CN103077201B - A kind of unknown position evaluation method based on the detection of internet active iteration - Google Patents

Abstract

The invention relates to an unknown location estimation method based on Internet active iterative detection. Including the following steps: 1) Check the location entered by the user, if the database query fails, use the network engine to obtain the location-related webpage collection; 2) Extract the location description in the webpage and classify it; 3) Calculate the credibility of the search result <i >C _s </i>, if <i>C _s </i> meets the threshold <i>C _min </i>, skip to step 5; 4) Repeat steps 1 to 3 for the fuzzy position in the search results , until the reliability rate meets the threshold or reaches the limit of times; 5) Calculate the geographic range of the location description, and fuse it to obtain the approximate geographic range of the target location; the present invention makes full use of the abundant and dynamically changing geographic knowledge resources in the Internet to estimate the geographic range of the unknown location Approximate range. Aiming at various text location descriptions in the Internet, a semantic-based multi-scale location extraction method is adopted, and the approximate geographical range of the location is estimated by the Point-Radius algorithm.

Description

An Unknown Position Estimation Method Based on Internet Active Iterative Detection

technical field

The invention relates to an unknown location estimation method, in particular to an unknown location estimation method based on Internet active iterative detection.

Background technique

With the continuous development and improvement of positioning technologies such as GPS, the application fields of LBS (Location-Based Service) continue to expand, such as various electronic map service platforms (Baidu Map, Google Map, Bing Map, etc.), travel information query system, daily life point of interest query system, traffic query system, social network, etc. There are two main methods for these location service platforms or systems to provide location information query: one is to use GPS positioning, map operations, etc. to obtain more accurate location coordinates for query; the other is to use natural language location description for query. There are many uncertainties in qualitative or semi-quantitative location descriptions, but they are more in line with human expression habits and cognition. For natural language location queries, location databases need to store the mapping relationship between location names and geographic ranges. However, existing location databases are difficult to store all location names due to high construction costs, long time consumption, limited scale, and difficulty in updating. Instead, it mainly focuses on the collection and preservation of important locations such as main place names, addresses, and significant POIs. Therefore, it becomes impossible to inquire about a large number of locations with little significance and relatively low importance in life, thus contradicting the all-round, multi-level, and multi-granularity location service requirements. (References: Gu Jing, Research and Development of Location-Based Information Service Application System [D]. Xidian University, 2004; Xia Baoguo, Design and Implementation of GIS-Based Wuhan Tourism Information Query System [D]. Huazhong University of Science and Technology , 2006; Gao Weisi, location-based service and design of urban traffic navigation system [D]. Yunnan University, 2011; Yang Yuyao et al., a mobile Internet social model based on geographic location information [J]. Computer Research and Development, 2011;)

