JP2015102933A - Place-name position estimation method, place-name position estimation apparatus, and place-name position estimation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of estimating the position of the name of a place.SOLUTION: A POI data storage section 12 stores multiple pieces of POI data formed by associating the name of a POI with position information of the POI. A POI related-document extraction section 13 extracts document data including the name of the POI from the document data. A document analysis section 14 extracts the name of a place related to the POI data from the extracted document data, and calculates the number of times that the name of the place appears in the document data. A place-name position estimation section 15 calculates the position of the name of the place, by use of weight of the POI data and the position information of the POI. The frequency of appearance of the name of the place is the weight of the POI data with respect to the name of the place.

Description

  The present invention relates to a technique for estimating the position of a place name.

  Currently, there is a need for a technique for estimating the geographical position of a place name included in document data. For example, this is a technique for displaying the location and latitude / longitude on the map of “Chinatown” corresponding to the place name from document data “I went to a store called AA in Chinatown with a friend”.

  As an example of such a technique, there is a technique for converting a place name into a real-world position using dictionary data in which the place name is associated with its position information (Prior Art 1). Further, Patent Document 1 discloses a technique of extracting a plurality of points related to a place name and calculating a polygon including each point as a place name (Prior Art 2).

JP 2009-37316 A

Imai, “Extracting Place Names from Web Documents Using POI Information”, Information Processing Society of Japan, 12th Information Science and Technology Forum Proceedings Vol. 2, FIT2013, D-019, p.127-128 Okazaki, 1 other, “Concise and Fast Similar String Search Algorithm for Inter-set Similarity”, Language Processing Society of Japan, Natural Language Processing, Vol.13, No.2, April 2006, p.1-29

  However, according to the prior art 1, since the place name is converted into the position using the dictionary data, it cannot be converted when the place name to be estimated is not registered in the dictionary data. Even if the place name is in the dictionary data, the registered place name position may be different from the generally recalled position.

  For example, when converting the place name “Shinagawa” to a position, even if the central point of “Shinagawa-ku, Tokyo” is displayed as the position of “Shinagawa” from the dictionary data of the address, “Shinagawa” generally recalled is This is the case of Shinagawa Station on the north side of the center point.

  On the other hand, according to the prior art 2, since a point related to the place name is used, a position close to a generally recalled position can be obtained. However, since the place name is converted into a polygonal face, it is impossible to grasp which position in the face matches the place name.

  The present invention has been made in view of the above circumstances, and an object thereof is to improve the estimation accuracy of the position of a place name.

  The place name position estimation method according to claim 1 is a storage step of storing a plurality of point data in which a name of a predetermined point and position information of the point are associated with each other by a computer, and the storage unit. An extraction step of reading the name of the spot from the document data and extracting the document data including the name of the spot from a plurality of document data, and extracting the place name related to the spot from the extracted document data, and the place name is the document data An analysis step for calculating the frequency of occurrence of the place name, and the calculated appearance frequency of the place name as the importance of the point with respect to the place name, and using the importance of the point and the position information of the point, the position of the place name is determined It has an estimation step to calculate.

  According to the present invention, a plurality of point data in which a name of a predetermined point and position information of the point are associated are stored, document data including the name of the point is extracted from the plurality of document data, and extracted. The location name related to the location is extracted from the document data, the frequency of the location name appearing in the document data is calculated, and the importance of the location and the location of the location are calculated using the frequency of appearance of the location as the importance of the location with respect to the location name. Since the position of the place name is calculated using the information, the position of the place name can be reliably estimated.

  The place name position estimation method according to claim 2 is the place name position estimation method according to claim 1, wherein, in the analysis step, a plurality of combinations in which the name of the point, the place name, and the appearance frequency of the place name are associated with each other. Output data, and in the estimating step, combine the plurality of combination data to generate an integrated graph in which the name of the corresponding point and the name of the place are combined, and use the integrated graph to determine the name of the point corresponding to the name of the place The gist is to obtain and calculate the location of the place name using the importance of the location and the location information of the location.

