JP6411800B2 - Information management apparatus, information management system, and information management program - Google Patents

Description

  The present invention relates to an information management apparatus, an information management system, and an information management method.

  In recent years, in order to support users' going-out behavior in the real world, local information such as sightseeing spots, local events, local products (special products), etc. are provided via information terminals such as smartphones and PCs. Regional information services are provided. The usefulness of the above regional information service is improved by providing a lot of up-to-date local information.

JP 2011-145879 A

  According to the applicant's survey, the number of information related to events (hereinafter referred to as event information) is significantly smaller than the number of information related to sightseeing spots and products. From the viewpoint of the usefulness of the regional information service, it is preferable that more event information is provided. In addition, since events are held with a limited date and time, it is important to maintain the freshness of event information through frequent information collection. However, it is difficult to manually collect a large number of event information manually from a viewpoint of work load and work cost.

  In view of the above circumstances, an object of the present invention is to appropriately collect event information.

  The information management device according to the present invention includes a spot post sorting unit that sorts post information related to a spot from a plurality of pieces of post information based on geographical information included in the post information, and a plurality of pieces of post information. Based on the wording included in the information, the event posting sorting unit that sorts the posted information related to the event, and the sorted posted information sorted by the spot post sorting unit and the event post sorting unit, An event information extraction unit that extracts event information including an event name included in the posted information and a holding spot name where the event corresponding to the event name is held.

  The information management system of the present invention includes a spot post sorting unit that sorts post information related to a spot from a plurality of pieces of post information based on geographical information included in the post information, and a plurality of pieces of post information. Based on the wording included in the information, the event posting sorting unit that sorts the posted information related to the event, and the sorted posted information sorted by the spot post sorting unit and the event post sorting unit, An event information extraction unit that extracts event information including an event name included in the posted information and a holding spot name where the event corresponding to the event name is held.

  The information management method of the present invention includes selecting post information related to a spot from a plurality of post information based on geographical information included in the post information, and including the post information from a plurality of post information. The posting information related to the event is selected based on the wording, and the selected posted information is analyzed, and the event name included in the selected posted information and the event corresponding to the event name are Extracting event information including names of held spots.

  According to the present invention, event information is appropriately collected from a plurality of post information.

It is a figure which shows the information management server which concerns on embodiment, and the computer apparatus relevant to an information management server. It is a physical configuration block diagram of the server of the embodiment. It is a physical configuration block diagram of the client terminal of the embodiment. It is a logical configuration block diagram of the information management server of the embodiment. It is a logic block diagram of the contribution information acquisition server of an embodiment. It is a flowchart which shows roughly the event information collection operation | movement of embodiment. It is a flowchart which shows spot contribution selection operation | movement. It is a flowchart which shows event posting selection operation | movement. It is a flowchart which shows the flow of machine learning (SVM). It is a figure which shows an example of learning data. It is a flowchart which shows event information extraction operation | movement. It is a flowchart which shows the flow of machine learning (CRF). It is a figure which shows an example of learning data. It is a flowchart which shows event acquisition information integration operation | movement.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

1. Embodiment 1 (1). Overview of Information Management Server With reference to FIG. 1, an information management server MS according to an embodiment of the present invention and a plurality of computer devices related to the information management server MS are schematically shown. The computer apparatus according to the present embodiment executes communication according to a predetermined protocol (TCP / IP or the like).

The blog server BS receives and stores post information PI posted from a user terminal (for example, a smartphone), and provides the post information PI in response to an external request. Hereinafter, examples of post information PI in the present embodiment will be enumerated without limitation.
-Short postings (eg, tweets [registered trademark]) of 140 characters or less in a microblog service (such as Twitter [registered trademark])
-Posting diaries and reviews on blog (web log) services-Status posting on social networking services (SNS)

  The posting information acquisition server AS acquires and accumulates the posting information PI provided from the blog server BS. The posting information acquisition server AS may accumulate all the posting information PI posted to the blog server BS, or may accumulate a part of the posting information PI (for example, only Japanese posting information PI). .

