JP6763732B2 - Extractor - Google Patents

Description

The present invention relates to an extraction device that extracts words from a document.

Conventionally, a technique for predicting (estimating) congestion in a certain place (for example, an event venue) has been proposed. For example, Patent Document 1 proposes a technique for predicting congestion using the position information of a moving body.

Japanese Unexamined Patent Publication No. 2004-213098

However, when the position information of the moving body as described above is used, it is possible to estimate the congestion at the place where the moving body was located, but it is difficult to estimate the congestion at another place. Therefore, it is conceivable to estimate the congestion based on the contents of the document including the information indicating the location (for example, the name of the location and the name of the event).

For example, a document containing the event name or announcement of the event to be estimated is divided into words by morphological analysis, and a set of words (Bag of Words) is used as the feature of the event, and the feature and congestion in the past. A method of determining whether or not the estimated target location is likely to be congested can be considered by calculating the degree of similarity with the feature quantity of the event. However, since many words (noise) that are not related to the estimation of congestion are included, it may not be possible to accurately estimate whether or not the event to be estimated is congested. Even if the inverse document frequency (IDF) is simply applied to a set of documents including the event name, words specific to the event that are not related to the estimation of congestion will be extracted, so it is appropriate. I can't extract words.

The present invention has been made in view of the above problems, and appropriate terms for estimating the congestion status of the location of the congestion status estimation target based on the contents of the document corresponding to the location of the congestion status estimation target. It is an object of the present invention to provide an extraction apparatus for extracting into.

In order to solve the above-mentioned problems, the extraction device of the present invention is a term used when estimating the congestion status of the location of the congestion status estimation target based on the contents of the document corresponding to the congestion status estimation target location. It is an extraction device that extracts related feature words, and is acquired by an information acquisition unit that acquires a document that is an extraction source and congestion information that indicates the congestion status of a place corresponding to the content of the document, and an information acquisition unit. It is provided with an extraction unit that extracts congestion-related characteristic words from a document that is an extraction source, based on the degree of appearance of words in a document corresponding to a location of a congestion situation in which the congestion information satisfies a preset condition.

According to the present invention, it is possible to appropriately extract congestion-related characteristic words peculiar to a document corresponding to a place in a congestion situation that satisfies a predetermined condition. As a result, the location corresponding to the document based on the result of determining whether or not the extracted word is included in the document containing the information indicating the congestion status estimation target (for example, the document regarding the location of the event scheduled to be held in the future). It is possible to estimate the congestion situation of. In this way, it is possible to appropriately extract congestion-related characteristic words for estimating the congestion status of the location of the congestion status estimation target based on the content of the document including the information indicating the congestion status estimation target.

According to the present invention, it is possible to appropriately extract words for estimating the congestion status of the location of the congestion status estimation target based on the content of the document corresponding to the congestion status estimation target location.

It is a block diagram which shows the functional structure of the estimation apparatus which concerns on one Embodiment. It is a figure which shows the data example of a POI database. It is a figure which shows the data example of the dynamic person database. It is a figure which shows the data example of the event database. It is a figure which shows the hardware configuration of the estimation apparatus. It is a flowchart which shows the process executed by the estimation apparatus. It is a figure explaining the process of extracting a plurality of words.

Hereinafter, embodiments of the extraction device according to the present invention will be described in detail together with the drawings. In the description of the drawings, the same elements are designated by the same reference numerals, and duplicate description will be omitted.

FIG. 1 shows the estimation device 10 according to the present embodiment. The estimation device 10 (extraction device) is a device that estimates the congestion status of the location of the congestion status estimation target based on the contents of the document corresponding to the location of the congestion status estimation target. In addition, the estimation device 10 uses congestion-related characteristic words when estimating the congestion status of the location where the congestion status is estimated. Prior to the estimation, the estimation device 10 extracts the congestion-related characteristic words from documents other than the document corresponding to the location of the congestion situation estimation target. The place where the congestion situation is estimated is, for example, the place where the event is held. An "event" is any event that is uniquely identified by, for example, the event name, venue, and duration. Specific examples of events include large-scale events such as sports events and band live events (events that are expected to have a large number of participants), and small-scale events such as CD release events and mini-lives (events that are expected to have a large number of participants). There are also events) where the number of participants is expected to be small.

The above document is, for example, text data posted on a microblog by various users on the Internet. In this case, the text data of one post is regarded as one document. This document may contain information about the event. The estimation device 10 extracts event information from the document. Here, the event information is information indicating an event, and includes information indicating an event name, a venue, and a period of the event.

The estimation device 10 is connected to a text output device (for example, a server that provides a microblogging service) that outputs a document so that the above document can be acquired (received) via a network such as the Internet, and information. Can be sent and received.

The estimation device 10 extracts event information from the above document, determines the event congestion status of the event information, and extracts congestion-related characteristic words from the document based on the determination result. Further, the estimation device 10 refers to a document (a document corresponding to a location for which the congestion status is estimated) related to an event (for example, an event to be held in the future), and uses the extracted congestion-related characteristic words to congest the event. Estimate the situation.

The estimation device 10 includes an information acquisition unit 11, an extracted document database 12, a POI database 13, a dynamic number database 14, an event database 15, an extraction unit 16, an estimation target document database 17, and an estimation unit 18. To be equipped.

The information acquisition unit 11 acquires a document to be extracted and congestion information indicating a congestion status of a place corresponding to the content of the document. Further, the information acquisition unit 11 specifies a place corresponding to the content of the document based on the content of the acquired document, acquires an index value indicating the degree of congestion of the place, and obtains congestion information based on the index value. By generating it, congestion information is acquired.

