JP6082085B1 - Genre estimation server, genre estimation method, and genre estimation program - Google Patents

Abstract

A genre estimation server, a genre estimation method, and a genre estimation program capable of efficiently performing genre assignment for unknown spots not prepared in a correspondence database. A genre estimation server 10 is a genre estimation server that estimates a genre corresponding to a user's stay spot information. The genre estimation server 10 includes a spot information acquisition unit 13 that acquires user's stay spot information, The genre classification unit 14 classifies the genre by combining the suffix name prefix or prefix feature included in the acquired spot information and the description on the Internet describing the spot name spot. [Selection] Figure 3

Description

  The present invention relates to a genre estimation server, a genre estimation method, and a genre estimation program for estimating a genre corresponding to user stay spot information.

  In recent years, it is expected to analyze user location information and use it for infrastructure development and marketing. For example, the user's location information (latitude / longitude) is acquired using a GPS sensor attached to a smartphone or the like, and converted into an address to identify the stay spot of the user. By regularly acquiring the stay spot information, the user's daily behavior can be easily estimated.

  On the other hand, acquiring detailed stay spot information may infringe on user privacy. Therefore, as shown in FIG. 21, it is considered to perform behavior analysis while protecting user privacy by abstracting stay spots into “genres” such as “station”, “shopping street”, and “shrine”. (Patent Documents 1 and 2). In order to abstract stay spots into genres, it is common to use a correspondence database of spots and genres. If such a correspondence database is used, as shown in FIG. 22, a stay spot such as “XX country club” can be abstracted into a genre such as “golf course”.

JP 2012-164227 A JP2011-221665A

  It is desirable that the number of existing spots be prepared for the construction of the correspondence database of spots and genres. However, since the preparation work of the correspondence database is complicated, the spots that can be abstracted into genres are biased. The association between the spot and the genre is basically manual work, and the more the amount, the more complicated it becomes. Therefore, as shown in FIG. 23 (a), a spot that is not prepared on the correspondence database cannot be given a genre in advance and becomes an unknown spot U1.

  In order to reduce the preparation work of the correspondence database, a technique for collecting correspondence information between spots and genres from the Internet has been studied (Patent Document 2). For example, as shown in FIG. 23B, information corresponding to the corresponding information (facility name “xx gymnasium”, category “gymnasium”) may be extracted from the Internet. However, a genre cannot be assigned in advance to spots that do not have information corresponding to the corresponding information on the Internet, and become unknown spots U2 and U3. That is, there is a limit to the number that can be collected in the technology for collecting the correspondence information between spots and genres from the Internet.

  The present invention provides a genre estimation server, a genre estimation method, and a genre estimation program that can efficiently perform genre assignment for an unknown spot that is not prepared in a correspondence database in view of the above-described conventional technology. With the goal.

In order to achieve the above object, the invention according to the first aspect is a genre estimation server for estimating a genre corresponding to a user's stay spot information, the spot information acquisition unit for acquiring the user's stay spot information, A genre classification unit that classifies a genre by combining a suffix or prefix feature of a spot name included in the spot information acquired by the spot information acquisition unit and a description on the Internet that describes the spot of the spot name. The genre classification unit performs a spot name determination process for determining a genre based on a suffix or prefix feature of the spot name, and matches two or more genres in the spot name determination process. Analyzing the description on the Internet, focusing on only two or more genres And summarized in that to implement the descriptions determination process determines genre.

The invention according to a second aspect is the invention according to the first aspect, wherein the spot information acquisition unit counts frequent words that are frequent words in the description on the Internet prior to the spot name determination process. Then, the genre classification unit performs the explanatory sentence determination process using the frequent words counted by the spot information acquisition unit after performing the spot name determination process.

The invention according to a third aspect is a genre estimation method in which a genre estimation server estimates a genre corresponding to a user's stay spot information, wherein the computer acquires a user's stay spot information, A genre classification step for classifying a genre by combining features of the suffix or prefix of the spot name included in the spot information acquired in the spot information acquisition step, and a description on the Internet that describes the spot of the spot name. In the genre classification step, a spot name determination process for determining a genre from the suffix or prefix feature of the spot name is performed, and when the spot name determination process matches two or more genres, Narrow down candidates to only two or more genres that match, And summarized in that to implement the descriptions determination process determines genres by analyzing the description of the Internet.

