CN101393544A - Chinese Address Semantic Analysis Method Oriented to Address Coding - Google Patents

Abstract

The invention discloses an address coding-oriented Chinese address semantic analysis method, the steps of which are as follows: Step 1: construct an address feature font library according to sample data; a. establish sample data; b. filter feature words; c. filter subsidiary features Characters; the selected characteristic words and subsidiary characteristic words constitute the characteristic character library; the second step: according to the characteristic character library, according to the address representation rules, convert the Chinese address into a string in digital form; the third step: build the address analysis rule library; Step 4: Semantic analysis, including: address representation, converting the original address into a digital representation, address resolution, splitting the address represented by the number into address elements; restoring the address, restoring the parsing result of the digital representation to the corresponding to the original address string.

Description

Chinese Address Semantic Analysis Method Oriented to Address Coding

technical field

The present invention proposes a Chinese address semantic analysis method that does not rely on a gazetteer, and is suitable for geographic information systems (GIS) in the fields of real estate management, land management, urban planning, public security, postal services, taxation, telecommunications, public health, and number know-how address encoding.

Background technique

In daily production and life, addresses are one of the most commonly used reference systems for describing geographic locations using natural language. Address description is the most commonly used means to describe the spatial location in various business systems. Using address coding technology can convert a large amount of address positioning information that already exists in the management information system (MIS) into one that can be used in geographic information systems ( The geographic coordinate method of GIS) enables GIS to integrate, store, retrieve, operate, and analyze geographic data, link the data scattered in various departments through the spatial reference system, and provide information for land use, resource management, environmental monitoring, and transportation. Transportation, urban planning, etc. provide decision-making services, thereby greatly promoting the application of GIS technology.

Address coding refers to the process of intelligently analyzing the address information described in natural language according to the address model and coding rules, and establishing the association with the corresponding spatial coordinate information and geographic coding by matching with the database. The basic principle is shown in Figure 1. Show.

Address coding needs to solve the following three key technical problems: ①Address semantic analysis: refers to splitting the address described in natural language into address elements within a certain limited area that can specify a specific geographical range. For example, "No. 122, Ninghai Road, Gulou District, Nanjing City" is resolved into four address elements of "Nanjing City", "Gulou District", "Ninghai Road" and "No. 122". Each address element in the address is arranged according to the relationship from large to small, and the following address elements must be meaningful relative to the previous address elements. ②Address model: The address model is used to describe the composition rules of address elements in various types of addresses; ③Address matching: refers to matching the address analyzed by computer semantics with the standard address in GIS according to the established address model and coding rules. And give the intelligent process of geographical coordinate value.

Since the 1970s, the United States has begun to establish a national geographic coding system, the "Dual Independent Map Encoding System" (Dual Independent Map Encoding, DIME). The development of DIME is a milestone in the development history of GIS technology. In the late 1980s, the US Census Bureau developed the DIME system into a topologically integrated address coding reference system (Topologically Integrated Geographic Encoding and Referencing, TIGER). Due to its wide coverage, good accuracy, guaranteed update, and low cost, the TIGER database has become a recognized address coding reference standard in the United States. At present, foreign geocoding databases and geocoding software tools have been commercialized, and there are many content standards and specifications for address data, such as the public draft of FGDC address data content standards. Most foreign GIS software have address coding functions, such as MapInfo's MapMaker, ArcGIS's Geocoding and GeoMedia's Geocodes Addresses, which have high response speed and accuracy in specific applications. The key factors for the successful popularization and application of foreign address coding technology are: first, from the perspective of language, there are space separators between words in the address description of English and other western languages; from the perspective of standardization, the data naming and expression of addresses, as well as software development and Application services follow the principles of standardization and normalization. Therefore, address semantic analysis in foreign address coding adopts the method of simple character matching with place names in the standard address database (collectively referred to as "dictionary matching method").