The information acquisition unit 11 acquires a document from the text output device by requesting the text output device to transmit a document to be an extraction source at a predetermined timing (for example, once a month).

The information acquisition unit 11 may acquire all the documents of each user from the text output device, or for example, acquire only the documents posted by the user in a certain period such as the past one month. May be good. The information acquisition unit 11 stores the acquired document in the extracted document database 12. The extracted document database 12 is a means for storing documents.

The information acquisition unit 11 acquires a document from the extracted document database 12 at a predetermined timing (for example, once a month) and extracts event information from the document. The information acquisition unit 11 is, for example, Reference 1 (Event Information Extraction Technology Using Microblog, Wataru Yamada, Yu Kikuchi, Keiichi Ochiai, Daisuke Torii, Hiroshi Inamura, Ken Ota, Information Processing Society of Japan Journal Vol.57 No. Event information is extracted by a method using SVM (Support Vector Machine) and CRF (Conditional Random Fields) as shown in .1 123-132 (Jan.2016)). According to this method, character strings indicating the event name, the holding period, and the holding place can be extracted from the above-mentioned documents. In this way, the information acquisition unit 11 specifies the venue corresponding to the content of the document based on the content of the acquired document.

For example, "5/2 (Sat) A team vs B team left outfield reserved seat pair ticket 2015 professional baseball A team sponsored game A team vs B team left outfield reserved seat 2 consecutive number ticket. Date May 2, 2015 (Saturday) ) 14:00 Match start C Baseball field seat Left outfield 8 http://t.co/xxyyzz If there is a document, the event name will be "Team A vs Team B" and the event will be held on "May 2, 2015". "C baseball field" is extracted as a facility name (POI (Point of interest) name) indicating a place, and the extracted event information is associated with the document of the extraction source.

When the information acquisition unit 11 refers to the document stored in the extracted document database 12 and extracts the event information from the document by the above method, the information acquisition unit 11 associates the event information with the document. Subsequently, the information acquisition unit 11 determines the congestion status of the POI indicated by the POI name of the extracted event information, and generates congestion information based on the determined result. As a method of judging the congestion status of POI indicated by the POI name of the extracted event information, the congestion status is determined by the method of judging based on the popularity of POI (method 1) and the ratio of the increase in the number of dynamic people due to the event holding. There is a method for determining (method 2).

The method for determining the congestion status according to the method 1 will be described first. First, the information acquisition unit 11 refers to the POI database 13 and acquires the popularity (index value) of the POI corresponding to the POI name of the extracted event information. The popularity of a POI indicates the degree to which people gather at the POI. Here, an example of data of the POI database 13 is shown in FIG. If the popularity is high, it indicates that there is a high possibility of congestion because many people tend to gather. The popularity of the POI is, for example, an evaluation value of the POI input by the user, a value obtained by totaling the number of searches with a keyword indicating the POI, or the like, and is a value calculated by a known method for calculating the popularity of the POI. is there.

The POI database 13 is a database in which information indicating the popularity of POI is associated with each POI. The database shown in FIG. 2 stores "POI name", "latitude / longitude", "mesh code", and "popularity" in association with each other.

The "POI name" is a name indicating POI. This POI name is information (ID) unique among the records of the POI database 13. The "latitude / longitude" is information indicating the position of the POI, and is a value of latitude / longitude. The "mesh code" indicates the identifier of the mesh corresponding to the latitude and longitude of the POI. The mesh refers to a range in which an area is divided into a mesh shape (rectangular shape) based on latitude and longitude. "Popularity" is the popularity of the POI.

The information acquisition unit 11 refers to the POI database 13 and acquires the popularity of the record including the POI name of each of the extracted event information. That is, the information acquisition unit 11 uses the POI name of each of the extracted event information as a search key, refers to the POI database 13, and acquires the popularity corresponding to the POI name. When the information acquisition unit 11 acquires the popularity corresponding to the POI names of all the event information, the information acquisition unit 11 compares the popularity among the POIs indicated by the POI names of all the event information. As a result of the comparison, the information acquisition unit 11 has a POI event indicated by the POI name, which is the popularity included in a predetermined high-ranking fixed ratio (for example, the top 10%) corresponding to the POI name. Judge that the congestion situation is congestion. As a result of the above comparison, the information acquisition unit 11 has a POI event indicated by the POI name, which is the popularity level included in a predetermined lower fixed ratio (for example, the lower 10%) of the popularity level corresponding to the POI name. Judge that the congestion situation is non-congested. Here, non-congested means that it is vacant (the number of participants is small). As described above, in the method 1, the congestion status of the POI event indicated by the POI name is determined based on the popularity corresponding to the POI name.

Subsequently, a method of determining the congestion status by the method 2 will be described. First, the information acquisition unit 11 refers to the holding period of the extracted event information, and determines whether or not it is before the current time. That is, the information acquisition unit 11 determines whether or not the event of the event information is a thing of the past (whether or not the event has ended). If the event of the event information is not a past event, the information acquisition unit 11 discards the extracted event information. Subsequently, the information acquisition unit 11 refers to the POI database 13 and refers to the POI database 13 using the POI names of all the event information excluding the discarded event information as the search key among the extracted event information. Obtain the mesh code corresponding to the POI name. Subsequently, the information acquisition unit 11 refers to the dynamic number database 14 and acquires the estimated number of people at each time corresponding to the mesh code. Here, an example of data of the dynamic number database 14 is shown in FIG.