The invention according to a fourth aspect is summarized in that a genre estimation program for causing a computer to function as each processing unit of the invention according to the first or second.

  According to the present invention, it is possible to provide a genre estimation server, a genre estimation method, and a genre estimation program that can efficiently perform genre assignment for an unknown spot that is not prepared in the correspondence database.

It is a block diagram of the genre estimation system in this Embodiment. It is a figure for demonstrating the outline | summary of the genre estimation method in this Embodiment. It is a functional block diagram of the genre estimation server in this Embodiment. It is a block diagram of the user latitude longitude information in this Embodiment. It is a block diagram of the spot information table in this Embodiment. It is a block diagram of the spot information cache in this Embodiment. It is a block diagram of the learning dictionary for spot name classification | category in this Embodiment. It is a block diagram of the learning database for explanatory text classification in this Embodiment. It is a block diagram of the genreization position information database in this Embodiment. It is a sequence diagram of the genre estimation in this Embodiment. It is a flowchart which shows operation | movement of the spot information acquisition part in this Embodiment. It is a flowchart which shows the detail of the spot information acquisition process in this Embodiment. It is a flowchart which shows the operation | movement of the genre classification | category part in this Embodiment. It is a flowchart which shows the detail of the spot name determination process in this Embodiment. It is a flowchart which shows the detail of the explanatory note determination process in this Embodiment. It is a flowchart which shows the detail of the similarity calculation process in this Embodiment. It is a figure for demonstrating the specific example of the spot name determination process in this Embodiment. It is a figure for demonstrating the specific example of the explanatory note determination process in this Embodiment. It is a figure for demonstrating the method of automatically producing the learning database for description sentence classification | category in this Embodiment. It is a figure for comparing a prior art document and this Embodiment. It is a figure for demonstrating background art. It is a figure for demonstrating background art. It is a figure for demonstrating the subject of background art.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The following embodiments exemplify a genre estimation system for embodying the technical idea of the present invention, and the configuration of the apparatus and the configuration of data are limited to the following embodiments. is not.

(Genre estimation system)
FIG. 1 is a configuration diagram of a genre estimation system according to an embodiment of the present invention. This genre estimation system is a system that estimates a genre corresponding to user stay spot information, and is used, for example, when collecting user stay spot information periodically for user behavior analysis. Specifically, as shown in FIG. 1, a plurality of terminals 1A, 1B,... (Hereinafter collectively referred to as “terminal 1”) are connected to the genre estimation server 10 via the Internet 3. The terminal 1 is a smartphone or the like having a GPS sensor. The genre estimation server 10 is a device that estimates a genre corresponding to the user's stay spot information. A log collection server 2 that performs preprocessing such as stay point extraction may be interposed between the terminal 1 and the genre estimation server 10.

  The genre estimation server 10 is connected to external systems 4A and 4B via the Internet 3. The external system 4A is a system that provides an address conversion service for converting latitude and longitude into addresses. The external system 4B is a system that provides a Web search service (search engine). The function of providing such an address conversion service or Web search service may be provided in the genre estimation server 10 instead of the external systems 4A and 4B.

(Genre estimation method)
FIG. 2 is a diagram for explaining the outline of the genre estimation method according to the embodiment of the present invention. That is, the genre of a spot (unknown spot) that does not exist in the spot-genre correspondence database is automatically estimated by combining the features of the suffix / prefix and the description on the Internet, and genre assignment to the unknown spot is realized. It is like that.

  FIG. 2A shows an estimation method based on suffix / prefix characteristics. As shown in this figure, a genre is estimated from a spot name by using a suffix / prefix feature dictionary of each genre. In this example, since the suffix of the spot name is “Lake”, it is estimated as “Lake / pond” genre.

  FIG. 2B shows an estimation method based on the similarity of explanatory texts on the Internet. As shown in this figure, for a spot name that cannot be estimated by the characteristics of the suffix / prefix, the genre is estimated from information on the Web including the name of the spot. First, using the existing spot / genre correspondence database as a search key, collecting explanatory texts on the Web representing each spot “golf course”, “shrine”, “ramen”,... Build a vessel. Next, for the unknown spot “XX-chan”, an explanatory text representing the location is acquired from information on the Web. Finally, the degree of similarity is calculated by applying the explanation sentence obtained in this way to the document classifier, and the genre “Ramen” having a high degree of similarity is estimated as the genre for the unknown spot “XX-chan”.