Specifically, the dictionary matching method uses the place-name data in GIS as a place-name dictionary, uses the string matching algorithm for lexical and grammatical analysis, and matches the address string with the place-name in the dictionary to achieve the purpose of address resolution. The dictionary matching method can only analyze the place names in the address string that are exactly the same as those in the dictionary. For example, assuming that the dictionary contains the three place names of "Nanjing City", "Gulou District" and "Ninghai Road" but not "Wenyuan Road", the address "No. 122, Ninghai Road, Gulou District, Nanjing City" can be resolved into "Nanjing City ", "Gulou District", "Ninghai Road" and "No. 122" are four address elements, and "Nanjing City Wenyuan Road No. 12" is resolved into "Nanjing City" and "Wenyuan Road No. 12". Therefore, the effect of the dictionary matching method is directly proportional to the dictionary size and update speed, but the efficiency is inversely proportional to the dictionary size. However, due to the historical and cultural characteristics of the Chinese language itself, as well as the serious lack of address coding standards, there are serious irregularities in Chinese addresses, not only the regularity of address naming is poor, the format is complex, and there are multiple names in one place. , and adding, deleting, and modifying are more random. Obviously, the dictionary matching method cannot meet the needs of large-scale data processing in terms of scope of application, update maintenance, accuracy and response speed. Semantic analysis of Chinese addresses has become a key problem to be solved in current Chinese address coding technology (Huang Song. Research on Chinese Address Coding Technology [D]. Peking University: Master's Degree Thesis, 2005).

Since the 1980s, domestic address coding technology research has focused on address standard libraries and address matching algorithms. For example, Beijing and Shanghai have promulgated a series of coding standards for urban roads and road intersections (Zhu Jianwei, Wang Zemin. Geography Coding Principles and Localization Solutions [J]. Beijing Surveying and Mapping, 2004(2), 24-27.). The semantic analysis of addresses all adopts the dictionary matching method, such as the "Addressing God" of Beijing Changdi Computer Company, the "Map Searcher" of Founder of Peking University, the "Beijing Address Code Database System and Standard Address Matching Engine" of Zhaoxi Technology, and the Beijing SuperMap Company The "customer relationship management system" of Shanhai Yihui, EzGeoCoding and other systems.

Contents of the invention

The technical problem to be solved by the present invention is to overcome the defects in the prior art and provide an address coding-oriented Chinese address semantic analysis method without relying on a gazetteer for address analysis.

The present invention is oriented to the Chinese address semantic analysis method of address coding, and its specific technical process is as follows:

The first step: according to the sample data, construct the address feature font library

The Chinese address includes four types of address elements: administrative division, street, gate number and supplementary information, which are arranged from large to small according to the geographical area of the address elements (the last is supplementary information) to form a Chinese address string; the composition of the Chinese address string will Some types of address elements are missing;

The above-mentioned administrative divisions are divided into four levels according to the <Administrative Area Code of the People's Republic of China> (GB 2260-1995), and the administrative areas above the village are sorted from large to small (there will be missing): the first level is provinces, autonomous regions, Municipalities directly under the central government and special administrative regions; the second level is cities, districts, autonomous prefectures, leagues, and municipal districts and counties directly under the central government; the third level is counties, city districts, county-level cities, and banners; the fourth level is townships, towns, and villages;

Generally speaking, an address often contains the names of multiple administrative divisions of different levels. For example, "No. 122, Ninghai Road, Gulou District, Nanjing City" includes the names of two different levels of administrative divisions, "Nanjing City" (second level) and "Gulou District" (third level).

said street is a road name and/or street name;

The gate number refers to house number, building number, building name and/or room number;

The supplementary information refers to the name of the institution or the vocabulary (east, west, south, north, etc.) "Market" is the name of an institution, and "West" in "Xige Street West, Yongning Town, Jiangpu County, Nanjing City" is a vocabulary that expresses the relationship of spatial direction.

A Chinese address string can be split into multiple address elements of different types; the address element is a combination of ordinary characters + characteristic words (except for supplementary information); among them

The characteristic characters of administrative divisions are: provinces, autonomous regions, municipalities directly under the central government, special administrative regions, cities, regions, autonomous prefectures, leagues, districts, counties, banners, townships, towns, villages, villages, villages, etc.;

The characteristic characters of streets are: road, street, street, avenue, avenue, road, li, alley, alley, alley, strip, etc.;

The characteristic characters of the gate number are: number, building, dormitory, fasting, hall, hall, etc.;

Building the address feature font library includes the following steps:

1. Establish sample data: Separate each address element in the original address data to form sample data.

Table 1 Raw data sample

Raw data No. 87, Houzaimen West Village, Xuanwu District, Nanjing No. 2 Shilin Road, Zhuzhen Town, Liuhe District Shangen Formation, Naner Village, Qiaolin Town, Pukou District, Nanjing Longnan Village, Longshan Township, Jiangpu County, Nanjing City No. 3, Xiyun Village, Yuhuatai District, Nanjing No. 037-054, B1 Floor, Huanbei Market, Hangzhou, Nanjing No. 153, Wengjiaying Village, Wengjiaying Village, Honghua Street, Qinhuai District