The dynamic number database 14 is a database that stores the estimated number of people at each time for each mesh code. The database shown in FIG. 3 stores the "mesh code", the "time", and the "estimated number of people" in association with each other.

The "mesh code" is a mesh code indicating an estimated mesh. The "time" is the time when the estimated number of people is measured. In addition, information (for example, measurement start time and measurement end time) indicating the period in which the estimated number of people was measured may be input to this "time". The "estimated number of people" is the estimated number of people located in the mesh indicated by the mesh code at the above "time". This estimated number of people is, for example, a value calculated based on the number of terminal devices located in each mesh by specifying the number of terminal devices based on the position information of the terminal devices.

The information acquisition unit 11 refers to the dynamic number database 14 and acquires the estimated number of people at each time corresponding to the mesh code. Specifically, the information acquisition unit 11 uses the mesh code as a search key to acquire the estimated number of people for each event information holding period and a certain period (for example, 2 weeks) before the holding period. If the estimated number of people at each time in the mesh code of the POI name of the event information cannot be obtained, it cannot be determined whether or not the event in the event information is congested, so the event information is discarded.

After acquiring the estimated number of people at each time corresponding to the mesh code, the information acquisition unit 11 calculates the average number of people estimated at the time corresponding to the holding period (average number of people during the event). Subsequently, the information acquisition unit 11 calculates the average number of people in the past fixed period before the holding period, and sets this number as the normal number of people. The information acquisition unit 11 calculates a subtraction value obtained by subtracting the normal number of people from the average number of people during the event, or a division value obtained by dividing the average number of people during the event by the normal number of people. When the subtraction value or the division value (index value) is equal to or greater than a predetermined congestion threshold value, the information acquisition unit 11 determines that the event of the target event information is congested. Further, when the subtraction value or the division value is less than a predetermined non-congestion threshold value, the information acquisition unit 11 determines that the event of the target event information is non-congestion. The congestion threshold is equal to or higher than the non-congestion threshold. When the subtraction value or the division value is less than the congestion threshold value and equal to or more than the non-congestion threshold value, the information acquisition unit 11 determines that the event of the target event information is not congested and is not non-congested. .. As described above, in the method 2, the congestion status of the event in the event information is determined based on the result of comparing the estimated number of people during the holding period of the event with the estimated number of people in the normal time.

When the information acquisition unit 11 determines the event congestion status of each event information by the method 1 or the method 2, based on the determination, the information acquisition unit 11 includes flag information (congestion flag) indicating whether or not the event is congested as the congestion information. Generates flag information (non-congested flag) indicating whether or not it is non-congested. When the information acquisition unit 11 determines that the congestion status is congested as a result of determining the congestion status, the congestion flag is set to "True" and the non-congestion flag is set to "False". When the information acquisition unit 11 determines that the event of the event information is non-congested, the congestion flag is set to "False" and the non-congested flag is set to "True". Further, when the information acquisition unit 11 determines that the event of the event information is neither congested nor non-congested as a result of determining the congestion status, the congestion flag and the non-congested flag are set to "False". That is, there may be an event in which the congestion flag and the non-congestion flag are set to "False". In this way, the information acquisition unit 11 acquires the congestion information by generating the congestion flag and the non-congestion flag indicating the congestion status of the POI name of the event information. The information acquisition unit 11 may generate either a congestion flag or a non-congestion flag. After generating the congestion flag and the non-congestion flag of the event of each event information, the information acquisition unit 11 uses the event information, the document, the congestion flag, and the non-congestion flag as event data as the event data. The data is stored in the event database 15. Here, a data example of the event database 15 is shown in FIG.

The event database 15 stores the "event name", the "POI name", the "holding period", the "posted text", the "congestion flag", and the "non-congestion flag" in association with each other.

The "event name" is the event name of the event information. The "POI name" is the POI name of the event information. The "holding period" is the holding period of event information. The "posted text" is a document corresponding to the event information. The “congestion flag” is a congestion flag generated by the information acquisition unit 11. The “non-congestion flag” is a non-congestion flag generated by the information acquisition unit 11.

The information acquisition unit 11 extracts event information from the document, stores the event data of all the extracted event information in the event database 15, and then notifies the extraction unit 16 that the event data has been stored. The extraction unit 16 extracts congestion-related characteristic words accordingly.

The extraction unit 16 extracts congestion-related feature words from a document to be extracted based on the degree of appearance of words in a document corresponding to a place in a congestion situation in which a congestion flag or a non-congestion flag satisfies a preset condition. Is. In addition, the extraction unit 16 has a second condition in which the degree of appearance of words in a document corresponding to a place in a congestion situation in which a congestion flag or a non-congestion flag is set in advance and satisfies the first condition is different from the first condition. Congestion-related characteristic words are extracted based on the degree of appearance of words in the document corresponding to the location of the congestion situation that satisfies the conditions.