  The place name position estimation method according to claim 3 is the place name position estimation method according to claim 1 or 2, wherein in the estimation step, the importance of the point is converted into a distance from the point, and the position of the place name is calculated. The gist is to calculate.

  The place name position estimation apparatus according to claim 4 stores a plurality of point data in which a name of a predetermined point and position information of the point are associated with each other, and reads the name of the point from the storage unit Extracting means for extracting document data including the name of the point from a plurality of document data, extracting a place name related to the point from the extracted document data, and determining a frequency at which the place name appears in the document data. An analyzing means for calculating, and an estimation means for calculating the position of the place name using the importance of the point and the position information of the point, using the calculated appearance frequency of the place name as the importance of the point with respect to the place name, It is summarized as having.

  The place name position estimation device according to claim 5 is the place name position estimation apparatus according to claim 4, wherein the analysis means associates the name of the point, the place name, and the appearance frequency of the place name. Output the data, the estimation means integrates the plurality of combination data to generate an integrated graph in which the name of the corresponding point and the name of the place are combined, and using the integrated graph, the name of the point corresponding to the name of the place The gist is to obtain and calculate the location of the place name using the importance of the location and the location information of the location.

  The place name position estimation apparatus according to claim 6 is the place name position estimation apparatus according to claim 4 or 5, wherein the estimation means converts the importance of the point into a distance from the point and calculates the position of the place name. The gist is to calculate.

  The place name position estimation program according to claim 7 causes a computer to execute the place name position estimation method according to any one of claims 1 to 3.

  According to the present invention, the position of a place name can be reliably estimated.

It is a figure which shows the functional block structure of a place name position estimation apparatus. It is a figure which shows the example of document data. It is a figure which shows the example of POI data. It is a figure which shows the operation | movement flow of a place name position estimation apparatus. It is a figure which shows the example of extraction results, such as a text document. It is a figure which shows the functional block structure of a document analysis part. It is a figure which shows the operation | movement flow of a document analysis part. It is a reference figure at the time of description of the operation | movement flow of a document analysis part. It is a figure which shows the example of output results, such as the frequency | count of appearance of a place name. It is a figure which shows the functional block structure of a place name position estimation part. It is a figure which shows the operation | movement flow of a place name position estimation part. It is a figure which shows the example of an integrated graph. It is a figure which shows the example of an estimation result of the position of a place name.

  The present invention pays attention to the fact that place names do not exist in some place names, and registered addresses that are different from the addresses recognized by the user, and information other than position information related to place names to be estimated (POI described later). The location of the place name is estimated using data etc.).

  Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings.

  First, the definition of the words used in this embodiment will be described.

  A “text document” is a plain text representation of a document described in a natural language such as Japanese. For example, a text extracted from a blog article published on the Internet.

  A “place name” is a place in the real world expressed in a natural language such as Japanese. For example, “Tokyo”, “Yokohama”, “Dobita Street”, “Tokyo Solamachi” and the like can be mentioned. The “location of the place name” represents the place name as coordinates that are one point. Generally expressed as a pair of latitude and longitude.

  “POI (Point of Interest)” refers to a point having some meaning in the real world. For example, shops such as restaurants and sightseeing spots, and landmarks such as “Yokohama Marine Tower” can be mentioned.

  Next, the function of the place name position estimation apparatus 1 according to the present embodiment will be described.

  FIG. 1 is a diagram showing a functional block configuration of the place name position estimation apparatus 1. The place name position estimation apparatus 1 includes a document data storage unit 11, a POI data storage unit 12, a POI related document extraction unit 13, a document analysis unit 14, a place name position estimation unit 15, and a place name position data storage unit 16. It is configured with.

  The document data storage unit 11 stores a plurality of text documents from which place names are extracted. These text documents are collected in advance and stored in the document data storage unit 11 in advance. An example of document data is shown in FIG.