  Further, the posting information acquisition server AS supplies the posting information PI to the information management server MS in response to a request from the information management server MS. The posted information acquisition server AS may supply all of the accumulated posted information PI to the information management server MS, or may supply a part of the posted information PI.

  The information management server MS is a computer device that analyzes the posting information PI supplied from the posting information acquisition server AS, collects and accumulates event information EI, and presents the event information EI in response to a request from the client terminal CL. is there. In the present embodiment, Japanese posting information PI is an analysis target. Post information PI limited according to a predetermined standard may be the analysis target. For example, only post information PI posted by a specific account (such as an official account on Twitter) may be analyzed.

  The event information EI collected based on the posted information PI is information related to events such as festivals, exhibitions, and concerts. The event information EI includes at least the name of the event and the name of the spot where the event is held. The event information EI may include the date and time when the event is held, the latitude / longitude of the place where the event is held, a URL that specifies a website related to the event, a feature word related to the event, and the like. For example, certain event information EI includes information “event name: science festival”, “host spot name: discovery park Yaizu”, and “date and time of holding: from 13:00 on March 23, 2014”.

  The client terminal CL requests and acquires the event information EI from the information management server MS, and displays information on the display in response to a request from the user. For example, the client terminal CL displays the event information EI superimposed on the electronic map.

  The electronic map data used for displaying the electronic map may be acquired from the map server (not shown) by the client terminal CL in response to a request from the user, or may be stored in the client terminal CL in advance. Further, it may be provided from the information management server MS together with the event information EI. In that case, the information management server MS may store electronic map data in advance, or may acquire it from the map server each time or periodically.

1 (2). Physical configuration 1 (2) -1. Configuration of Server Device FIG. 2 is a block diagram showing the physical configuration of each server device of this embodiment including the information management server MS. Each server device (AS, BS, MS) includes a network interface 10, an input unit 12, an output unit 14, a CPU (Central Processing Unit) 16, a RAM (Random Access Memory) 18, a ROM (Read Only Memory) 20, and an HDD (HDD). Hard Disk Drive) 22.

  The network interface 10 communicates with other computer devices via a network. The input unit 12 receives an input signal from an input device such as a keyboard. The output unit 14 transmits an output signal to an output device such as a display. The CPU 16 performs various controls and calculations by executing programs stored in the RAM 18 and the ROM 20 which are main storage devices. The HDD 22 is an auxiliary storage device that stores programs and data that can be expanded on the RAM 18. Note that a storage medium such as an SSD may be employed instead of or in addition to the HDD 22.

  As a person skilled in the art understands naturally, one server may be configured by a plurality of computer devices, or one computer device may include a plurality of virtualized servers.

1 (2) -2. Configuration of Client Terminal FIG. 3 is a block diagram showing a physical configuration of the client terminal CL of the present embodiment. The client terminal CL includes a network interface 30, an input unit 32, an output unit 34, a CPU 36, a RAM 38, a ROM 40, and an HDD 42. Each of the above elements has the same configuration as the element described with reference to FIG.

1 (3). Logical Configuration FIG. 4 is a block diagram showing a logical configuration of the information management server MS. The information management server MS includes a posted information selection unit 100, an event information extraction unit 110, an event information integration unit 120, and an information presentation unit 130 as functional blocks. The posted information sorting unit 100 includes a spot posted sorting unit 102 and an event posted sorting unit 104. The information management server MS also includes a spot information table 140, a spot posting table 150, an event posting table 160, and an event information table 170 as logical tables.