When the extraction unit 16 receives the notification from the information acquisition unit 11 that the event data is stored in the event database 15, the extraction unit 16 starts the extraction process of the congestion-related feature words. Specifically, when extracting the congestion-related feature words at the time of congestion, the extraction unit 16 morphologically analyzes the document of the event in which the congestion flag is "True" (first condition) by a known technique, and the congestion-related features. Extract morphemes as word candidates. The extraction unit 16 calculates a score for determining whether or not the morpheme is a congestion-related characteristic word for each extracted morpheme. The extraction unit 16 calculates the frequency of appearance of the extracted morphemes in the set of event data documents (congestion event set) whose congestion flag is "True". In addition, the extraction unit 16 is a set of documents of all event data (a set of documents of event data in which the congestion flag is "True" or the congestion flag is "False" (second condition)). Calculate the frequency of appearance. The extraction unit 16 calculates the number of documents in the congestion event set, the calculated number of documents, the appearance frequency of the extracted morphology in the congestion event set, and the extracted morphology in the document set of all event data. Calculate the score based on the frequency of appearance. For example, the score of a morpheme that is a candidate for a congestion-related feature word at the time of congestion is calculated by the following equation (1) (when the dice coefficient is used). Score _word is a score of a morpheme that is a candidate for a congestion-related feature word at the time of congestion. df (word in set _congestion ) is the frequency of appearance of morphemes in the congestion event set. num (set _congestion ) is the number of documents in the above-mentioned congestion event set. df (word in set _event ) is the frequency of appearance of morphemes in a set of documents of all event data.
Score _word = 2 × df (word in set _setting ) / {num (set _setting ) + df (word in set _event )} ・ ・ ・ (1)

The above formula is not only the set of documents whose congestion flag is "True", but also the appearance frequency of morphological elements in the set of documents of all event data (the morphological elements in the set including the documents of event data whose congestion flag is not "True"). It is also based on the frequency of appearance of. That is, a score is calculated based on the appearance degree of the morpheme in the document in which the congestion flag of the event data is "True" and the appearance degree of the morpheme in the document of all event data, and the congestion-related characteristic word is calculated based on the score. Is extracted. If the frequency of appearance of morphemes in the set of documents of all event data is high, the score will not be high even if the frequency of appearance of morphemes in the set of events during congestion is high. Therefore, if the morpheme is a word peculiar to the event set at the time of congestion, the score will be high. Instead of the set of all the above event data documents, the set of event data documents whose congestion flag is "True" and the set of event data documents whose non-congestion flag is "True" are combined. It may be a set.

When extracting the congestion-related feature words during non-congestion, the extraction unit 16 morphologically analyzes the document of the event in which the non-congestion flag is "True", and morphologically analyzes the document as a candidate for the congestion-related feature words during non-congestion. Is extracted, and the score shown in the following formula (2) is calculated for each morpheme. Score _word is a score of a morpheme that is a candidate for congestion-related feature words during non-congestion. df (word in set _vacant ) is the frequency of appearance of morphemes in a non-congested event set (a set of event data documents whose non-congested flag is "True"). num (set _vacant ) is the number of documents in the non-congested event set. df (word in set _event ) is the frequency of appearance of morphemes in a set of documents of all event data.
Score _word = 2 × df (word in set _vacant ) / {num (set _vacant ) + df (word in set _event )} ・ ・ ・ (2)

When the extraction unit 16 calculates the score, if the calculated score is equal to or higher than a predetermined score threshold value (for example, 0.3), the extraction unit 16 uses the candidate morpheme of the congestion-related feature word of the score as the congestion-related feature word. Remember. In this way, the extraction unit 16 extracts the congestion-related characteristic words.

For example, the extraction unit 16 sets the “outfield reserved seat” and the “ticket” included in the document of the record R1 of the event database 15 shown in FIG. 4 as congestion-related characteristic words at the time of congestion (suggesting that the scale of the event is large). Extract as word). Further, the extraction unit 16 refers to the “mini-live”, “signing session”, and “CD” included in the document of the record R2 of the event database 15 as congestion-related characteristic words (suggesting that the scale of the event is small) during non-congestion. Extract as word). In addition, the extraction unit 16 extracts "chike" included in the document of record R3 of the event database 15 as a congestion-related characteristic word at the time of congestion. In addition, the extraction unit 16 extracts "live broadcast" included in the document of record R4 of the event database 15 as a congestion-related characteristic word at the time of congestion. Further, the extraction unit 16 extracts "release" and "tower record" included in the document of record R5 of the event database 15 as congestion-related characteristic words at the time of non-congestion. When the extraction unit 16 completes the extraction of the congestion-related characteristic words, the extraction unit 16 notifies the estimation unit 18 that the extraction of the congestion-related characteristic words is completed.

The estimation unit 18 acquires a document corresponding to the location of the congestion status estimation target, and determines the congestion status of the location based on whether or not the congestion-related characteristic words extracted by the extraction unit 16 are included in the document. presume.

First, the estimation unit 18 creates a discriminator (model) by machine learning using data (event data) to which a flag for the presence or absence of congestion is added in a past event. Specifically, when the estimation unit 18 receives the notification from the extraction unit 16 that the extraction of the congestion-related feature words is completed, the estimation unit 18 refers to the event database 15 and acquires each event data. Further, the estimation unit 18 acquires the congestion-related characteristic words at the time of congestion and the congestion-related characteristic words at the time of non-congestion from the extraction unit 16. Subsequently, the estimation unit 18 indicates whether or not each of the elements corresponding to the congestion-related feature words at the time of congestion extracted by the extraction unit 16 includes the congestion-related feature words in the event data document (for example,). , Flags the presence or absence of each of the congestion-related feature words at the time of congestion) Generates a feature vector for congestion judgment. The number of dimensions of this feature vector is the number of congestion-related feature words. In addition, the estimation unit 18 indicates whether or not each of the elements corresponding to the congestion-related feature words at the time of non-congestion extracted by the extraction unit 16 includes the congestion-related feature words in the event data document. Generate a feature vector for judgment. Subsequently, the estimation unit 18 generates learning data for congestion determination using the feature vector for congestion determination as an explanatory variable and the congestion flag of the event data as an objective variable. Further, the estimation unit 18 generates learning data for non-congestion determination using the feature vector for non-congestion determination as an explanatory variable and the non-congestion flag of the event data as an objective variable. Then, the estimation unit 18 executes machine learning using the learning data for congestion determination, inputs a feature vector indicating whether or not each of the congestion-related feature words at the time of congestion is included in the document, and outputs a congestion flag. Generate a model to do. Further, the estimation unit 18 executes machine learning using the learning data for non-congestion determination, and inputs a feature vector indicating whether or not each of the congestion-related feature words at the time of non-congestion includes a document, and is non-congested. Generate a model that outputs flags.