  The POI data storage unit 12 stores a plurality of POI data (point data). The POI data stores at least a POI name and POI position information in association with each other. The position of the POI is represented as a pair of latitude and longitude, for example. These POI data are collected in advance and stored in the POI data storage unit 12 in advance. An example of POI data is shown in FIG.

  The POI related document extraction unit 13 extracts a text document including the name of the POI from a plurality of text documents using each text document and each POI data.

  The document analysis unit 14 extracts a place name related to the POI data from the text document extracted by the POI related document extraction unit 13 using the morphological analysis technique and the specific expression extraction technique, and the place name appears in the text document. The number of times (frequency of appearance of place names) is calculated.

  The place name position estimation unit 15 sets the number of appearances of the place name calculated by the document analysis unit 14 as the weight (importance) of the POI data for the place name, and uses the weight of each POI data and the position information of each POI. The position of is calculated.

  The place name position data storage unit 16 stores the position information of the place name calculated by the place name position estimation unit 15.

  The above is the function with which the place name position estimation apparatus 1 is provided. Note that some functions of the document data storage unit 11, the POI data storage unit 12, the POI-related document extraction unit 13, and the document analysis unit 14 are based on the technique disclosed in Non-Patent Document 1 described above. The difference from this disclosed technique is that, in addition to outputting the POI name in the document analysis unit 14, the number of appearances of the place name is calculated, and the place name position estimation unit 15 determines the position of the place name using the number of appearances of the place name. It is in the point which is trying to estimate.

  Next, the overall operation of the place name position estimation apparatus 1 will be described. FIG. 4 is a diagram illustrating an operation flow of the place name position estimation apparatus 1.

  First, in step S101, the POI related document extracting unit 13 reads out the names of the text documents and the POIs from the document data storage unit 11 and the POI data storage unit 12, respectively, and includes the POI names of the text documents. Extract text documents. At that time, a text document in which the POI name is described in slightly different notation due to the notation fluctuation may be extracted using the similar character string search technique disclosed in Non-Patent Document 2 described above.

  Then, the extracted text document, the POI name included in the text document, and the appearance position such as the number of digits and the number of lines where the POI name appears on the text document are output in association with each other. An example of the output result is shown in FIG.

  For example, if there is a text document that says “Yesterday, the weather was nice, so I went to Yokohama street and ate a cake from Yokohama cafe.” If there is POI data named “Yokohama Cafe,” this text document Extract and output “22-27th characters” as the appearance position of the POI name.

  Next, in step S102, the document analysis unit 14 extracts a place name related to the POI data from the text document using the text document extracted in step S101, the POI name, and the POI appearance position. Calculate the number of appearances of the place name. This process will be described in detail later.

  Finally, in step S103, the place name position estimation unit 15 uses the position information of each POI read from the POI data storage unit 12, and includes the POI name, the place name, and the number of appearances output in step S102. The position of each name is calculated from the set. Then, the place name and its position information as the calculation result are stored in the place name position data storage unit 16. This process will also be described in detail later.

  The above is the overall operation of the place name position estimation apparatus 1.

  Next, the operation of the document analysis unit 14 in step S102 described above will be described. As shown in FIG. 6, the document analysis unit 14 includes a document input unit 141, a morpheme analysis unit 142, a specific expression extraction unit 143, a place name extraction unit 144, and a place name output unit 145. FIG. 7 is a diagram illustrating an operation flow of the document analysis unit 14.

  First, in step S102-1, the document input unit 141 receives input of the text document extracted in step S101, the POI name, the POI name appearance position (see FIG. 5), and then the text document. The morpheme analysis unit 142 outputs the POI name and the POI notation to the place name extraction unit 144.

  The “POI notation” is a notation in which the POI name is actually described at the appearance position of the POI name in the text document. Since the POI related document extracting unit 13 can extract a text document even if the POI name is described with slightly different notation due to the notation as described above, the POI related document extracting unit 13 extracts the actual description as the POI notation here. For example, if the POI name is “Yokohama Cafe” and “Yokohama cafe” is actually described at the appearance position, “Yokohama cafe” is output as the notation of POI.