  A schematic operation of each functional block will be described. The posted information sorting unit 100 sorts posted information PI supplied from the posted information acquisition server AS. More specifically, the spot posting sorting unit 102 sorts the posting information PI related to the spot name based on the geographical information included in the posting information PI. The event posting sorting unit 104 sorts the posting information PI related to the event based on the wording included in the posting information PI. The event information extraction unit 110 analyzes the selected posted information PI (selected posted information SPI), and the event spot included in the selected posted information SPI and the event spot corresponding to the event name is held. Event information EI including the name is extracted. The event information integration unit 120 integrates a plurality of event information EI corresponding to the same event among the extracted event information EI into one integrated event information IEI. The information presentation unit 130 presents the integrated event information EI (integrated event information IEI) to the client terminal CL.

  FIG. 5 is a block diagram showing a logical configuration of the posting information acquisition server AS. The posting information acquisition server AS includes a posting information processing unit 200 that accumulates the posting information PI acquired from the blog server BS as a functional block, and includes a posting information table 210 as a logical table.

  The above functional blocks are realized by the CPU 16 executing a computer program stored in the main storage device of each server (AS, MS). Further, the above logical table is a constituent element of a relational database in which a plurality of tables are linked based on a predetermined relation (relation). The relational database may be implemented as an on-disk database in which data is stored in the HDD 22 that is a non-volatile memory, or may be implemented as an in-memory database in which data is stored in the RAM 18 that is a volatile memory.

1 (4). Event Information Collection Operation The event information collection operation of this embodiment will be described below. FIG. 6 is a flowchart schematically showing the event information collection operation of the present embodiment. Event information EI is extracted from the posted information PI selected stepwise in steps S1010 and S1020 (S1030). A plurality of pieces of event information EI relating to one event are integrated (name identification) in step S1040.

1 (4) -1. Spot post sorting operation (step S1010)
In step S1010, the spot post selection unit 102 executes a spot post selection operation. Hereinafter, a specific operation will be described with reference to FIG. The spot posting selection unit 102 performs morpheme analysis on each of the plurality of posting information PIs acquired from the posting information acquisition server AS, and acquires the morphemes included in each posting information PI (S1012).

  Next, the spot post selection unit 102 performs matching between the morpheme acquired in step S1012 and the spot name included in the spot information table 140 for each post information PI (S1014). The spot information table 140 includes a plurality of spot information entries. Each spot information entry may include at least the spot name and the latitude / longitude of the spot, and may include additional information such as an explanatory note regarding the spot. In addition, it is preferable that a plurality of spot names corresponding to the same spot (for example, “Kinkakuji” and “Rokuji”) are included in the spot information table 140 because fluctuations in the spot name included in the posted information PI can be absorbed. is there.

  As a result of the matching process in step S1014, the spot post selection unit 102 performs a filtering process on the post information PI determined to include the spot name (S1016). Any filtering process may be applied. For example, a phrase that can be a spot name or a person name (for example, “Matsushima” or “Kawasaki”) corresponds to person name filtering that excludes post information PI that is used as a person name instead of a spot name, or 1 corresponding to a plurality of spots. The same name filtering is performed to determine which spot names (for example, “Maruyama Park” in Kyoto Prefecture and “Maruyama Park” in Hokkaido) indicate. A plurality of filtering processes may be executed in parallel. It is preferable that a flag indicating that is added to a spot information entry corresponding to a spot name that can be used as a person name. The same applies to spot information entries corresponding to overlapping spot names.

  An arbitrary algorithm can be adopted for the above filtering process. For example, the above filtering (personal name filtering, same name filtering) is executed depending on whether or not the post information PI includes a phrase acquired from an explanatory text related to a spot based on Inverse Document Frequency (IDF). May be. Further, the personal name filtering is constructed by machine learning after causing the spot posting selection unit 102 to execute machine learning for determining whether the same word or phrase is used as a person name or a spot name. Whether a person name or a spot name may be determined based on the model. The above machine learning is executed based on, for example, a conditional random field (CRF). Regarding person name filtering, filtering by IDF and filtering by CRF may be executed in parallel.