As such a machine learning method, for example, supervised machine learning such as a support vector machine (SVM), a neural network, and boosting represented by Naive Bayes can be used.

The estimation unit 18 uses the model to generate information indicating the congestion state of the event indicated by the document to be estimated for congestion. When the estimation unit 18 acquires a document as an estimation target document from the text output device after the congestion-related characteristic words have been extracted (after acquiring the information that the congestion-related characteristic words have been extracted from the extraction unit 16). , Event information is extracted from the document by the same method as described above, and the event information and the document are associated with each other and stored in the estimation target document database 17.

The estimation unit 18 generates a feature vector indicating whether or not the document includes each of the congestion-related feature words during congestion and the congestion-related feature words during non-congestion, and uses the feature vector as an input value to generate the model. Is used to output a congestion flag or a non-congestion flag. The estimation unit 18 stores the output congestion flag or non-congestion flag in association with the above-mentioned document stored in the estimation target document database 17. When the estimation unit 18 receives a request for confirmation of the congestion status of an event from another device, the estimation unit 18 refers to the estimation target document database 17 and determines the congestion status based on the information of the congestion flag or the non-congestion flag corresponding to the event. Output.

As described above, the estimation unit 18 not only stores the congestion flag or the non-congestion flag indicating the congestion status of the event shown in the document acquired from the text output device, but also is based on the congestion flag or the non-congestion flag. Therefore, information on sentences indicating the congestion status of the event may be generated, and the information on the generated sentences may be provided to the user. For example, the estimation unit 18 may transmit the information of the sentence to the text output device, and the text output device may transmit the information of the sentence to the user.

In this way, the estimation unit 18 outputs the result of estimating the congestion status of the event by outputting the congestion flag or the non-congestion flag for the event shown in the document acquired from the text output device.

Subsequently, FIG. 5 shows the hardware configuration of the estimation device 10 according to the present embodiment. The functional block (constituent part) of the estimation device 10 is realized by any combination of hardware and / or software. Further, the means for realizing each functional block is not particularly limited. That is, each functional block may be realized by one physically and / or logically coupled device, or directly and / or indirectly by two or more physically and / or logically separated devices. (For example, wired and / or wireless) may be connected and realized by these plurality of devices.

For example, the estimation device 10 and the like according to the embodiment of the present invention may function as a computer for extracting congestion-related feature words. The estimation device 10 described above may be physically configured as a computer device including a processor 101, a memory 102, a storage 103, a communication module 104, and the like.

In the following description, the word "device" can be read as a circuit, a device, a unit, or the like. The hardware configuration of the estimation device 10 may be configured to include one or more of the devices shown in the figure, or may be configured not to include some of the devices.

Each function in the estimation device 10 is performed by loading predetermined software (program) on hardware such as the processor 101 and the memory 102, so that the processor 101 performs an operation, and communication by the communication module 104, the memory 102, and the storage 103. It is realized by controlling the reading and / or writing of the data in.

The processor 101 operates, for example, an operating system to control the entire computer. The processor 101 may be composed of a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, registers, and the like. For example, the extraction unit 16 and the estimation unit 18 described above may be realized by the processor 101.

Further, the processor 101 reads a program (program code), a software module, and data from the storage 103 and / or the communication module 104 into the memory 102, and executes various processes according to these. As the program, a program that causes a computer to execute at least a part of the operations described in the above-described embodiment is used. For example, the estimation device 10 may be realized by a control program stored in the memory 102 and operated by the processor 101, and may be realized for other functional blocks as well. Although it has been explained that the various processes described above are executed by one processor 101, they may be executed simultaneously or sequentially by two or more processors 101. The processor 101 may be mounted on one or more chips. The program may be transmitted from the network via a telecommunication line.

The memory 102 is a computer-readable recording medium, and is composed of at least one such as a ROM (Read Only Memory), an EPROM (Erasable Programmable ROM), an EPROM (Electrically Erasable Programmable ROM), and a RAM (Random Access Memory). May be done. The memory 202 may be called a register, a cache, a main memory (main storage device), or the like. The memory 202 can store a program (program code), a software module, or the like that can be executed to carry out the method according to the embodiment of the present invention.

The storage 103 is a computer-readable recording medium, and is, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, or a magneto-optical disk (for example, a compact disk, a digital versatile disk, or a Blu-ray). It may consist of at least one (registered trademark) disk), smart card, flash memory (eg, card, stick, key drive), floppy (registered trademark) disk, magnetic strip, and the like. The storage 103 may be referred to as an auxiliary storage device. The storage medium described above may be, for example, a database, server or other suitable medium that includes memory 102 and / or storage 103.