  Next, in step S102-2, the morpheme analyzer 142 converts the input text document into a sequence of sentences delimited by morphemes by performing morphological analysis. A “sentence string” refers to an ordered collection of one or more sentences. For example, when the text document shown in FIG. 8A is input, a sentence string as shown in FIG. 8C is output. Note that the morphological analysis technique is a known technique.

  Next, in step S102-3, the specific expression extraction unit 143 extracts specific expressions such as date and time from the text document (sentence column) after the morphological analysis, and sets the type of each specific expression to the extracted morphemes. A mark for identification is given. For example, when a sequence of sentences delimited by morphemes in FIG. 8C is input, a sequence of sentences with a date and time mark as shown in FIG. 8D is output. The specific expression extraction technique is also a known technique.

  Next, in step S102-4, the place name extraction unit 144 extracts only the sentence including the POI notation from the sentence column to which the unique expression mark is added. At that time, the sentence can be selected using conditions such as part of speech as in the technology disclosed in Non-Patent Document 1. After that, morpheme strings to which place marks are assigned are extracted from the marks of specific expressions, and these are extracted as place names. However, as in the technology disclosed in Non-Patent Document 1, if the place mark overlaps the POI notation, it is not extracted.

  For example, when a sentence including the POI notation shown in FIG. 8B is extracted from the sentence column to which the mark shown in FIG. 8D is assigned, the result is as shown in FIG. When a morpheme sequence to which a place mark is assigned is extracted from here, a place name as shown in FIG.

  Finally, in step S102-5, the place name output unit 145 outputs the combination data in which the POI name, the place name, and the number of appearances of the place name are associated with each other for the place name extracted in step S102-4. The output result is shown in FIG.

  The above is the operation of the document analysis unit 14. Note that steps S102-2 to S102-4 are executed once for each text document. When there are a plurality of text documents, it is repeatedly executed according to the number of the text documents, and a plurality of combination data is output.

  Subsequently, the operation of the place name position estimation unit 15 in step S103 described above will be described. As shown in FIG. 10, the place name position estimation unit 15 includes a POI / place name input unit 151, a POI / place name integration unit 152, a position calculation unit 153, and a position output unit 154. FIG. 11 is a diagram illustrating an operation flow of the place name position estimation unit 15.

  First, in step S103-1, the POI / place name input unit 151 inputs combination data (see FIG. 9) that associates the POI name, the place name, and the number of appearances of the place name output in step S102-5. After receiving the POI position information (see FIG. 3) from the POI data storage unit 12, the combination data is output to the POI / place name integration unit 152, and the POI position information is output to the position calculation unit 153.

  Next, in step S103-2, the POI / place name integration unit 152 graphs the correspondence between the POI name and the place name using the combination data. Specifically, POI names and place names that are the same are integrated, and each POI name and each place name are connected by a line, and the number of appearances of the place name is given on the line, so that multiple combinations One integrated graph in which data is integrated is generated. An example of the integrated graph is shown in FIG. Thus, by tracing a POI linked to a certain place name on the integrated graph, all POI data from which the place name is extracted and the number of appearances of the place name at that time can be checked.

  Next, in step S103-3, the position calculation unit 153 obtains the name of the POI corresponding to the place name to be estimated using the integrated graph, and uses the number of appearances of the place name as the weight of the POI data for the place name. Is converted into a distance from the POI position, and the position of the place name is calculated using the POI position information. Various calculation methods can be considered depending on the application destination of the present technology. For example, a value obtained by multiplying the coordinate value of the POI position by a weight may be averaged, or fitting with an appropriate probability distribution may be considered.

  Finally, in step S103-4, the position output unit 154 stores the location information (latitude / longitude values) of the place name calculated in step S103-3 in the place name position data storage unit 16. An image of the estimated position information is shown in FIG. The position of “Chinatown” is displayed at a position from the position of each POI calculated based on the weight of the POI data.