  As described above, in the embodiment of the present invention, the estimation technique based on the features of the suffix / prefix and the estimation technique based on the similarity of the explanatory text on the Internet are combined. Combining these two estimation techniques can offset each other's disadvantages. That is, when the genre is estimated from the features of the suffix / prefix of the spot name such as XX river and XX sea, the genre cannot be estimated for a word having no feature such as “XX-chan”. On the other hand, when collecting “sentences explaining spots” on the Internet using spot names as keys and estimating the genre using document classification technology, the spot names such as “XX River” will explain the spots. The variations of the document become too large, making it difficult to grasp the characteristics of the document and to perform classification. Combining the two estimation techniques can cancel each other's disadvantages, so that genre assignment can be performed efficiently.

  In order to abstract position information into genre information, it is necessary to prepare in advance a database that links spots and genres. Therefore, in the embodiment of the present invention, a document classification technique is applied, and a genre is estimated using “sentences that explain a spot”. The document classification technique is used for news sentence classification and the like, but by applying it to position information genre estimation, it is possible to estimate the genre of a spot without creating a database that links the spot and the genre. However, a database in which a sentence explaining a spot and a genre are linked as learning data is necessary (described later).

  Furthermore, in the embodiment of the present invention, “sentences explaining the spots” are collected from the Web, so it is possible to deal with unknown spots.

(Genre estimation server)
FIG. 3 is a functional block diagram of the genre estimation server 10 according to the embodiment of the present invention. As shown in this figure, the genre estimation server 10 includes a spot information database 11, a genre position information database 12, a spot information acquisition unit 13, a genre classification unit 14, and a learning database 15.

  The spot information acquisition unit 13 is a functional unit that acquires user stay spot information and stores the information in the spot information database 11. The spot information acquisition unit 13 can communicate with the terminal 1 and the external systems 4A and 4B via the Internet 3.

  The genre classification unit 14 is a functional unit that classifies the spot information stored in the spot information database 11 into genres and stores it in the genreized position information database 12. Specifically, the genre is classified by combining the suffix name prefix or prefix feature included in the spot information acquired by the spot information acquisition unit 13 and the description on the Internet that describes the spot name spot. . In this classification, the learning database 15 is referred to. Hereinafter, the function unit that determines the genre from the feature of the suffix or prefix of the spot name is referred to as “spot name determination unit 14A”. Further, a function unit that analyzes the explanatory text on the Internet and determines the genre is referred to as an “explanatory text determination unit 14B”.

  The spot information database 11 is a database that stores spot information. The genre position information database 12 is a database that stores genre position information. The learning database 15 is a database that stores learning data. Details of these databases will be described later. Here, the genreized position information database 12 is mounted on the genre estimation server 10, but the genreized position information database 12 may be mounted on another server.

(Database)
FIG. 4 is a configuration diagram of the user latitude / longitude information 3A. The user latitude / longitude information 3 </ b> A is user position information acquired by the spot information acquisition unit 13 via the Internet 3. As shown in this figure, the user latitude / longitude information 3A associates time, user ID, latitude, longitude, and the like. As the user ID, terminal-specific information (UUID) can be used. The user latitude / longitude information 3A may be GPS data or calculated data calculated by the log collection server 2 or the like.

  FIG. 5 is a configuration diagram of the spot information table 11A. The spot information table 11 </ b> A is a table that stores spot information, and is included in the spot information database 11. As shown in this figure, the spot information table 11A associates item numbers, times, user IDs, latitudes, longitudes, spot names, addresses, explanations, frequent words, and the like. The item number is a serial number for convenience.

  FIG. 6 is a configuration diagram of the spot information cache 11B. The spot information cache 11 </ b> B is a cache that stores spot information, and is included in the spot information database 11. As shown in this figure, the spot information cache 11B associates latitude, longitude, spot name, address, description, frequent words, and the like. Frequent words are words that frequently appear in the explanation.