The communication module 104 is hardware (transmission / reception device) for communicating between computers via a wired and / or wireless network, and is also referred to as, for example, a network device, a network controller, or a network card.

Further, each device such as the processor 101 and the memory 102 is connected by a bus 105 for communicating information. The bus 105 may be composed of a single bus, or may be composed of different buses between the devices.

Further, the estimation device 10 includes hardware such as a microprocessor, a digital signal processor (DSP: Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array). It may be configured, and the hardware may realize a part or all of each functional block. For example, the processor 101 may be implemented on at least one of these hardware. The above is the configuration of the estimation device 10 according to the present embodiment.

Subsequently, the extraction method, which is the operation method of the estimation device 10 according to the present embodiment, will be described with reference to the flowchart of FIG. In this process, first, when the information acquisition unit 11 acquires a document that is an extraction source document, the information acquisition unit 11 registers the document in the extraction document database 12 (step S1). The information acquisition unit 11 extracts the event name, POI name, and holding period from the documents stored in the extracted document database 12 at a predetermined timing. Further, the information acquisition unit 11 generates event data by associating the event name, the POI name, and the congestion flag with the holding period based on the popularity of the POI name of the POI database 13 corresponding to the extracted POI name. Then, the generated event data is registered in the event database 15 (step S2).

The extraction unit 16 performs morphological analysis on the event data document whose congestion flag is "True" among the event data stored in the event database 15, calculates a score for each morpheme, and obtains the score. Is equal to or greater than a predetermined threshold, the morpheme is extracted as a congestion-related feature word. Further, the extraction unit 16 performs morphological analysis on a document whose non-congestion flag is "True" among the event data stored in the event database 15, calculates a score for each morpheme, and obtains the score. Is equal to or greater than a predetermined threshold, the morpheme is extracted as a congestion-related feature word. In this way, the extraction unit 16 extracts the congestion-related characteristic words (step S3). After extracting the congestion-related characteristic words, the estimation unit 18 acquires a document to be an estimation target document and stores it in the estimation target document database 17 (step S4). The estimation unit 18 generates a model based on the information stored in the event database 15 and the congestion-related feature words extracted by the extraction unit 16. The estimation unit 18 extracts event information from the document stored in the estimation target document database 17, and estimates the congestion status of the event extracted from the document using the model (step S5).

In the above-described embodiment, the case where one word morpheme is used as a congestion-related feature word has been described, but a phrase composed of a plurality of words may be used as a congestion-related feature word. For example, as shown in FIG. 7, when there is a document containing the sentence "If you come to XX, there are too many people and it is difficult." , "Come", "Tara", "People", "Ga", "Many", "Sugi", "Te", "Difficult." Subsequently, the extraction unit 16 removes particles and punctuation marks by a known method, and among the above morphemes, the target morpheme is "XX", "come", "tara", "person", "person", " "Many", "too much", "difficult". Subsequently, the extraction unit 16 combines the suffix (tara, too much) with the stem (coming, many) by a known method, and "XX", "coming", "person", "too many", Divide into "difficult". Subsequently, the extraction unit 16 generates a set of two morphemes as candidates for congestion-related characteristic words among the above morphemes (that is, extracts 2-gram assuming that each of the above morphemes is one character). .. For example, the extraction unit 16 generates a set of morphemes that are candidates for congestion-related feature words such as (XX, when it comes), (when it comes, people), (people, too many), and (too many, hard). .. The method of determining whether or not the generated morpheme set is the target for extracting the congestion-related feature words is the same as the method of calculating the above-mentioned score and making a determination based on the score. As a result, when the score of the group (people, too many) exceeds the threshold value, the extraction unit 16 extracts the group (people, too many) as a congestion-related characteristic word. In this way, the extraction unit 16 may extract a combination of a plurality of words as the congestion-related characteristic words. In this case, it is possible to extract a combination of a plurality of words indicating "many people", "great people", "people in Kuroyama", "immobility", "long queue", and "heavy congestion". That is, the congestion-related superfamily may be defined or extracted by "a combination of a specific word and a word appearing in the vicinity" instead of the word. For example, "great people", "many people", "people in Kuroyama", etc. are extracted by combining "people" and "great, many, Kuroyama". A part of the set of morphemes forming N-gram (2-gram in the above example) is memorized in advance, and the memorized word and the set of the morpheme obtained as a result of morphological analysis of the document are stored. It may be generated. Further, all or a part of the set of morphemes forming N-gram (2-gram in the above example) is stored in advance, and the score for the set of morphemes is calculated and calculated by the same method as described above. Congestion-related feature words may be extracted based on the score obtained.

In the above-described embodiment, the case of determining the congestion of the event location has been described, but the information (name, location indicating the location) indicating the location (for example, a tourist spot) included in the document is extracted and the extracted location. You may decide whether or not there is congestion. In this case, when the congestion is determined by the method 2, the time when the document is posted corresponds to the time indicating the above-mentioned event holding period.

In the above-described embodiment, the case of acquiring the document including the information indicating the place (information of the venue of the event) has been described, but the document which is predetermined to be related to the place may be acquired. ..

In the above-described embodiment, as method 2, a case where it is determined whether or not the event is crowded depending on the magnitude of the increase in the number of dynamic people due to the holding of the event has been described, but whether or not the event venue is crowded depending on the number of check-ins. You may try to judge.

As described above, the information acquisition unit 11 of the estimation device 10 acquires the document to be extracted and the congestion flag and non-congestion flag of the event location corresponding to the contents of the document. The extraction unit 16 extracts the congestion-related characteristic words from the document based on the degree of appearance of the words in the document in which the congestion flag satisfies the condition set in advance.