  Through the above steps S1010 (steps S1012 to S1016), post information PI related to the spot name is selected from the entire post information PI and stored in the spot post table 150. The selected posting information PI may be directly supplied to the event posting selection unit 104.

1 (4) -2. Event posting selection operation (step S1020)
In step S1020, the event posting sorting unit 104 executes an event posting sorting operation. Hereinafter, a specific operation will be described with reference to FIG. The event posting sorting unit 104 extracts a feature from each of the plurality of posting information PI acquired from the spot posting table 150 (or the spot posting sorting unit 102) (S1022).

  The extracted feature is an arbitrary index related to the wording included in the posting information PI. For example, a word vector indicating a word included in the posting information PI is an example of the feature. The above word vectors are typically vectors that also indicate the appearance frequency of words, but word vectors that do not indicate the appearance frequency of words (only indicate the presence or absence of words) may be employed. In addition, an index indicating the unusualness of the word based on the reverse document frequency may be adopted as the feature.

  The event posting sorting unit 104 sorts the posting information PI related to the event based on the extracted feature (S1024). Whether the posting information PI is related to the event is determined based on an arbitrary criterion. For example, it is preferable that the event posting selection unit 104 determines whether or not the posting information PI is related to the event based on the result of machine learning related to the relationship between the posting information PI and the event.

  A specific example of machine learning will be described with reference to FIGS. 9 and 10. FIG. 9 is an operation flow of the learning phase executed prior to the estimation phase of step S1024, and FIG. 10 is an example of learning data used in the learning phase. In this embodiment, a support vector machine (Support Vector Machine, SVM), which is a type of supervised machine learning, is employed. The use of a linear kernel as the SVM kernel function is preferable because the processing load is light, but other kernel functions may be adopted.

  In the learning phase, first, a label indicating whether or not the event is related is given to the posting information PI (S800). The post information PI including the event name (the unique name of the event) is given a label indicating “related”, and the post information PI not including the event name is assigned a label indicating “not related”. FIG. 10 shows the learning data after the label is assigned.

  A feature extraction process similar to that in step S1022 is executed for each post information PI to which a label is assigned (S810). Thereafter, the posting information PI assigned with the “related” label is taken as a positive example, and the posting information PI given the “not related” label is taken as a negative example, and the event posting sorting unit 104 performs machine learning (S820). .

  In step S1024, post information PI estimated to be related to the event is selected based on the model constructed by the above machine learning. The selected posting information PI is stored in the event posting table 160. The selected posting information PI may be directly supplied to the event information extraction unit 110.

  Post information PI that has not been selected (post information PI estimated not to be related to an event) may be discarded or may be accumulated for some purpose (for example, as a negative example of machine learning).

1 (4) -3. Event information extraction operation (step S1030)
In step S1030, the event information extraction unit 110 executes an event information extraction operation. Hereinafter, a specific operation will be described with reference to FIG. The event information extraction unit 110 extracts features from each of the plurality of post information PI acquired from the event post table 160 (or the event post selection unit 104) (S1032).

  The extracted feature is arbitrary information regarding the wording included in each post information PI. In the present embodiment, information including word notation, part of speech, original form, reading, group, number of characters, and character type acquired by morphological analysis of the posted information PI is extracted as a feature.

  The event information extraction unit 110 extracts event information EI based on the extracted features (S1034). As described above, the event information EI includes an event name and a holding spot name. The event name is extracted by an arbitrary method. For example, it is preferable that the event information extraction unit 110 extracts the event name by extracting the unique expression based on the result of machine learning related to the event name included in the posted information PI. It is more preferable that the date and time of the event is also extracted by the above-described extraction of the unique expressions.

  A specific example of machine learning will be described with reference to FIGS. 12 and 13. FIG. 12 is an operation flow of the learning phase executed prior to the estimation phase of step S1034, and FIG. 13 is an example of learning data used in the learning phase. In this embodiment, a conditional random field (CRF), which is a type of supervised machine learning, is employed. However, other machine learning algorithms may be employed.