Table 2 sample data sample

2. Filter feature words: feature words indicate the end of an address element, which can be regarded as the unit of address elements; in most cases, addresses can be more accurately divided into independent semantic units according to feature words. Feature word screening process: count the frequency of the last character and the two characters of all address elements in the sample data respectively, and sort them in descending order; filter the single character whose cumulative frequency accounts for more than 80% of the percentage as feature words ( Referred to as "single characteristic word"); Accumulated frequency accounts for more than 80% two characters (must last character is not single characteristic word) is screened as characteristic word (referred to as " multiple characteristic word ");

3. Screening of subsidiary feature words: Chinese addresses usually contain some vocabulary expressing spatial relations, such as east, south, west, north, etc., which can be used to assist in judging the split position of address elements, and filter these words into subsidiary feature words;

The selected characteristic words and subsidiary characteristic words constitute the characteristic word library;

Step 2: Convert the Chinese address into a string in digital form according to the characteristic font library and according to the address representation rules;

In order to facilitate computer processing, it is necessary to convert the Chinese address string into a digital representation, where 1 represents a characteristic character, 2 represents a subsidiary characteristic character, 3 represents the last character of two consecutive repeated characteristic characters, 0 represents a common character, and 9 represents a terminator . Ordinary characters are meaningless for the formulation of splitting rules, and consecutive 0 characters can be compressed into one 0 character. For example, "No. 32, Xiaolin, Yedong Village, Babaiqiao Town, Liuhe County, Jiangsu Province" is expressed as "01010110212019", and "The Village Department of Yexu Village, Yingtian Road, Jianye District" is expressed as "0101011309".

Step 3: Build an address resolution rule library

After the Chinese address is converted into a numeric string, its composition follows the following rules:

● "0" can only be a number among "1", "2" and "9";

● "1" can only be followed by one of "0", "1", "2", "3" and "9";

● "2" can only be one of "0", "1", "2" and "9";

● "3" can only be one of "0", "1", "2" and "9";

●It can only start with one of "0", "1", "2" and "3";

●It can only end with "9".

According to the above rules, the address can be expressed as a tree structure, and each path represents a resolution rule. The first-level nodes of the tree are "0", "1", "2", and "3", and their descendant nodes are organized according to the above rules. However, when each path reaches a certain length, the split point of the address element can be determined, thereby terminating the further expansion of the path. At the same time, each parsing rule must specify a specific split position, which is represented by "f+ split position". Taking the address in the sample data as an example, the application frequency of the rules is counted, and the analysis paths with a cumulative frequency of more than 95% are selected as analysis rules (as shown in Table 3). The tree structure of parsing rules is shown in Figure 3.

Table 3 address resolution rules

serial number parsing rules serial number parsing rules 1. 010f2 93. 0202019f6 2. 019f2 94. 02110101f4 3. 2010f3 95. 011122101f3 4. 0110f3 96. 112f2 5. 09f1 97. 020210101f5 6. 0210f3 98. 309f2 7. 012f2 99. 02021019f7

8. 0120f2 100. 012101021f2 9. 0112f3 101. 2020109f2 10. 0119f3 102. 21119f4 11. 2110f3 103. 012101011f2 12. 21010f2 104. 202010101f2 13. 21019f2 105. 211111f3 14. 21210f4 106. 21112f3 15. 2012f3 107. 119f2 16. 2019f3 108. 3101f2 17. 20210f4 109. 202021f2 18. 01110f4 110. 012101010109f2 19. 10f2 111. 02221f1 20. 0219f3 112. 21211019f5 twenty one. 02010f4 113. 11109f4 twenty two. 201101f4 114. 2209f3 twenty three. 219f2 115. 0111219f3 twenty four. 0212f3 116. 0202109f6 25. 02210f4 117. 202020101f2 26. 012101019f2 118. 2120f2 27. 11101f3 119. 2013f3 28. 011110f2 120. 1211f1 29. 2210f3 121. 0121010101010f4 30. 2109f3 122. 02211f5 31. 02019f4 123. 210111f2 32. 2119f3 124. 2111101f2 33. 209f2 125. 020119f5 34. 2112f3 126. 020209f5 35. 013f2 127. 021119f3 36. 201102f4 128. 202120f4 37. 301f3 129. 020219f5 38. 2219f3 130. 20111019f3 39. 1210f3 131. 1212f3 40. 0209f3 132. 2021109f4 41. 2201f4 133. 129f2 42. 201109f3 134. 201121019f3 43. 2221f4 135. 2021101f2 44. 21021f2 136. 020210109f8 45. 01210101019f4 137. 02110109f3 46. 21219f4 138. 22119f4 47. 19f1 139. 0113f3 48. 2101101f2 140. 02111019f3 49. 21012f2 141. 2102019f2 50. 29f1 142. 213f2 51. 110f2 143. 20221f5