  FIG. 7 is a configuration diagram of the spot name classification learning dictionary 15A. The spot name classification learning dictionary 15 </ b> A is a learning dictionary used when classifying spot names, and is included in the learning database 15. As shown in this figure, the spot name classification learning dictionary 15A associates prefix / suffix patterns, genres, and the like. The spot name classification learning dictionary 15A may be created manually, or may be created using a commercially available database such as “Japanese vocabulary system”.

  FIG. 8 is a configuration diagram of the explanatory sentence classification learning database 15B. The explanatory sentence classification learning database 15 </ b> B is a learning database used when classifying explanatory sentences, and is included in the learning database 15. As shown in this figure, the explanatory sentence classification learning database 15B is an association of genre names, feature words, and the like. A characteristic word is a characteristic word of the genre. The explanatory sentence classification learning database 15B may be created manually, or may be created by manually collecting explanatory sentences matching each genre, performing morphological analysis, and automatically extracting feature words. . Morphological analysis is a natural language processing technique that divides a sentence into linguistically meaningful units and discriminates the part of speech in each divided unit, and is used to extract verbs, nouns, and the like.

  FIG. 9 is a configuration diagram of the genreization position information database 12. The genreized position information database 12 is a database that stores genreized position information, and associates time, user ID, genre, and the like.

(sequence)
FIG. 10 is a sequence diagram of genre estimation in the embodiment of the present invention. Hereinafter, the configuration of the genre estimation server 10 will be described together with its operation with reference to FIG.

  First, the spot information acquisition unit 13 extracts latitude / longitude from the user position information and transmits it to the external system (address conversion service) 4A. The external system 4A converts the latitude / longitude into a spot name / address and returns the spot name / address to the spot information acquisition unit 13.

  Next, the spot information acquisition unit 13 transmits the spot name / address received from the external system 4A to the external system (Web search service) 4B. The external system 4B collects information on the Web that matches the spot name / address received from the spot information acquisition unit 13 and returns the information to the spot information acquisition unit 13.

  Next, the spot information acquisition unit 13 counts words (frequent words) that frequently appear in information on the Web received from the external system 4B. Then, spot information composed of latitude / longitude, spot name / address, frequent words, etc. is stored in the spot information database 11.

  Next, the spot name determination unit 14A reads spot information from the spot information table 11A and performs spot name determination processing. Here, when the spot name determination process ends normally, genreized position information is generated based on the result and stored in the genreized position information database 12. On the other hand, when the spot name determination process ends in error, the fact is notified to the explanatory note determination unit 14B. The explanatory note determination unit 14B performs an explanatory sentence determination process, generates genreized position information based on the result, and stores it in the genreized position information database 12.

(Operation of spot information acquisition unit)
FIG. 11 is a flowchart showing the operation of the spot information acquisition unit 13. First, when receiving the user position information, the spot information acquisition unit 13 performs a spot information acquisition process (S1 → S2), and then performs a frequent word count process (S3 → S4). Specifically, the frequent words in the explanatory text are counted using a morphological analyzer (such as Mecab) (S3), and the frequent words are stored in the spot information table 11A of the spot information database 11 (S4). Finally, the item number of the spot information table 11A is notified to the genre classification unit 14 (spot name determination unit 14A) (S5).

  FIG. 12 is a flowchart showing details of the spot information acquisition process (S2 in FIG. 11). First, the spot information acquisition unit 13 determines whether or not the user position information is latitude / longitude (S11). When it is determined in S11 that the user position information is latitude / longitude, it is determined whether the latitude / longitude is included in the spot information cache 11B (S12). When it is determined in S12 that the latitude / longitude is not included in the spot information cache 11B, the latitude / longitude is transmitted to the external system (address conversion service) 4A to acquire the address / spot name (S13). If the acquisition process of S13 is not an error, it is determined whether the address / spot name is included in the spot information cache 11B (S14 → S15). If it is determined in S15 that the address / spot name is not included in the spot information cache 11B, the address / spot name is transmitted, the explanatory text is searched for on the web, and the explanatory text is received (S16 → S17).