In the learning phase, first, a label related to the unique expression is given to a word / phrase included in the posting information PI (S900). As shown in FIG. 13, the following five types of labels are given to words (B means Begin, I means Inside, and O means Outside).
-B-Event: Indicates the start of a specific expression corresponding to the event name.
-I-Event: Indicates that it is a part of the unique expression corresponding to the event name.
-B-Time: Indicates the start of a specific expression corresponding to the date and time of the event
-I-Time: Indicates that it is a part of the unique expression corresponding to the date and time of the event.
・ O: Indicates other elements.

  The feature extraction process similar to step S1032 is executed for each piece of posted information PI to which a label is assigned (S910). Thereafter, the event information extraction unit 110 performs machine learning based on the assigned label and the extracted feature (S920).

  In step S1034, an event name to be included in the event information EI is extracted based on the model constructed by the above machine learning. More specifically, the event information extraction unit 110 extracts a series of morphemes provided with one B-Event label and one or more I-Event labels as event names. Preferably, the date and time to be included in the event information EI is also extracted.

  As for the holding spot name to be included in the event information EI, the spot name specified in the above step S1014 may be adopted as the holding spot name as it is, or machine learning (for example, machine learning using the above-mentioned CRF). ) May be used to extract the name of the holding spot. When extracted by machine learning using CRF, labels (B-Spot and I-Spot) about the name of the holding spot are also given to the phrase in the learning phase described above.

  The extracted event information EI is stored in the event information table 170. The extracted event information EI may be directly supplied to the event information integration unit 120.

  The event information collection operation may be terminated in step S1030 (steps S1032 and S1034) without performing the event information integration operation described later. However, at least according to conventional event information collection, a plurality of event names are often extracted for the same event. In the present embodiment, the above problems are solved by the following event information integration operation.

1 (4) -4. Event information integration operation (step S1040)
In step S1040, the event information integration unit 120 executes an event information integration operation (name identification). Hereinafter, a specific operation will be described with reference to FIG. The event information integration unit 120 groups a plurality of event information EI acquired from the event information table 170 (or event information extraction unit 110) for each holding spot name (S1042). In this example, for a certain holding spot name, event information EIa having an event name [future exhibition of the 21st century] (the bracket is not included in the character string of the event name; the same shall apply hereinafter) and [21st century It is assumed that event information EIb having an event name “Mirai Exhibition” is grouped.

When multiple event information EI exists for one holding spot name, the event information integration unit 120 creates all event name pairs (spot identical event pairs) included in these event information EI (S1044). In this example, one spot-identical event pair of [21st Century Future Exhibition]-[21st Century Mirai Exhibition] is created. When the number of event information EI for one holding spot name is _n , it is naturally understood that the number of spot identical event pairs is _n C ₂ .

  When only one event information EI exists for one held spot name, the event information integration operation for the held spot name ends.

  Next, the event information integration unit 120 performs standardization after obtaining the longest common partial sequence length for each spot identical event pair, and calculates the standardized longest common partial sequence length (S1046). The standardized common subsequence length is an example of an index indicating the similarity between two event names corresponding to the same spot event pair. The longest common partial string length means the length of the longest character string among the partial character strings that appear in common in two character strings. In this example ([21st Century Future Exhibition]-[21st Century Mirai Exhibition]), [21st Century Exhibition] is the longest common substring, and its column length is “6”. The standardization in step S1046 is a process for eliminating the influence of the length of the character string, and is a process of dividing the longest common substring length by the longer character string length of the spot identical event pairs. In the standardization of this example, the longest common substring length “6” is divided by the longer [21st century Mirai exhibition] character string length “9”. Therefore, the standardized longest common subsequence length in this example is “0.67” (= 6/9. The third decimal place is rounded off).

