JP4382634B2 - Address analysis apparatus, address analysis method, and address analysis program - Google Patents

Description

  The present invention relates to a technique for analyzing an input address character string and converting it into an address code and / or a standardized address character string.

  Search the database with the address that you have offered, or inquire at each address level (address level) in prefectures, cities, towns, towns, towns, and villages, etc. It is more efficient to enter the address that the customer proposes as it is, without entering it separately. However, in the address that the customer proposes, the capital letter or letter at the beginning of the place name is omitted, or the city and county or city and town have the same place name, and the city and county or town are mistaken. Even if an address table including an address code for each address level and a standardized place name (address string) was used, an accurate address code or a standardized address string could not be obtained. .

Therefore, conventionally, the address table is searched by deriving the swing of the notation, or the address derived from the swing of the notation is registered in the address table in advance (see Patent Document 1).
JP 2001-134602 A

  However, in the case of street names, because it consists of Tozai Street, Namboku Street, and street name annexes, Tozai Street and Namboku Street are expressed in reverse, or even though they are the same street (“Muromachi” A lot of fluctuations such as “Shinmachi” and “Shinmachi Muromachi”), aliases of street names (Kuromon-dori and Shin-shimachi-dori), street name annotations (east-in, east-in, east-in, east) Therefore, it is difficult to derive all combinations and register them in the address table.

  Further, when a street name ancillary word is not input, the street name ancillary word is ignored and the search cannot be performed. Therefore, there is a problem that even an address code or a standardized address character string candidate cannot be obtained.

  In the present invention, in order to solve the above-described problem, the street name is divided into East-West street, North-South street, and street name ancillary words, and the combination of fluctuations is reduced by using the street name code in which each coded and standardized data is registered. .

For example, assuming that there are two fluctuations on average in the name of Tozai-Dori and Namboku-dori, and that there are four fluctuations in the street name ancillary, and that there are two ways of reversing the names of the East-West and North-South streets, the derivation method is 1 There are 2 x 2 x 4 x 2 = 32 street names, and there are about 7,000 street names, so about 220,000 derived data are required. Tozai street names, north-south street names, street name annexes If it is divided into two, the total number of street names in the east and west and the streets in the north and south is about 200. Even if the fluctuation of the street name is derived, all street names can be searched in 400 ways.
[Action]
In order to encode and search East-West Street and Namboku Street separately, streets with many fluctuations, such as reverse input of Tozai Street and Namboku Street, alternative names of street names, alternative names of street names, different names of street names, etc. Names can also be searched for address codes and standardized address strings.

  Moreover, even if the street name ancillary word is omitted, the address code and the standardized address character string candidate can be shown. Furthermore, if there is a contradiction in the combination of street names such as East-West streets or North-South streets, it is possible to point out mistakes.

  By dividing the street name into East-West Street, Namboku Street, and street name ancillary words, it is possible to input a number of reverse names, street name alternatives, street name aliases, street name ancillary names, etc. It is possible to obtain an address code and a standardized address character string by correctly analyzing the address character string having fluctuations.

  Moreover, even if the street name ancillary word is omitted, the address code and the standardized address character string candidate can be shown.

  Furthermore, if there is a contradiction in the combination of street names such as East-West streets or North-South streets, it is possible to point out mistakes. In addition, if only one street name is entered, even if it is not one-way street, it is possible to point out the lack of input information.

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

  FIG. 1 shows an example of an embodiment of an address analyzing apparatus according to the present invention. In the figure, 1 is a word dictionary storage unit, 2 is an address table storage unit, 3 is a street name code table storage unit, and 4 is a street name. A table storage unit, 5 is a word division unit, 6 is an address candidate storage unit, 7 is an address candidate extraction unit, and 8 is a narrowing unit.

  As shown in FIG. 2, the word dictionary storage unit 1 divides a large number of words constituting an address character string into part-of-speech such as nouns, numbers, and suffixes, and place names, street names, street names, and other word meanings. A registered word dictionary is stored. Each word having the meaning of the place name in the word dictionary storage unit 1 further indicates a prefecture where the word may exist as an address, as shown in FIG. It is assumed that a prefecture flag represented for each address level of the town large character, character / chome, that is, a city level prefecture flag, a town level prefecture flag, and a character level prefecture flag are stored together (however, omitted in FIG. 2).