52. 0202010f6 144. 0221201f4 53. 20119f4 145. 212112f4 54. 01119f4 146. 229f2 55. 01112101f3 147. 021120f4 56. 02012f4 148. 212119f2 57. 20201019f2 149. 01112019f3 58. 1119f3 150. 0202210f6 59. 021109f3 151. 2011109f4 60. 20212101f2 152. 21020109f2 61. 211101f4 153. 011129f3 62. 022010f5 154. 021102f3 63. 2122f2 155. 3109f3 64. 2212f3 156. 021101101f4 65. 0211019f3 157. 022019f5 66. 21212f4 158. 2121109f4 67. 1112f3 159. 021110101f5 68. 0121010109f2 160. 021111f3 69. 22110f4 161. 02209f4 70. 20219f4 162. 210201019f6 71. 02119f4 163. 2102210f2 72. 11102f3 164. 01112010f4 73. 029f2 165. 11110f4 74. 2202f5 166. 20201109f5 75. 202010109f5 167. 211119f2 76. 011119f2 168. 3219f3 77. 120f1 169. 39f1 78. 3210f3 170. 0111211f2 79. 210119f2 171. 201119f3 80. 011111f3 172. 3021f4 81. 211109f5 173. 0202021f7 82. 2101109f4 174. 021110109f3 83. 02219f4 175. 0211109f3 84. 0201101f5 176. 20201011f2 85. 0201109f6 177. 20212109f2 86. 011112f2 178. 210201011f6 87. 0121010101019f2 179. 210209f2 88. 2011201f4 180. 0121010201010f4 89. 1219f3 181. 201111019f3 90. 12210f4 182. 202011019f2 91. 202019f5 183. 2101102f2 92. 20209f4 184. 3221f4

In FIG. 3 , the rule "0120f2" indicates that the nodes on each layer are "0", "1", "2", and "0", respectively. When scanning the address string, if there is a string sequence matching "0120", it can be determined to split after the second number, and "f2" is used to indicate the split position. For example, "Nantai Lane, Baixia District" is represented as "01201", which matches the rule "0120", that is, splits after "01" on the left, and parses it into "Baixia District" and Nantai Lane. The rule "029f2" indicates that when a part of the character sequence in the address string matches "029", it can be determined to split after the second number. Since "9" means the end of the digit string, the rule also means to split at the end.

Step 4: Semantic Analysis

On the basis of formulating characteristic words and analysis rules, the present invention designs a Chinese address analysis algorithm (referred to as "RBAI algorithm"). The algorithm includes three parts: address representation, address resolution and address restoration. The specific analysis process is as follows:

Input: a piece of original address data, represented by Address_Before;

Output: The parsing result of the original address, represented by Address_After.

(1) Address representation: convert the original address into a digital representation, and the result is Numbers_Before.

Step 1: Set Numbers_Before to an empty string; use n to represent the length of the current original address to be parsed

Step 2: i is from 1 to n, execute in a loop:

If the i-th character of the original address is the main feature word, then Numbers_Before[i] represents 1;

If the i+1th character of the original address is the main feature word, then Numbers_Before[i] is represented as 3;

If the i-th character of the original address is an auxiliary character, then Numbers_Before[i] is represented as 2;

If the i-th character of the original address is an ordinary character, then Numbers_Before[i] represents 0;

i is assigned the value i+1;

end loop;

Step 3: Add 9 at the end of Numbers_Before; compress consecutive multiple 0s in Numbers_Before into one 0.

(2) Address analysis: Split Numbers_Before into address elements according to the analysis rules, and the result is Numbers_After.