  On the other hand, if it is determined in S12 that the latitude / longitude is included in the spot information cache 11B, if it is determined in S15 that the address / spot name is included in the spot information cache 11B, and if the reception process in S17 is not an error, S19 Migrate to That is, the user ID, time, latitude, longitude, address, spot name, and description are stored in the spot information table 11A and spot information cache 11B of the spot information database 11 (S19). If it is determined in S11 that the user position information is not latitude / longitude, if the acquisition process of S13 is an error, and if the reception process of S17 is an error, an error is returned (S20).

(Operation of genre classification unit)
FIG. 13 is a flowchart showing the operation of the genre classification unit 14. First, the genre classification unit 14 reads item number data to be determined from the spot information table 11A, and performs spot name determination processing (S21 → S22). Next, when the spot name determination process is normally completed, the genreized position information is stored in the genreized position information database 12 (S23). On the other hand, when the spot name determination process ends in error, an explanatory sentence determination process is performed (S24). Next, when the explanatory sentence determination process ends normally, the genreized position information is stored in the genreized position information database 12 (S23). On the other hand, when the description sentence determination process ends in error, the genreized position information set as “genre unknown” is stored in the genreized position information database 12 (S25).

  FIG. 14 is a flowchart showing details of the spot name determination process (S22 in FIG. 13). First, the spot name determination unit 14A refers to the spot name classification learning dictionary 15A and identifies the genre associated with the spot name (S31). Here, when only one genre is matched, the genre position information to which the specified genre is assigned is returned (S32). On the other hand, if it matches two or more genres or does not match any of the dictionaries, an error is returned (S33).

  FIG. 15 is a flowchart showing details of the explanatory note determination process (S24 in FIG. 13). First, the explanatory note determination unit 14B calculates the similarity with each genre by using the frequent words in the explanatory sentence and the explanatory database classification learning database 15B (S41). As a result, when there is a genre having a similarity higher than a predetermined threshold, genre position information to which the genre having the highest similarity is assigned is returned (S42 → S43). On the other hand, if there is no genre with a similarity higher than a predetermined threshold, an error is returned (S42 → S44). Of course, the threshold value can be changed as appropriate.

  In addition, when it matches both the characteristic suffix and prefix like “Maison XX Park”, it is determined as two genres in the spot name determination process (S31 in FIG. 14). In this case, candidates may be narrowed down to only the two determined genres in the subsequent explanatory sentence determination processing (S41 in FIG. 15).

  FIG. 16 is a flowchart showing details of the similarity calculation process (S41 in FIG. 15). First, the explanatory note determination unit 14B sets one of the appearing words among the frequent words in the genre estimation target record of the spot information table 11A for the feature word existing in the explanatory sentence classification learning database 15B, the word that does not appear. Is vectorized as 0 (S51). Similarly, each record in the explanatory sentence classification learning database 15B is vectorized (S51). Thereafter, the cosine distance is calculated for each record in the explanatory sentence classification learning database 15B (S52).

(Specific example of similarity calculation processing)
Hereinafter, a specific example of the similarity calculation process will be described. Here, it is assumed that the feature word existing in the explanatory sentence classification learning database 15B is D = (marine, creature, ancient, Asia, Koturi, Tonkotsu, fine noodles). In this case, the frequent word in the explanatory text is a vector D = (0, 0, 0, 0, 1, 1, 0) corresponding to “Stick, Tonkotsu”. Also, the vector D1 = (1, 1, 1, 1, 0, 0, 0) for the genre “museum” with the characteristic word “marine, creature, ancient, Asia”. Further, the vector D2 = (0, 0, 0, 1, 1, 0, 1) for the genre “ramen” of the feature word “Asian, heavy, thin noodles” is obtained.

  In such a case, the similarity to the museum genre can be calculated by the following equation.

  Similarly, the similarity with the ramen genre can be calculated by the following equation.

  This similarity calculation method is an example, and various calculation methods can be employed. For example, the vector may be weighted based on the number of appearances. The similarity calculation method may not be the cosine distance. For example, it is possible to use open source Jubatus.

(Specific example of spot name determination processing)
FIG. 17 is a diagram for explaining a specific example of the spot name determination process. As already described, the spot name determination unit 14A determines a genre using a suffix / prefix dictionary unique to the genre. Specifically, the spot name classification learning dictionary 15A is created from the genre and its suffix / prefix pattern. Then, the spot name to be estimated is compared with the spot name classification learning dictionary 15A, and if the corresponding pattern exists in the spot name classification learning dictionary 15A, the genre is specified as the genre of the spot. On the other hand, if the corresponding pattern does not exist in the spot name classification learning dictionary 15A, the genre is undetermined, and the process proceeds to the next process. In this way, since it is based on a spot name pattern, genre estimation is possible even for unknown spot names.