The Internet, as a large knowledge base, provides rich geographical knowledge, which can be used as an extended data source for location services. Location reference information for web searches requires the use of natural language understanding to extract location descriptions from large amounts of text information. Natural language understanding is a variety of theories and methods that can realize effective communication between humans and computers in natural language. The natural language understanding of location description is mainly to identify location names and location relationships. Regarding the recognition of location names, existing studies have focused on extracting geographically named entities or place names. There are two main methods: one is the rule-based method, which establishes a corpus and construction rules for geographically named entities or place names, and adopts rule matching. Recognition, this method has strict requirements on concept construction rules, which can improve the accuracy of the extraction results, but it makes the recall rate drop a lot, and it is difficult to solve the problem of fuzzy position and new position recognition; the other is based on statistics. Considering syntactic and semantic information, there are inevitably some problems in the acquisition of some low-frequency numerals and the noise introduced by adjacent high-frequency words. Regarding the identification of positional relationships, existing research has focused on extracting basic spatial relationships (topological relationships, metric relationships, orientation relationships, etc.), and there are two main methods: one is based on sentence analysis, which requires a thorough understanding of syntax structure and sentence semantics, there are problems of vulnerability and multiple ambiguities; one is a pattern-based method, which can avoid a thorough analysis of the sentence, but due to the richness of natural language expression, there are multiple ways of expressing the same information, which will make the pattern The number expanded rapidly. (References: Le Xiaoqiu et al., Extraction of Spatial Concepts in Natural Language Based on Spatial Semantic Roles [J]. Journal of Wuhan University Information Science Edition, 2005; Jiang Lin et al., Acquisition and Verification of Geographical Entity Concepts and Their Location Relationships [J] ].Computer Science, 2007; Li Lishuang et al., Recognition of Place Names in Chinese Text Based on Support Vector Machines [J]. Journal of Dalian Institute of Technology, 2007; Li Hanjing, Research on Spatial Concept Modeling Based on Natural Language Processing [D]. Harbin Institute of Technology, 2007; Li Yusen, Research on Geographically Named Entity Recognition for GIS [J]. Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition), 2008; Ma Long, Research on Chinese Place Name Recognition Based on Conditional Random Domain Model [D]. Dalian University of Technology, 2009; Tang Xuri et al., Research on text-based Chinese place name recognition [J]. Chinese Journal of Information, 2010; Jiang Wenming, Research on the extraction method of spatial orientation relationship for Chinese texts [D]. Nanjing Normal University, 2010; Shen Qijun, Chinese texts Research on Spatial Relationship Labeling Method [D]. Nanjing Normal University, 2010; Zhang Xueying et al., Rules-Based Analysis Method of Chinese Address Elements [J]. Journal of Geo-Information Science, 2010; Li Haiguang, Chinese Named Entity Relationship Based on Location and Semantic Features Extraction Research [D]. Hefei University of Technology, 2011; Du Ping et al., Chinese Place Name Recognition and Ambiguity Elimination——A Case Study of Administrative Division Place Names Above the County Level in China [J]. Remote Sensing Technology and Application, 2011.)

The location database has problems of limited scale and difficulty in updating. Geographical location information queries based on location databases (especially fuzzy location queries) may cause location names to be difficult to identify or coverage missing, which is not enough to meet user needs. The Internet contains a wealth of geographic knowledge, which can provide a large amount of descriptive information for locations of interest for estimating the coverage of "unknown" locations. And how to search location-related information from the Internet, and obtain the approximate geographical range of the "unknown" location therefrom, is the main work of the present invention.

Contents of the invention

The present invention mainly solves the technical problems existing in the prior art; it provides a geographical knowledge resource that can make full use of abundant quantities and dynamic changes in the Internet to realize the approximate range estimation of the target position.

Above-mentioned technical problem of the present invention is mainly solved by following technical scheme:

A method for estimating an unknown position based on Internet active iterative detection, characterized in that it comprises the following steps:

Step 1, check the user's input location query words; if the location cannot obtain geographic coverage from the spatial database, then actively start Internet iterative detection, that is, take the target location as the theme and use the web search engine to crawl the relevant information of the target location from the Internet;

Step 2, take the location query word as the theme to carry out initial detection, and use the network engine to obtain a collection of web pages containing the description of the target location from the Internet;

Step 3, performing geographic location analysis on the network document described in the target position obtained in step 2, that is, extracting a natural language position description from the network document, the natural language position description including a reference position and a spatial relationship;

Step 4, use the natural language location description obtained in step 3 to classify the location description; if the reference location of the location description can obtain geographical coverage from the location database, the location description is stored in the precise description set P, otherwise it is stored in the fuzzy description set A;

Step 5: Evaluate the current search credibility rate C _s ; if C _s is less than the search credibility threshold C _min , conduct a new round of Internet text search on the subject of the reference position in the fuzzy description set A; if Cs is greater than or equal to the search credibility threshold threshold C _min , skip to step 7;

Step 6, repeat steps 1 to 5 until the credibility rate of each round of search results meets the threshold or reaches the limit of the number of searches;

Step 7, calculating the approximate geographic scope and credibility of all location descriptions;

Step 8, integrating and refining the geographic coverage of multiple location descriptions to obtain the geographic range of the target location;

In step 3 of the above-mentioned unknown location estimation method based on Internet active iterative detection, the natural language location description recognition mainly includes location name recognition and spatial relationship recognition, and the semantic-based multi-scale extraction method is used to extract the natural language location description , including the following sub-steps:

In step 3.1, a corpus of location descriptions is established, which stores characteristic vocabulary expressing location names and spatial relationships and syntactic patterns of location descriptions; here, the establishment of a corpus can be established through manual induction and machine learning.