In this case, the estimation device 10 can appropriately extract the congestion-related characteristic words peculiar to the document corresponding to the location of the congestion situation satisfying the predetermined condition. That is, the estimation device 10 can extract words that suggest congestion of the event or words that suggest that the event is not congested from the characteristic words of the event. As a result, the estimation device 10 determines whether or not the extracted word is included in the document including the information indicating the congestion status estimation target (for example, the description of the event such as the document regarding the location of the event scheduled to be implemented in the future). Based on the results, it is possible to estimate the congestion status of the place corresponding to the document. That is, the future congestion state can be estimated by determining whether or not the above-extracted words are included in the document regarding the location of the event scheduled to be held in the future. In addition, the current or past congestion state can be estimated by determining whether or not the above-extracted words are included in the document regarding the location of the event currently being held or the event held in the past. In this way, the estimation device 10 can appropriately extract congestion-related feature words for estimating the congestion status of the location of the congestion status estimation target based on the contents of the document including the information indicating the congestion status estimation target. it can.

Further, the extraction unit 16 satisfies the degree of appearance of words in a document satisfying a condition in which the congestion flag is set in advance (congestion flag is True) and another condition in which the congestion flag is set in advance (congestion flag is False). Congestion-related characteristic words are extracted based on the degree of appearance of words in. In this case, the estimation device 10 extracts the congestion-related characteristic words based on the documents corresponding to the locations of the appearance degrees of the words in the set of documents having different congestion statuses, and therefore corresponds to the locations of the congestion status satisfying one of the conditions. It is possible to identify words that frequently appear only in documents that are used. As a result, it is possible to appropriately extract congestion-related characteristic words peculiar to the congestion situation satisfying the condition. The extraction unit 16 may extract congestion-related characteristic words only based on the degree of appearance of words in a document in which the congestion flag satisfies a preset condition (congestion flag is True). When the degree of appearance is equal to or higher than a predetermined threshold value, the extraction unit 16 extracts candidates for the congestion-related feature words as congestion-related feature words.

The information acquisition unit 11 identifies a location (POI name of event information) corresponding to the content of the document based on the content of the acquired document (document stored in the extracted document database 12), and the degree of congestion at the location. The index value (popularity of POI, ratio of dynamic number of people) indicating the above is acquired, and the congestion flag is acquired by generating the congestion flag based on the index value.

In this case, the estimation device 10 does not need to acquire the congestion flag from another device, and can extract words for estimating the congestion status of the place corresponding to the contents of the document with a simple configuration. Note that the information acquisition unit 11 acquires event data including a congestion flag or a non-congestion flag based on the determination of the congestion status from another device without determining the congestion status of the place corresponding to the content of the document. You may.

The extraction unit 16 extracts a combination of a plurality of words as congestion-related characteristic words. In this case, it is not possible to estimate the congestion status with one word, but it is possible to extract words that can estimate the congestion status by combining a plurality of words.

The estimation unit 18 acquires a document indicating a congestion status estimation target, and based on whether or not the congestion-related characteristic words extracted by the extraction unit 16 include the document, the congestion in the place corresponding to the document is congested. Estimate the situation.

In this case, the congestion status of the place corresponding to the document can be appropriately estimated based on the result of determining whether or not the congestion-related characteristic words are included. Further, since the estimation device 10 estimates using the congestion-related feature words, it is possible to present the estimation result of the congestion situation even when the statistical processing of the dynamic number of people is not real-time. Further, the estimation device 10 can present easy-to-understand information by displaying a sentence indicating the congestion status of the event based on the above-mentioned congestion flag instead of the numerical value based on the number of dynamic people. In the estimation device 10, the congestion status of the location corresponding to the document indicating the congestion status estimation target is estimated using the congestion-related feature words extracted by the extraction unit 16, but the estimation process is performed by another device. It may be.

In the method of determining the presence or absence of congestion based on the popularity and scale of the POI where the event is held, events with the same POI but different scales of participants may be held at the same time, so the congestion situation is estimated appropriately. Can't. However, if the congestion status is estimated based on whether or not the congestion-related characteristic words are included in the document as described above, the characteristic words peculiar to the congestion status are not used without using the popularity of the place where the congestion status is estimated. Since the estimation is based on the presence or absence of, the congestion situation can be estimated appropriately.

In addition, when estimating congestion using information on the number of dynamic people in the area including the POI where the event is held, the congestion status at the location of the event can be estimated, but the congestion status of other events is estimated. Can't. However, if the congestion status is estimated based on whether or not the congestion-related feature words are included in the document as described above, the feature words peculiar to the congestion status are not used for the dynamic number of people at the location where the congestion status is estimated. Since the estimation is based on the presence or absence of, the congestion situation can be estimated appropriately.

Software is an instruction, instruction set, code, code segment, program code, program, subprogram, software module, whether called software, firmware, middleware, microcode, hardware description language, or another name. , Applications, software applications, software packages, routines, subroutines, objects, executable files, execution threads, procedures, features, etc. should be broadly interpreted to mean.

Further, software, instructions, and the like may be transmitted and received via a transmission medium. For example, the software uses wired technology such as coaxial cable, fiber optic cable, twisted pair and digital subscriber line (DSL) and / or wireless technology such as infrared, wireless and microwave to websites, servers, or other When transmitted from a remote source, these wired and / or wireless technologies are included within the definition of transmission medium.