Step 4: Set Numbers_After to an empty string; use k to represent the length of Numbers_before

Step 5: m from 1 to k, execute the loop:

If the m characters on the left of Numbers_Before match a certain parsing rule, the left and right substrings of Numbers_before will be split according to the rules; the left substring Numbers_Left will be saved as an address element in the parsing result and will not be split; the right substring Numbers -right to continue splitting; define Numbers_Before as Numbers-right;

m is assigned the length of Numbers_Left + 1;

otherwise

m assignment is m+1;

end loop;

(3) Address Restoration: Step 6: restore the parsing result represented by the number to a character string corresponding to the original address, and the result is Address_After.

Compared with the address resolution technology adopted in the existing Chinese address coding, the present invention mainly has the following advantages:

a. Not dependent on dictionaries: avoid dictionary construction and updating, and can resolve address element names that are not included in GIS (or dictionaries);

b. Do not rely on natural language processing technologies such as Chinese word segmentation;

c. High efficiency: Since no word-based string matching algorithm is used, but a single address is operated, the efficiency is significantly improved;

d. Strong applicability: According to the actual application situation, by updating the sample data, the feature font library and analysis rules can be quickly updated, while the analysis algorithm does not need to be updated.

e. Simple to implement and easy to popularize: the feature font library and analysis rules are highly reusable, the algorithm is simple, and it is easy to embed in various application systems.

Description of drawings

Figure 1 is a schematic diagram of the basic principle and process of address coding;

Fig. 2 is a schematic flow chart of the method of the present invention;

Figure 3 shows the parsing rules based on the tree structure;

Figure 4. Schematic diagram of address resolution (Address: No. 108, Chaotian Street, Xiongzhou Town, Liuhe County);

Figure 5. Schematic diagram of address resolution (Address: Jinshan Village, Babai Town, Liuhe County, Jiangsu Province);

Figure 6. Schematic diagram of address resolution (Address: No. 82, Taishan Village, Liucheng Town, Liuhe County);

Figure 7. Schematic diagram of address resolution (address: Babaiqiao Town Street, Liuhe District);

Figure 8. Schematic diagram of address resolution (Address: No. 1-2, Jiankang Lane, Xiongzhou Town, Liuhe District);

Figure 9. Schematic diagram of address resolution (address: No. 4, Ming Palace, Xuanwu District, Nanjing);

Figure 10. Schematic diagram of address resolution (address: Xiongzhou Town Central Farmer’s Market, Liuhe District);

Figure 11. Schematic diagram of address resolution (address: Room 302, No. 5, Beimenqiao Road);

Figure 12. Schematic diagram of address resolution (address: beside the East Bridge, Chengqiao Town, Liuhe District);

Figure 13. Schematic diagram of address resolution (Address: Room 104, No. 14, Xiangfuying, Xuanwu District);

Figure 14, the flow chart of Chinese address semantic analysis algorithm;

Figure 15, address representation algorithm flow chart;

Figure 16, address resolution algorithm flow chart.

Detailed ways

The method of the present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments.

On a PC with a main frequency of 1.73GHz and a memory of 1GB, the software system for realizing the technology of the present invention developed by the Visual C#.NET2003 platform takes the address data of 194,719 enterprises in Nanjing City, Jiangsu Province as sample data, and builds an address feature font library according to the sample data , according to the feature font library, according to the address representation rules, the address semantic analysis operation is performed.

Example 1: Chinese address string: No. 108, Chaotian Street, Xiongzhou Town, Liuhe County, as shown in Figure 4, parsed as: Liuhe County/Xiongzhou Town/No. Chaotian Street/108 ("/" indicates the segmentation position).

Example 2: Chinese address string: Jinshan Village, Babai Town, Liuhe County, Jiangsu Province, as shown in Figure 5, is parsed as: Jiangsu Province/Liuhe County/Babai Town/Jinshan Village.

Example 3: The Chinese address string: No. 82, Taishan Village, Liucheng Town, Liuhe County, as shown in Figure 6, is parsed as: Liuhe County/Liucheng Town/Taishan Village/No.82.

Example 4: Chinese address string: Babaiqiao Town Street, Liuhe District, as shown in Figure 7, parsed as: Liuhe District/Babaiqiao Town Street.

Example 5: Chinese address string: No. 1-2, Jiankang Lane, Xiongzhou Town, Liuhe District, as shown in Figure 8, parsed as: Liuhe District/Xiongzhou Town/Kangkang Lane/No. 1/-2.

Example 6: Chinese address string: Ming Palace No. 4, Xuanwu District, Nanjing City, as shown in FIG. 9 , resolved as: Nanjing City/Xuanwu District/Ming Palace/No.4.