Step 3.2, with the support of the corpus, pattern matching is performed on the network text to obtain the location description;

Step 3.3, geographic and non-geographical semantic disambiguation of place names.

In step 4 of the above-mentioned unknown position estimation method based on Internet active iterative detection, the premise of using the reference position and spatial relationship to estimate the target position is that the reference position can obtain the precise geographical range from the position database, and set a single The location description is expressed according to Formula 1, RO is the name of the reference location, SR is the spatial relationship of the location, T is the occurrence time of the location description, C is the credibility of the location description, and S is the search reference of the reference object RO; The first K positions describe Loc _i and classify them according to the preconditions. When Loc _{i.RO meets the preconditions, Loc i is} stored in the precise description set P, otherwise it is stored in the fuzzy description set A;

Loc={RO, SR, T, C, S} Formula 1

In step 5 of the above-mentioned unknown position estimation method based on active iterative detection of the Internet, the specific method for evaluating the current search credibility C _s is to define the search credibility C _s as an evaluation index, and the search credibility is The ratio of the sum of the credibility of all location descriptions in P to the total number of location descriptions, as shown in Equation 2, m is the number of location descriptions in P, K is the total number of location descriptions, _Loci.C is the reliability of a location description Reliability.

$C_{\mathrm{the\; s}}=\frac{\underset{i=0}{\overset{m-1}{\mathrm{\&Sigma;}}}{\mathrm{Loc}}_i.C}{K}$ formula two

The credibility of the location description is calculated according to Equation 3, where ε is the attenuation parameter, n is the number of searches, and the credibility of the location description is set to 1 at the first search, and decays with the increase of the number of searches;

Loc _i .C=1*(ε) ⁿ Equation 3

When C _s satisfies the minimum credible threshold C _min , the precise description set P is directly output to estimate the target position; when C _s does not meet the conditions, a method based on multiple iterative searches on the Internet is used to ensure the search credibility, that is, A A new round of Internet search is carried out based on the fuzzy reference position in the network, and the geographical range of the reference position is first estimated through network resources, and then the target position is estimated using the reference position.

Step 6 in the above-mentioned unknown position estimation method based on Internet active iterative detection is the iterative search of the fuzzy reference position; according to the processing of steps 4 and 5, the set search result is expressed in formula 4, and n is the number of searches , m is the position number of the current search, P is the precise description set, A is the fuzzy description set, C _s is the search credibility.

WS[n][m]={P, A, C _s } Formula 4

The iterative search process includes the following sub-steps:

Step 6.1, store the fuzzy location description WS[0][0].A of the target location search result into the search set Q, set n=0, m=0;

Step 6.2, take the fuzzy description set WS[n][m].A in Q, judge whether n+1 has reached the limit of search times, if so, exit the search;

Step 6.3, sequentially take the location description Loc _i in WS[n][m].A to perform the n+1th search, obtain the search result WS[n+1][i], and associate it with the reference object RO search of the location description Reference, ie Loc _i.S = WS[n+1][i];

Step 6.4, remove the searched fuzzy description set WS[n][m].A from Q, check

Whether WS[n+1][i].C _s satisfies the threshold C _min , if not, put WS[n+1][i].A into the search set Q;

Step 6.5, check whether there is a fuzzy description set in Q, if yes, repeat step 6.2 to step 6.4 for iterative search.

In step 7 of the aforementioned unknown location estimation method based on Internet active iterative detection, since the fuzzy location description of the kth search result needs to refer to the k+1th search result, the calculation method is reversed, that is, from the last The search starts to calculate the geographic range, which includes the following sub-steps:

Step 7.1, define the number of searches in the search result WS as n, the number of search positions for the nth time is m, m=WS[n-1].size; define the geographic range set FC to store the geographic range of each search result;

Step 7.2, take the search result WS[n-1][m-1] of the nth search for the mth position;

Step 7.3, sequentially take the location Loc _y in WS[n-1][m-1].P, query the reference location coordinates based on the location database, and use the Point-Radius algorithm to calculate the geographic coverage FP(y) and its reliability CP (y);