  The event information integration unit 120 determines whether or not the event information EI should be integrated based on the standardized longest common subsequence length (S1048). When the standardized longest common partial sequence length is equal to or greater than the threshold Th, the event information integration unit 120 sets the event information table 170 to integrate two event information EI corresponding to the same spot event pair into one integrated event information IEI. Update (S1050). On the other hand, when the standardized longest common partial sequence length is less than the threshold Th, the event information EI is not integrated (S1052).

  The threshold value Th used in step S1048 is arbitrarily set. For example, according to the trial using the test data by the applicant, a suitable example of the threshold value Th is “0.44”. An evaluation method for determining whether the threshold value Th is suitable is also arbitrary.

  In step S1050, the criteria for determining which event name of the spot identical event pair is adopted as the event name of the integrated event information IEI is arbitrary. For example, a shorter event name may be employed, or an event name included in more post information PI may be employed. Further, when the posting information PI is a tweet (registered trademark), an event name included in the posting information PI retweeted more may be adopted. In addition, an event name having a larger number of hits when input to the search engine may be employed, or an event name included in the posting information PI including additional information such as a URL may be employed.

1 (5). Example (experimental result)
One example of the embodiment is described below. In the experiment of this example, the event information collection operation (S1010 to S1040) of this embodiment was executed for Japanese tweets posted in November 2013 (about 1,692 million). SVM machine learning (S800-S820) uses 2000 tweets posted in 2012 (200 positive examples including event names, 1800 negative examples not including event names), and CRF machine learning ( S900-S920) used 254 tweets including event-related information posted in 2012. As a result, 9781 event information EI was collected. The event information EI before the event information integration operation in step S1040 was executed was about 137,800.

  400 items were randomly extracted from the collected event information EI of 9781 cases, and the accuracy of the collected event names was evaluated. Of the 400 events, 276 (69%) event names completely matched the actual event name, 55 (13%) event names that partially matched the actual event name, and falsely detected events The name was 69 (17%).

1 (6). Effects of the Embodiment According to the configuration of the present embodiment described above, the event information EI is extracted from the selected posted information PI after the stepwise selection of the posted information PI is performed. Therefore, the event information EI is appropriately collected. A plurality of event information EI corresponding to the same event is integrated into one event information EI (integrated event information IEI). Therefore, the collected event information EI is arranged more appropriately.

2. Modifications The above embodiment can be variously modified. Specific modifications are exemplified below. Two or more aspects arbitrarily selected from the above embodiments and the following examples can be appropriately combined as long as they do not contradict each other.

2 (1). Modification 1
In the above embodiment, the word analysis by the morphological analysis is executed in the spot posting selection operation (S1010). However, morphological analysis is not essential, and wording analysis by other methods, for example, analysis by N-gram may be executed.

  Further, in the above embodiment, the posted information PI is selected by the lexical matching process. However, the selection may be performed by other selection methods, for example, latitude / longitude information (geotag) included in the posted information PI. For example, post information PI having a latitude and longitude within a predetermined distance from the latitude and longitude of a spot included in the spot information table 140 may be selected.

  The event information EI in this example includes at least the name of the event and the latitude and longitude of the place where the event is held. The “latitude / longitude of the place where the event is held” is arbitrarily acquired. For example, the event information EI includes latitude / longitude information (geotag) of the position where the post information PI estimated to be related to the event is posted.

  As understood from the above, in step S1010, based on the geographical information (spot name, geotag, etc.) included in the post information PI, post information PI related to the spot (post information PI including the spot name, spot) The posting information PI etc. posted in the vicinity of is selected.

2 (2). Modification 2
In the above embodiment, SVM is adopted as the machine learning algorithm in the event posting selection operation (S1020), but other machine learning algorithms may be adopted. Further, post information PI related to an event may be selected by a method other than machine learning. For example, the posting information PI may be selected (filtered) based on whether or not it has a word related to an event such as “held” or “participation”.

