The temporal similarity information acquisition unit specifies text information to which time information indicating a time within a certain time range is given with respect to the time indicated by the generation time information from among the plurality of text information, and the text A temporally similar information acquisition step for acquiring information;
  Spatial similarity information acquisition means specifies text information to which position information indicating a position spatially related to the position indicated by the generated position information is assigned from the plurality of text information, and the text A spatial similarity information acquisition step for acquiring information;
  An important word extracting means extracts a plurality of important words given importance from the text information acquired by the temporal similar information acquiring means and the spatial similar information acquiring means, and stores the plurality of important words in the memory A key word extraction step to output as metadata candidates to be added to the data;
With
  In the important word extraction step,
Frequency of appearance of a plurality of words in the text information, degree of temporal coincidence of the time information given to the text information from which the plurality of words are extracted to the generation time information, and the position given to the text information The weight of the appearance frequency, the temporal coincidence, and the spatial coincidence of the plurality of words based on the spatial coincidence with respect to the generation position information of information, thereby calculating the importance of the plurality of important words. Calculate and determine the degree,
A metadata candidate generation method characterized by:

Images