Example 7: Chinese address string: central farmer's market in Xiongzhou Town, Liuhe District, as shown in Figure 10, parsed as: Liuhe District/Xiongzhou Town/central farmer's market.

Example 8: Chinese address string: Room 302, No. 5, Beimenqiao Road, as shown in Figure 11, is parsed as: Beimenqiao Road/No. 5/Room 302.

Example 9: Chinese address string: the side of the East Bridge, Chengqiao Town, Liuhe District, as shown in Figure 12, and parsed as: Liuhe District/Chengqiao Town/East Bridge/Bian.

Example 10: Chinese address string: Room 104, No. 14, Xiangfuying, Xuanwu District, as shown in Figure 13, parsed as: Xuanwu District/Xiangfuying/No. 14/Room 104.

Claims (1)

1, the semantic analytic method of a kind of Chinese address towards geocoding, its step is as follows: the first step: according to sample data, make up address feature character library

A, set up sample data: each address key element in the original address data is separated, formed sample data;

B, screening tagged word: last character of all address key elements in the sample data and the frequency of two characters are added up respectively, and according to descending ordering; Cumulative frequency is accounted for the single character screening of number percent more than 80% be tagged word; Cumulative frequency is accounted for 80% above two characters screening for tagged word, must last character not be single tagged word;

C, screening subsidiary characteristic word;

Tagged word that is screened and subsidiary characteristic word have constituted the feature character library;

Second step: according to the feature character library, represent rule, Chinese address is converted to the character string of digital form according to the address, 1 representation feature word wherein, 2 expression subsidiary characteristic words, a back character of two continuous repeated characteristic words of 3 expressions, the common character of 0 expression, 9 expression end marks; With 0 continuous character compression is one 0 character;

The 3rd step: make up the address resolution rule base

Chinese address is converted to after the numeric string, and it constitutes all follows following rule:

Can only the number in " 1 ", " 2 ", " 9 " after " 0 ";

Can only the number in " 0 ", " 1 ", " 2 ", " 3 ", " 9 " after " 1 ";

Can only the number in " 0 ", " 1 ", " 2 ", " 9 " after " 2 "; .

Can only the number in " 0 ", " 1 ", " 2 ", " 9 " after " 3 ";

Can only begin with the number in " 0 ", " 1 ", " 2 ", " 3 ";

Can only finish with " 9 ";

According to above-mentioned rule, address table is shown tree construction, each paths is represented a resolution rules, and the first order node of tree is respectively " 0 ", " 1 ", " 2 ", " 3 ", and its descendant's node is organized according to above-mentioned rule; When each paths arrives certain-length, can determine the fractionation point of address key element, thereby stop the continuation expansion in this path; Simultaneously, every resolution rules is stipulated concrete fractionation position, and represents with " f+ splits the position ";

The 4th step: the semantic parsing

Input: original address data, represent with Address_Before;

A, address are represented: original address is converted to numeral, and the result is Numbers_Before,

A, Numbers_Before are changed to empty string; The length of representing current original address to be resolved with n

B, i are from 1 up to n, and circulation is carried out:

If i character of original address is main tagged word, then Numbers_Before[i] be expressed as 1;

If i+1 character of original address is main tagged word, then Numbers_Before[i] be expressed as 3;

If i character of original address is the subsidiary characteristic word, then Numbers_Before[i] be expressed as 2;

If i character of original address is common character, then Numbers_Before[i] be expressed as 0;

The i assignment is i+1;

End loop;

C, add 9 at the end of Numbers_Before;

D, with one 0 of continuous a plurality of 0 boil down among the Numbers_Before;

B, address resolution: Numbers_Before is split as the address key element according to resolution rules, and the result is Numbers_After;

A, Numbers_After are changed to empty string; The length of representing Numbers_before with k;

B, m carry out circulation from 1 up to k:

If the left side m character of Numbers_Before and certain bar resolution rules coupling, two substring about then Numbers_before being split according to rule; Left side substring Numbers_Left saves as an address key element in the analysis result, no longer splits; Right substring Numbers-right proceeds to split; Numbers_Before is defined as Numbers-right;

The m assignment is length+1 of Numbers_Left;

Otherwise

The m assignment is m+1;

End loop;

C, address reduction: the analysis result of numeral is reduced to and original address corresponding characters string, and the result is Address_After.

Images