  As shown in FIG. 3, the address table storage unit 2 stores an address table in which each name in a word constituting an address character string is registered together with a suffix, an address level, and an address code. The address code is a code stipulated by the Geospatial Information Authority of Japan, and the inclusion relationship is understood on the code. From the left, the prefecture code, the ordinance-designated city / county / city / ward code, and the general town / village code It is.

  As shown in FIG. 4, the street name code table storage unit 3 stores the street name and street name of each east-west street in the words constituting the address character string, A street name code table registered together with street characteristics is stored. Here, as street name codes, codes of 9000 series on East-West street and 8000 series on Namboku street are attached.

  As shown in FIG. 5, the street name table storage unit 4 registers an address code and a standardized address character string (place name) corresponding to the combination of the east-west street name code, the north-south street name code, and the street name ancillary words. As you can see, the name table is stored.

  The word dividing unit 5 converts an address character string (including a street name) directly input from a keyboard or the like (not shown) or read from a storage medium or input from another device or the like via a communication medium to a word dictionary. Prefectures where the storage unit 1 is used to divide into words, find the prefecture code that is most inclusive at least at the prefecture level corresponding to the divided words, and each of the divided words may exist as an address Are output together with the meaning of words such as street flags and street name appendices, including prefecture flags representing city levels, that is, city level prefecture flags, town level prefecture flags, character level prefecture flags, prefecture flags including the whole.

  The address candidate extraction unit 6 includes a city level address candidate extraction unit 61, a town level address candidate extraction unit 62, a letter / chome level address candidate extraction unit 63, and a street name level address candidate extraction unit 64.

  Among them, the city level address candidate extraction unit 61, the town level address candidate extraction unit 62, and the character / chome level address candidate extraction unit 63 use a word output from the word dividing means 5 or a compound word obtained by combining a plurality of words. , Based on the prefecture flag, select a word corresponding to each address level of the prefecture, municipality, town, large letter, and letter / chome, and search the address table storage unit 2 with the selected word for each address level. Address candidate including the address code and the standardized address character string, and when the address candidate is not stored in the address candidate storage unit 7, the address candidate is stored in the address candidate storage unit 7, and the address candidate storage unit If an address candidate is already stored in 7, an address candidate having an inclusion relationship with the stored address candidate is stored.

  The street name level address candidate extraction unit 64 searches the street name code table storage unit 3 for a word that has the meaning of the street name in the words output from the word dividing unit 5 and responds. Get street name code, check that it is a combination of East-West street and Namboku street or one-way street, and store street name table with street name code, street name code (or street name code) and street name appendix Part 4 is searched for address candidates including standardized address strings and address codes at the street name level, and the address candidate storage unit 7 has the prefecture, municipality level, town large character level, and character / chome level. When the address candidate is not stored, the address candidate at the name level is stored in the address candidate storage unit 7, and when the address candidate is already stored in the address candidate storage unit 7, the stored address candidate is stored. Storing an address candidate of the nickname level with inclusion relation between.

  The address candidate storage unit 7 temporarily stores the address candidate obtained by the address candidate extraction unit 6.

  The narrowing-down unit 8 narrows down the address candidates stored in the address candidate storage unit 7 by combining the address level, the number of unused words, the number of suffix matches, and the distinction between formal and different notation.

  FIG. 6 shows the flow of processing in the address analyzing apparatus of the present invention, and the operation will be described in detail below according to this.

  The input address character string is divided into words by the word dividing unit 5 using the word dictionary storage unit 1, and each divided word indicates a prefecture in which each word may exist as an address for each address hierarchy. A city level prefecture flag, a town level prefecture flag, a character level prefecture flag, a prefecture flag including the whole, and the meaning of a word are output (s1).

  For example, when an address character string “Kyoto-shi Kita-ku Kuramaguchi Todera-cho Higashi-irijozenzenmonzen-cho” is input, it is output as shown in FIG. Since “Kuramaguchi” may be a street name or a street name, both the town level prefecture flag and the name as the meaning of the word are output. In Kyoto Prefecture, “Teramachi” is the street name only, so the town level prefecture flag is not output.