2 (3). Modification 3
In the above embodiment, the event posting selection operation is executed after the spot posting selection operation, but may be executed in the reverse order (that is, S1010 may be executed after S1020). Further, both operations may be executed in parallel. However, when a relatively simple process such as matching is executed in step S1010 and a relatively complicated process such as SVM is executed in step S1020 as in the embodiment, post information PI in step S1010. The processing load in step S1020 can be further reduced by executing step S1020 after sorting.

2 (4). Modification 4
In the event information extraction operation (step S1030), the event information extraction unit 110 may exclude the posted information PI posted by the NG user registered in the NG user list from the target for extracting the event information EI. Any user can be registered in the NG user list. For example, a user corresponding to an account frozen (suspended) in the blog server BS may be registered as an NG user.

2 (5). Modification 5
In the above embodiment, the event information integration operation (S1040) is executed based on the standardized common substring length indicating the similarity of the character strings, but based on another index (for example, edit distance) indicating the similarity. Integration of event information EI may be performed. For example, the event information EI may be integrated when the edit distance between two event names corresponding to the same spot event pair is equal to or less than the threshold Th2. The editing distance means the number of editing operations (insertion, deletion, and replacement) necessary to transform a character string into another character string. In the above modification, a standardized edit distance may be used. The standardization method is arbitrary. For example, as in the embodiment, a method of dividing the longest common substring length by the longer character string length of the spot identical event pairs may be employed.

  Further, the event information integration operation (S1040) may be executed based on the latitude / longitude included in the event information EI. For example, in step S1042, the event information integration unit 120 may group the plurality of event information EI for each mesh-like region (mesh) divided based on latitude and longitude. The above grouping is particularly suitable in the configuration of the first modification that selects the posting information PI based on the latitude and longitude (the configuration in which the posting information PI may not include the spot name). A specific example of the above mesh is a standard area mesh (for example, a quarter area mesh) provided by the Statistics Bureau of the Ministry of Internal Affairs and Communications in Japan.

2 (6). Modification 6
The event information extraction unit 110 may exclude unnecessary event information EI from the extracted event information EI. For example, the event information extraction unit 110 may exclude event information EI including words (for example, NG words, etc.) registered in the exclusion list.

  The spot names included in the spot information table 140 may be included in the above exclusion list. The name of the spot itself is not an event name. Therefore, it is preferable that the event information extraction unit 110 excludes the event information EI including the event name that matches the spot name registered in the exclusion list.

2 (7). Modification 7
The event information extraction unit 110 may shape the extracted event name. For example, when a delimiter (delimiter) such as a bracket is included in the event name, the event information extraction unit 110 preferably deletes the delimiter from the event name. This is because the delimiter itself is not likely to be included in the event name.

2 (8). Modification 8
One posting information PI may include a plurality of event names and a plurality of holding spot names. For example, the posting information PI is “[Event name 1] will be held at [Spot name 1]. [Event Name 2] will be held at [Spot Name 2]. Is included. In the above case, in step S1030, event information EI associated with an incorrect association may be extracted. For example, [Event Name 1] and [Event Spot Name 2] are associated with each other.

  In order to suppress the erroneous association as described above, when the event information extraction unit 110 extracts the event information EI, the distance between the event name and the holding spot name (the number of characters existing between them) is minimized. It is preferable to associate with. Also, the association between the event name and the event spot name is performed based on the weighted distance obtained by weighting each character type (for example, making the weight of punctuation marks higher than the weight of other characters) instead of the simple number of characters. May be.

2 (9). Modification 9
In the event information extraction operation, various information other than the event name, the holding spot name, and the holding date may be extracted. For example, the event information extraction unit 110 may analyze the posted information PI, extract external link information (such as a URL) included in the posted information PI, and include it in the event information EI. Further, the event information EI may analyze the posted information PI, extract a feature word included in the posted information PI, and include it in the event information EI. A technique for extracting feature words is arbitrary. For example, even if the event information extraction unit 110 calculates tf-idf (term frequency-inverse document frequency) for each word included in the posting information PI and extracts a word having the largest value of tf-idf as a feature word. Good. Further, one or more words whose tf-idf value is higher than a predetermined threshold may be extracted as feature words.