  For example, assume that spot name 1 is “Lake Biwa” and spot name 2 is “XX-chan”. In this case, as shown in FIG. 17, the lake / pond genre is specified as the genre of “Lake Biwa” of the spot name 1. On the other hand, the genre of “XX-chan” of spot name 2 is undecided, and the process proceeds to the next process.

(Specific example of explanation sentence determination processing)
FIG. 18 is a diagram for describing a specific example of the explanatory note determination process. As already described, the explanatory note determination unit 14B performs the determination based on the similarity of the explanatory text. Specifically, nouns, verbs, and adjectives are counted using a morphological analyzer such as Mecab. Apply these words to the document classifier to identify the closest genre. In this way, among unknown spots, it is possible to deal with spot names that are not characterized by suffixes / prefixes and spot names that are likely to exist in other genres.

  For example, as illustrated in FIG. 18A, it is assumed that the description of “XX-chan” in the spot name 2 is “Kokoritonkotsu Ramen ...”. When an explanatory text is acquired from the Internet using a spot name and an address as a key, a plurality of explanatory texts may be found as shown in FIG. Even in such a case, it is possible to specify the necessary explanatory text “Kokotoritonkotsu Ramen ...” by taking address information into account. As shown in FIG. 18 (c), the frequent words in this explanatory note are “Kokoteri, Tonkotsu, Ramen, Popularity, Number, Hakata, Rich”. Therefore, as shown in FIG. 18D, the ramen genre having the highest similarity is adopted as the genre of “XX-chan”.

(Automatic creation of learning database for explanatory text classification)
FIG. 19 is a diagram for explaining a method of automatically creating the explanatory sentence classification learning database 15B. First, explanations of the same genre are collected, and frequent words such as nouns, verbs, and adjectives are counted using a morphological analyzer such as Mecab. For example, when explanations 100 of ramen genres are collected, “thin noodles”, “thick noodles”, “soy sauce”, “miso”, “salt”, “tonkotsu”, “butter”, etc. are counted as frequent words. Among these frequently used words, the words “beer” and “delicious” appearing in other genres such as the cafeteria genre and the pub genre are deleted, and the word 101 that is likely to appear only in the ramen genre is left. This word 101 is used as a feature word of the ramen genre.

(Comparison with prior art documents)
FIG. 20 is a diagram for comparing the prior art document and the embodiment of the present invention.

  As shown in FIG. 20A, in Patent Document 1, a spot-genre correspondence database is used to abstract stay spots into genres. However, the construction of such a correspondence database is complicated, and genres can only be assigned if the location information is included in the genre conversion database.

  As shown in FIG. 20B, in Patent Document 2, in order to reduce the preparation work of the correspondence database, correspondence information between spots and genres is collected from the Internet. Therefore, the construction of the spot-genre correspondence database is easier than that of Patent Document 1. However, a genre cannot be assigned to a spot for which there is no correspondence information between the spot and genre on the Internet.

  As shown in FIG. 20 (c), in the embodiment of the present invention, an estimation technique based on suffix / prefix characteristics and an estimation technique based on similarity of explanatory text on the Internet are combined. Therefore, the construction of the spot-genre correspondence database is easier than that of Patent Document 2. Even if the correspondence information between the spot and the genre is not on the Internet, the genre can be estimated if the spot name itself or related information exists.

  As described above, the genre estimation server 10 according to the embodiment of the present invention is a genre estimation server that estimates the genre corresponding to the stay spot information of the user, and the spot information acquisition unit 13 that acquires the stay spot information of the user. And a genre classification unit that classifies a genre by combining a suffix or prefix feature of a spot name included in the spot information acquired by the spot information acquisition unit 13 and a description on the Internet that describes the spot name spot 14. Thereby, since two estimation techniques that cancel each other's disadvantages can be combined, it is possible to efficiently perform genre assignment for unknown spots that are not prepared in the correspondence database.