  Next, in the city level address candidate extraction unit 61, the town level address candidate extraction unit 62, and the character / chome level address candidate extraction unit 63 of the address candidate extraction unit 6, an address table storage unit based on the word division result 2 is used to extract address candidates at the city level, town level, and character level.

  First, in the city level address candidate extraction unit 61, “Kyoto” and “Kyoto City” are searched from “Kyoto”, and “Kita Ward” other than Kyoto is also searched from “North”. As a result of the inclusion relation check, “Kyoto Prefecture / Kita Ward” and “Kyoto City / Kita Ward” are stored in the address candidate storage unit 7 (s2). Here, since these two address candidates represent the same address, “Kyoto City / Kita Ward” with the same suffix is left and “Kyoto Prefecture / Kita Ward” is deleted.

  Next, in the town level address candidate extraction unit 62, “Kyoto City / Kita Ward / Kurumaguchi Town” is searched from “Kuramaguchi”, and “Kyoto City / Kita Ward / Jozenji Monzen Town” is searched from “Jozenji Monzen-cho”. These are stored in the address candidate storage unit 7 (s3). FIG. 8 shows the stored contents of the address candidate storage unit 7 at this time.

  Next, in the character / chome level address candidate extraction unit 63, character / chome level address candidate extraction (s4) is performed. However, since there is no corresponding word, nothing is performed.

  Next, the street name level address candidate extraction unit 64 performs street name level address candidate extraction (s5) using the street name code table storage unit 3 and the street name table storage unit 4, as shown in FIG. Hereinafter, the description will be made along the flow of the name level address candidate extraction process as shown in FIG.

  First, the street name code table storage unit 3 is searched to obtain the street name code “9001” from the first street name “Kuramaguchi”. Next, the street name code “8001” is obtained from the second street name “Teramachi”. For the street name ancillary word, the street name ancillary word “Toiri” (which may be coded) normalized by the word dictionary storage unit 1 is obtained (s5-1).

  After obtaining the street name code, check that it is a combination or one-way street of East-West Street and North-South Street. In the case of East-West streets (for example, “Kurumaguchi-dori Nijo”) and North-south streets (for example, “Kamogawasuji-cho”), the street name table storage unit 4 is not searched, and the processing is terminated because there is an input error. Also, if there is only one of East-West Street or North-South Street even though it is not one-way street, even if the street name table storage unit 4 is searched, many address candidates are returned and the solution cannot be narrowed down. Is finished (s5-2).

  Next, the street name table storage unit 4 is searched for the north-south street name code “8001”, the east-west street name code “9001”, and the street name ancillary word “Toiri”, and the address code “26/101/179/901”. , “26/101/253/901”, “26/101/309/901”, but only “26/101/253/901” address candidates that contain “Jozenji Monzen-cho” and address candidate storage unit 7 (s5-3, s5-4, s5-5).

  Finally, the narrowing-down process (s6) is performed in the narrowing-down unit 7. That is, “Teramachi”, “Toiri”, and “Jozenji Monzen-cho” are unused words in address candidates analyzed with “Kuramaguchi” as the town name, and there are no unused words in address candidates analyzed with “Kuramaguchi” as the street name. Therefore, the former address candidate having an unused word is rejected. Eventually, "Kenma-cho, Kita-ku, Kyoto, Higashi-Irikamizenzenji-monzen-mae (26/101/253/901)" is output as a candidate address solution.

  Here, the street name code table storage unit 3 and the street name table storage unit 4 can be searched even if the East-West street and the Namboku street are reversed and input, such as “Tamamachi-dori Kuramaguchi Higashi-Irigami-zenzen-mon-mae-cho, Kyoto-shi”. Are irrelevant to the input order, the street name code table storage section 3 is searched for the street name code “9001” for “Kuramaguchi”, the street name code “8001” for “Teramachi”, and the street name appendix “Toiri”. Since the name table storage unit 4 is searched, the correct solution of “Higashi-Irizenzenzenmonzen-mae (26/101/253/901)”, Kuramaguchi-dori-ji, Kita-ku, Kyoto-shi, Kyoto, is obtained.