2 (10). Modification 10
In the above embodiment, the information management server MS, which is one device, executes the event information collection operation. However, an information management system constituted by a plurality of devices may execute the event information collection operation. For example, the spot post selection unit 102, the event post selection unit 104, the event information extraction unit 110, the event information integration unit 120, and the information presentation unit 130 may be implemented by an information management system including two or more server devices.

  A configuration in which the posting information acquisition server AS and the information management server MS are integrated can also be employed.

2 (11). Modification 11
Each function executed by the CPU in the computer apparatus, particularly the information management server MS and the posted information acquisition server AS, may be executed by a programmable logic device such as an FPGA (Field Programmable Gate Array) or a DSP (Digital Signal Processor).

AS: Post information acquisition server, BS: Blog server, CL: Client terminal, EI: Event information, IEI: Integrated event information, MS: Information management server, PI: Post information, 100: Post Information sorting unit, 102 …… Spot post sorting unit, 104 …… Event post sorting unit, 110 …… Event information extraction unit, 120 …… Event information integration unit, 130 …… Information presentation unit, 140 …… Spot information table, 150... Spot posting table, 160... Event posting table, 170... Event information table, 200.

Claims (10)

A spot post sorting unit that sorts post information related to a spot by matching processing based on geographical information included in the post information from a plurality of post information;
Among the plurality of pieces of post information selected by the spot post selection unit, an event post selection unit for selecting post information related to the event by machine learning based on the wording included in the post information,
Analyzing the selected posted information selected by the spot posted selecting section and the event posted selecting section, the event name included in the selected posted information, and the name of the spot where the event corresponding to the event name is held And an event information extraction unit that extracts event information including the information management device. The information management device according to claim 1, further comprising: an event information integration unit that integrates a plurality of event information corresponding to the same event among the extracted event information into one integrated event information. The event information integration unit
The information management device according to claim 2, wherein a plurality of pieces of event information are determined to correspond to the same event based on the similarity of a plurality of event names corresponding to one holding spot name. The event information extraction unit
The information management device according to claim 1, wherein event information including an event name related to a word registered in an exclusion list is excluded from the extracted event information. The exclusion list includes a spot name indicating the name of the spot;
The event information extraction unit
The information management device according to claim 4, wherein event information including an event name that matches the spot name registered in the exclusion list is excluded from the extracted event information. The event information extraction unit
The information management apparatus according to claim 1, wherein post information posted by a user registered in an NG user list is excluded from extraction targets of the event information among the selected post information selected. The event information extraction unit
The information management apparatus according to claim 1, wherein the selected post information is analyzed to extract the event information further including a date and time when the event corresponding to the event name is held. The event information extraction unit
The information management apparatus according to claim 1, wherein the selected post information is analyzed to extract the event information further including a characteristic word included in the selected post information. From a plurality of posting information, based on the geographical information included in the posting information, a spot posting sorting unit that sorts the posting information related to the spot by matching processing,
Out of a plurality of pieces of post information selected by the spot post selection unit, an event post selection unit that selects post information related to an event by machine learning based on a word included in the post information;
Analyzing the selected posted information selected by the spot posted selecting section and the event posted selecting section, the event name included in the selected posted information, and the name of the spot where the event corresponding to the event name is held And an event information extraction unit that extracts event information including the information management system. From among a plurality of pieces of post information, based on geographical information included in the post information, selecting post information related to the spot by matching processing;
From among the selected plurality of posted information, based on the wording included in the posted information, to select the posted information related to the event by machine learning ,
Analyzing the selected posted information, and extracting event information including an event name included in the selected posted information and a holding spot name where an event corresponding to the event name is held. Information management program to be executed.

Images