  Specifically, the genre classification unit 14 performs a spot name determination process for determining a genre based on the suffix or prefix feature of the spot name and does not match one genre in the spot name determination process. An explanatory sentence determination process for analyzing the above explanatory sentence and determining the genre is performed. As a result, when the spot name determination process ends normally, it is not necessary to execute the explanatory sentence determination process, so that the genre can be estimated accurately and quickly.

  In addition, when the spot name determination process matches two or more genres, the genre classification unit 14 may perform the explanatory sentence determination process by narrowing down candidates to only the two or more genres that match. Thereby, since it is not necessary to calculate the similarity with each genre, it is possible to estimate the genre more quickly.

  Further, prior to the spot name determination process, the spot information acquisition unit 13 counts frequent words that are frequently appearing in the explanatory text on the Internet, and the genre classification unit 14 performs the spot name determination process, The explanatory sentence determination process may be performed using the frequent words counted by the information acquisition unit 13. Thus, frequently used words can be stored in the spot information database 11 together with information such as latitude and longitude, which is efficient.

  As described above, according to the embodiment of the present invention, it is possible to efficiently perform genre assignment for unknown spots that are not prepared in the correspondence database. That is, a genre can be automatically assigned to an unknown spot that has not been pre-registered if the spot name itself or related information is found. As a result, it is possible to reduce the database construction work for associating spots and genres. In addition, since more points can be genre converted than before, it is possible to acquire the genre of the user stay spot without any bias. If this technology is applied to the tourism field, it is possible to estimate the movement pattern of tourists and use it as basic information for infrastructure development in the local government, or to make flexible recommendations according to the movement pattern. .

  Note that the present invention can be realized not only as the genre estimation server 10 but also as a genre estimation method using characteristic function units included in the genre estimation server 10 as steps, or causing a computer to execute these steps. It can also be realized as a genre estimation program. Needless to say, such a program can be distributed via a recording medium such as a CD-ROM or a transmission medium such as the Internet.

DESCRIPTION OF SYMBOLS 1,1A, 1B ... Terminal 2 ... Log collection server 3 ... Internet 3A ... User latitude / longitude information 4A, 4B ... External system 10 ... Genre estimation server 11 ... Spot information database 11A ... Spot information table 11B ... Spot information cache 12 ... Genre Location information database 13 ... Spot information acquisition unit 14 ... Genre classification unit 14A ... Spot name determination unit 14B ... Description sentence determination unit 15 ... Learning database 15A ... Spot name classification learning dictionary 15B ... Explanation sentence classification learning database

Claims (4)

A genre estimation server that estimates a genre corresponding to user stay spot information,
A spot information acquisition unit for acquiring user stay spot information;
A genre classification unit for classifying a genre by combining a suffix or prefix feature of a spot name included in the spot information acquired by the spot information acquisition unit, and a description on the Internet that describes the spot of the spot name; equipped with a,
The genre classification unit performs a spot name determination process for determining a genre from features of the suffix or prefix of the spot name, and when two or more genres are matched in the spot name determination process, the matched 2 The candidate sentence is narrowed down to only one or more genres, and the explanatory sentence determination process for analyzing the explanatory sentence on the Internet and determining the genre is performed.
A genre estimation server characterized by that. Prior to the spot name determination process, the spot information acquisition unit counts frequent words that are frequent words in the description on the Internet,
The genre classification unit, genre of claim 1, wherein the after performing the spot name determination process, implementing the explanation determination process using frequent words counted by said spot information acquiring unit Estimated server. A genre estimation method in which a genre estimation server estimates a genre corresponding to user stay spot information,
Computer
A spot information acquisition step for acquiring user stay spot information;
A genre classification step for classifying a genre by combining a suffix or prefix feature of a spot name included in the spot information acquired in the spot information acquisition step, and a description on the Internet that describes the spot of the spot name; Run
In the genre classification step, a spot name determination process for determining a genre based on the suffix or prefix feature of the spot name is performed, and two or more genres are matched in the spot name determination process. The candidate sentence is narrowed down to only one or more genres, and the explanatory sentence determination process for analyzing the explanatory sentence on the Internet and determining the genre is performed
The genre estimation method characterized by this. A genre estimation program for causing a computer to function as each processing unit according to claim 1 .

Images