  Also, there are streets that are used for both East-West Street and North-South Street, so such streets are given a code for both streets, and whether the street is East-West Street or North-South Street is determined by the other street name. Rokugen-cho is generally on Tozai-dori Street, but it is a street name that is sometimes used on Namboku-dori, which is different from Tozai-dori. Here, in the case of “Ichijo-dori Rokukencho Higashi-in”, “Ichijo” is Tozai-dori (street code “9002”), so “Rokuken-cho” on both streets is determined to be Namboku-dori (street name code “8801”). The street name table storage unit 4 is searched with the street name code and the street name attached word “Toiri”, and the address code “26/102/109/902” is obtained.

  On the other hand, in the case of “going down Kamogawa-suji Rokugencho”, “Kamogawa-suji” is Namboku-dori (street name code “8002”), so “Rokugen-cho” is judged as East-West Street (street name code “8801”). The street name table storage unit 4 is searched for the street name code and the street name attached word “down” to obtain the address code “26/105/119/903”.

  Next, the names will be described as different notations.

  The name of the street is the same as “Muromachi Shinmachi” and “Shinmachi Muromachi”, but the expression is different, or “Kuromon Dori” and “Shinshi Machi Dori”. The same name code is registered in the code table storage unit 3. Thus, a desired address code can be obtained without changing the search of the street name table storage unit 4.

  For example, in the case of “Going down Sanjo between Muromachi Shinmachi” and “Going down Sanjo between Shinmachi Muromachi”, the street name table storage unit 4 is stored with street code “8003”, “9003” and street name ancillary “down”. The address code “26/104/144/905” is obtained by searching. In addition, in the case of “Kuromon Dori Sanjo Gou” and “Shin Shimachi Dori Sanjo Gou”, the street name table storage unit 4 is searched with the street name codes “8004” and “9003” and the street name ancillary “up”. The address code “26/104/065/901” is obtained.

  In the case of one-way street, even if there are no two street names, the street name table storage unit 4 is searched to obtain an address code. For example, in the case of “Roppa Ranan Dori Nishiiri”, the street name is only “Rokuba Ranan”, but it is understood that it is a one-way street, so the street name code “9501” and the street name appendix “Nishiiri” The name table storage unit 4 is searched to obtain the address code “26/105/011/903”.

  When the street name ancillary word is not input, the street name table storage unit 4 is searched with the east-west street code and the north-south street code to obtain address candidates. For example, when “Kuramaguchi Toderacho Kuramaguchicho” is input, there is no street name ancillary word. Therefore, the street name table storage unit 4 is searched with the east-west street name code “8001” and the north-south street name code “9001”. "Tsurujicho Higashi Iri", "Tsurumaguchi Tsurujicho Higashi Irigami", "Teramachi Dori Tsurumaguchi Ryu" are obtained, but "Kuramaguchi Tōjō Town Higashi Iri", which includes "Kuramaguchi Town", is selected. If no town name is input, such as “Kuramaguchi-dori-cho”, all three candidates are output.

  The present invention can also be realized by installing a program having the procedures shown in the flowcharts of FIGS. 6 and 9 via a medium or a communication line in a known computer.

The block diagram which shows an example of embodiment of the address analysis apparatus of this invention Explanatory drawing which shows an example of a word dictionary Explanatory drawing which shows an example of an address table Explanatory drawing which shows an example of a street name code table Explanatory drawing which shows an example of a street name table Flow chart of processing in address analysis apparatus of the present invention Explanatory drawing which shows an example of the result of word division Explanatory drawing which shows an example of the memory content of the address candidate memory | storage part after town level address candidate extraction Flow chart of street name level address candidate extraction processing

Explanation of symbols

  1: word dictionary storage unit, 2: address table storage unit, 3: street name code table storage unit, 4: street name table storage unit, 5: word division unit, 6: address candidate extraction unit, 7: address candidate storage unit , 8: Refinement unit, 61: City level address candidate extraction unit, 62: Town level address candidate extraction unit, 63: Character / chome level address candidate extraction unit, 64: Street name level address candidate extraction unit, s1: Word division Processing, s2: city level address candidate extraction processing, s3: town level address candidate extraction processing, s4: letter / chome level address candidate extraction processing, s5: street name level address candidate extraction processing, s6: narrowing processing, s5-1 : Street name code table search, s5-2: street name direction check, s5-3: street name table search, s5-4: inclusion check, s5-5: address candidate storage.