The information, signals, etc. described herein may be represented using any of a variety of different techniques. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or photons, or any of these. It may be represented by a combination of.

The terms described herein and / or the terms necessary for understanding the present specification may be replaced with terms having the same or similar meanings.

The terms "system" and "network" as used herein are used interchangeably.

Further, the information, parameters, etc. described in the present specification may be represented by an absolute value, a relative value from a predetermined value, or another corresponding information. ..

Also, the term "determining" as used herein includes a wide variety of actions. A "judgment" is, for example, judging, calculating, computing, processing, deriving, investigating, looking up (eg, table, database or another). (Search in the data structure of), confirming (ascertaining) can be regarded as "judgment", etc. Also, "judgment" includes receiving (eg, receiving information), transmitting (eg, transmitting information), input, output, accessing (accessing) ( For example, it may include the fact that (accessing the data in the memory) is regarded as "judgment". In addition, "judgment" may include that "judgment" is regarded as "resolving", "selecting", "choosing", "establishing", "comparing", and the like. That is, "judgment" may include considering some action as "judgment".

The phrase "based on" as used herein does not mean "based on" unless otherwise stated. In other words, the statement "based on" means both "based only" and "at least based on".

Further, the "means" in the configuration of each of the above devices may be replaced with a "part", a "circuit", a "device" and the like.

When the terms "first", "second", etc. are used herein, any reference to the elements does not generally limit the quantity or order of those elements. These designations can be used herein as a convenient way to distinguish between two or more elements. Thus, references to the first and second elements do not mean that only two elements can be adopted there, or that the first element must somehow precede the second element.

As long as "inclusion," "comprising," and variations thereof are used herein or in the claims, these terms are as comprehensive as the term "comprising." Intended to be targeted. Furthermore, the term "or" as used herein or in the claims is intended not to be an exclusive OR.

In the present specification, a plurality of devices shall be included unless the device is apparently only one in the context or technically.

In addition, each aspect / embodiment described in the present specification includes LTE (Long Term Evolution), LTE-A (LTE-Advanced), SUPER 3G, IMT-Advanced, 4G, 5G, FRA (Future Radio Access), W. -CDMA®, GSM®, CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB (Ultra-WideBand) ), Bluetooth®, and other systems that utilize suitable systems and / or extended next-generation systems based on them.

Further, the order of the processing procedures, sequences, flowcharts, etc. of each aspect / embodiment described in the present specification may be changed as long as there is no contradiction. For example, the methods described herein present elements of various steps in an exemplary order, and are not limited to the particular order presented.

The input / output information and the like may be stored in a specific location (for example, memory), or may be managed by a management table. Input / output information and the like can be overwritten, updated, or added. The output information and the like may be deleted. The entered information and the like may be deleted. The input information or the like may be transmitted to another device.

Each aspect / embodiment described in the present specification may be used alone, in combination, or switched with execution. Further, the notification of predetermined information (for example, the notification of "being X") is not limited to the explicit notification, but is performed implicitly (for example, the notification of the predetermined information is not performed). May be good.

The entire disclosure shall include both singular and plural, unless it clearly indicates that it is singular.

Although the present invention has been described in detail above, it is clear to those skilled in the art that the present invention is not limited to the embodiments described in the present specification. The present invention can be implemented as modifications and modifications without departing from the gist and scope of the present invention determined by the description of the claims. Therefore, the description of the present specification is for the purpose of exemplification and does not have any limiting meaning to the present invention.

10 ... estimation device, 11 ... information acquisition unit, 12 ... extracted document database, 13 ... POI database, 14 ... dynamic number database, 15 ... event database, 16 ... extraction unit, 17 ... estimation target document database, 18 ... estimation unit, 101 ... processor, 102 ... memory, 103 ... storage, 104 ... communication module, 105 ... bus.

Claims (4)

It is an extraction device that extracts congestion-related characteristic words, which are words used when estimating the congestion status of the location of the congestion status estimation target based on the contents of the document corresponding to the congestion status estimation target location.
An information acquisition unit that acquires a document to be extracted and congestion information indicating the congestion status of a place corresponding to the content of the document.
An extraction unit that extracts congestion-related characteristic words from a document that is an extraction source based on the degree of appearance of words in a document corresponding to a location in a congestion situation in which the congestion information acquired by the information acquisition unit satisfies a preset condition. When,
Equipped with a,
The information acquisition unit identifies a place corresponding to the content of the document based on the content of the acquired document, and acquires a popularity degree indicating the degree of gathering of people in the place or an index value indicating the number of people in the place. An extraction device that acquires congestion information by generating congestion information indicating the congestion status of the location of the document based on the index value . In the extraction unit, the degree of appearance of words in the document corresponding to the location of the congestion situation in which the congestion information acquired by the information acquisition unit satisfies the first preset condition is different from the first condition. The extraction device according to claim 1, wherein a congestion-related characteristic word is extracted based on the degree of appearance of the word in a document corresponding to a place of a congestion situation satisfying the condition of 2. The extraction device according to claim 1 or 2 , wherein the extraction unit extracts a combination of a plurality of words as congestion-related characteristic words. A document corresponding to the location of the congestion status estimation target is acquired, and the document is based on whether or not the congestion-related characteristic words extracted by the extraction unit are included in the document corresponding to the location of the congestion status estimation target. The extraction device according to any one of claims 1 to 3 , further comprising an estimation unit for estimating the congestion status of the place corresponding to the above.

Images