  The operation of the place name position estimation apparatus 1 has been described above. The place name position estimation apparatus 1 according to the present embodiment has been described above.

  As described above, according to the present embodiment, a plurality of POI data in which a POI name and POI position information are associated is stored, and document data including the POI name is extracted from the plurality of document data. The name of the place related to the POI data is extracted from the extracted document data, the number of times the place name appears in the document data is calculated, and the weight of the POI data is calculated using the calculated number of place names as the weight of the POI data for the place name. Since the position of the place name is calculated using the position information of the POI, the position of the place name can be reliably estimated.

  In other words, in order to convert the place name into one position in consideration of the POI data and its importance, information on which POI is associated with which place name and how important each relation is from the text document. By obtaining, it is possible to convert the position of the place name simply from the position of the POI, or to convert it to one position generally recalled than when using dictionary data prepared in advance. Can be converted.

  Finally, the place name position estimation device 1 described in the present embodiment can be realized by a computer having a memory and a CPU. Moreover, each operation | movement of the place name position estimation apparatus 1 can be constructed | assembled as a program, can be installed and executed in a computer, or can be distributed via a communication network.

DESCRIPTION OF SYMBOLS 1 ... Place name position estimation apparatus 11 ... Document data storage part 12 ... POI data storage part 13 ... POI related document extraction part 14 ... Document analysis part 141 ... Document input part 142 ... Morphological analysis part 143 ... Specific expression extraction part 144 ... Place name extraction Part 145 ... Place name output part 15 ... Place name position estimation part 151 ... POI / place name input part 152 ... POI / place name integration part 153 ... Position calculation part 154 ... Position output part 16 ... Place name position data storage part S101 to S103, S102-1 ~ S102-5, S103-1 to S103-4 ... step

Claims (7)

By computer
A storage step of storing a plurality of point data in which a name of a predetermined point and position information of the point are associated with each other in a storage unit;
An extraction step of reading the name of the point from the storage means and extracting document data including the name of the point from a plurality of document data;
An analysis step of extracting a place name related to the point from the extracted document data, and calculating a frequency at which the place name appears in the document data;
An estimation step of calculating the frequency of appearance of the place name as the importance of the point with respect to the place name, and calculating the position of the place name using the importance of the point and the position information of the point;
A place name position estimation method characterized by comprising: In the analysis step, a plurality of combination data in which the name of the point, the place name, and the appearance frequency of the place name are associated are output,
In the estimating step,
An integrated graph that combines the plurality of combination data and associates the name of the corresponding point and the name of the place is generated, the name of the point corresponding to the place name is acquired using the integrated graph, the importance of the point and the The place name position estimation method according to claim 1, wherein the position of the place name is calculated using position information of the point. In the estimating step,
The place name position estimation method according to claim 1 or 2, wherein the place name position is calculated by converting the importance of the place into a distance from the point. Storage means for storing a plurality of point data in which a name of a predetermined point is associated with position information of the point;
An extraction unit that reads the name of the point from the storage unit and extracts document data including the name of the point from a plurality of document data;
An analysis means for extracting a place name related to the point from the extracted document data, and calculating a frequency at which the place name appears in the document data;
Estimating means for calculating the frequency of appearance of the place name as the importance of the point with respect to the place name, and calculating the position of the place name using the importance of the point and the position information of the point;
A place name position estimation device characterized by comprising: The analysis means outputs a plurality of combination data that associates the name of the point, the place name, and the appearance frequency of the place name,
The estimation means includes
An integrated graph that combines the plurality of combination data and associates the name of the corresponding point and the name of the place is generated, the name of the point corresponding to the place name is acquired using the integrated graph, the importance of the point and the 5. The place name position estimation apparatus according to claim 4, wherein the place name position is calculated using the position information of the point. The estimation means includes
6. The place name position estimation apparatus according to claim 4, wherein the place name position is calculated by converting the importance of the place into a distance from the place.   A place name position estimation program that causes a computer to execute the place name position estimation method according to claim 1.

Images