Claims (3)

In the address analysis device that analyzes the input address string and converts it into an address code and / or a standardized address string,
The input address string is divided into words using a word dictionary, and each divided word is assigned to the prefecture where the word exists as an address, the municipality of the prefecture, the town square, and the address level Word division means for outputting together with the prefecture flag and the meaning of each word ,
A word corresponding to each address level is selected from the words output from the word dividing means based on the prefecture flag, an address table is searched with the selected word, and an address code and a standard for each address level are standardized. An address candidate including an address character string is obtained, stored in the address candidate storage unit, and if there is a word meaning a street name in the word output from the word dividing means, the street name code table is searched with the word. The corresponding street name code is obtained, and the street name table is searched with the street name code of the east-west street, the street name code of the north-south street, and street name ancillary words. If the address candidate is not stored in the address candidate storage unit, the street name address candidate is stored in the address candidate storage unit, and the address candidate is already stored in the address candidate storage unit. And address candidate extracting means for storing the candidate address in the nickname level with inclusion relation between candidate address that is the storage if it is 憶,
Addressing means comprising a narrowing means for narrowing down address candidates stored in the address candidate storage unit by combining the address level, the number of unused words, the number of suffix matches, and the distinction between formal notation and different notation Analysis device. In the address analysis method that analyzes the input address string and converts it to an address code and / or standardized address string,
Using a word dictionary, an address table, a street name code table, a street name table, a word dividing unit, an address candidate extracting unit, an address candidate storage unit, and a narrowing unit,
The word dividing means divides the input address character string into words using a word dictionary, and each divided word is assigned to the prefecture where the word is present as an address, a municipality, a town large character, a character A word division step for outputting together with the prefecture flag represented for each street address level and the meaning of each word ;
The address candidate extracting means selects a word corresponding to each address level from the words output from the word dividing means based on the prefecture flag, searches the address table with the selected word, and addresses for each address level. An address candidate including a code and a standardized address character string is obtained, stored in the address candidate storage unit, and if there is a word meaning a street name in the word output from the word dividing means, the word Search the name code table to find the corresponding street name code, search the street name table by East-West street name code, Namboku street name code and street name ancillary words, street name level address code and standardized address characters Address candidates including a column are obtained, and if the address candidate at the address level is not stored in the address candidate storage unit, the address candidate at the name level is stored in the address candidate storage unit as described above, And address candidate extracting step if the candidate address has already been stored for storing an address candidate of the nickname level with inclusion relation between candidate address that is the storage in section,
The narrowing-down means includes a narrowing-down step for narrowing down address candidates stored in the address candidate storage unit by combining the address level, the number of unused words, the number of suffix matches, and the distinction between formal and different notation. Address analysis method. An address analysis program that analyzes an input address string and converts it into an address code and / or a standardized address string,
In a computer having a word dictionary, an address table, a street name code table, a street name table, and an address candidate storage unit,
The input address string is divided into words using a word dictionary, and each divided word is assigned to the prefecture where the word exists as an address, the municipality of the prefecture, the town square, and the address level A word division process for outputting together with the prefecture flag and the meaning of each word ,
The word corresponding to each address level is selected from the divided words based on the prefecture flag, the address table is searched with the selected word, and the address code and the standardized address character string for each address level are obtained. An address candidate including the address candidate is stored in the address candidate storage unit, and if there is a word meaning a street name among the divided words, a street name code table is searched for the word to obtain a corresponding street name code. The street name table is searched with the street name code of the east-west street, the street name code of the north-south street, and a street name ancillary word to obtain an address candidate including a street name level address code and a standardized address character string. If no address candidate at the address level is stored, the address candidate at the name level is stored in the address candidate storage unit, and if the address candidate is already stored in the address candidate storage unit, it is stored. And address candidate extracting process of storing an address candidate of the nickname level candidate address as a containment relationship are,
An address analysis program for executing a narrowing process for narrowing down address candidates stored in an address candidate storage unit by combining the address level, the number of unused words, the number of suffix matches, and the distinction between formal and different notation.

Images