Step 7.4, sequentially take the location Loc _x in WS[n-1][m-1].A, use Loc _x .S to query the reference location coordinates in the geographical range set FC, if the coordinates are successfully obtained, use Point-Radius The algorithm calculates the geographical coverage FA(y) and its credibility CA(y);

Step 7.5, integrate the geographic ranges of all locations in P and A, and obtain the geographic range FC(WS[n-1][m-1]) of the current search result;

Step 7.6, judge whether m-1 is greater than 0; if it is greater than 0, calculate the position of the next search result, let m=m-1, skip to step b); if it is less than or equal to 0, proceed to the next step;

Step 7.7, judge whether n-1 is greater than 0; if it is greater than 0, calculate the position of the previous search result, let n=n-1, m=WS[n-1].size, skip to step b); if If it is less than or equal to 0, proceed to the next step;

Step 7.8, output FC(WS[0][0]).

Therefore, the present invention has the following advantages: it can make full use of abundant and dynamically changing geographical knowledge resources in the Internet to realize the approximate range estimation of the target position. Due to the complex relationship between location information and non-location information in the Internet, and the diversity of information expression forms, the present invention uses a semantic-based multi-scale extraction method to extract location descriptions from webpage texts for natural language text information in the Internet, and uses Point- The Radius algorithm calculates the approximate geographic extent of the target location. .

Description of drawings

Fig. 1 is a flow chart of an Internet active search method.

Figure 2. Flowchart of location calculation based on Internet search results.

detailed description

The technical solution of the present invention will be further specifically described below through the embodiments and in conjunction with the accompanying drawings.

Example:

1. Theoretical basis.

1.1. Geographic Information Retrieval (GIR).

Geographic information retrieval is to return documents related to geographic information query according to the limitation of geographic query scope. The basic idea is to use web crawlers to search webpage collections from the Internet, identify place names in webpages through named entity recognition and classification, and grammatical analysis, so as to determine the geographical range of query words and documents, and finally calculate the correlation between documents and query words ( Including text association and spatial association) to return and sort the search results. At present, most geographic information retrieval mainly adopts keyword matching algorithm. Both search words and place names in network documents need to have a clear geographical coverage for association technology. This method is difficult to adapt to the situation of vague place names (such as the middle and lower reaches of the Yangtze River). Therefore, Cannot be used directly for unknown location estimates based on web searches. With reference to the idea of geographic information retrieval, the present invention proposes a multi-scale iterative search algorithm (as shown in Figure 1), which acquires network documents related to unknown locations based on the Internet, extracts location descriptions containing unknown locations, and then uses The approximate geographic extent of the unknown location is calculated with reference to the location and the spatial relationship. The main process is to obtain the web page collection from the Internet through meta-search, and extract the location description containing the query word in the web page based on semantics. If the location description does not meet the credibility rate to estimate the location of the query word, then perform a new round of the identified fuzzy location. This process is an iterative process. As long as the credibility rate condition is not satisfied or the search limit is not reached, the network search is continuously carried out to obtain reference information that can estimate the geographic range of the fuzzy location.

1.2. Location description georeferencing (GeoreferencingLocalityDescriptions, GLD).

Location description georeferencing is the conversion of a location from a textual description to a numerical description in a coordinate system. The ideal location description georeferencing process is to convert the text description into a digital description and map it on the map, and express the spatial range of the location and the uncertainty of the location distribution. Currently, the more popular algorithms are the Point-Radius algorithm and the Probability algorithm. . The Point-Radius method uses a point and the maximum error to describe the position and its uncertainty. The sources of uncertainty mainly considered include reference position (spatial range of reference position, geodetic datum, coordinate accuracy, map scale) and spatial relationship (distance Uncertainty of relationship and uncertainty of directional relationship), all uncertainty metrics are projected into one dimension as the maximum error of the target position, and the target position is expressed in a circular area formed by the point and the maximum error as the radius. The Probability method uses the uncertainty probability density surface to express the target position and its uncertainty, mainly considering the source of uncertainty including the spatial distribution of the target object, the imprecision and ambiguity of the spatial relationship, the incompleteness of the reference object, and the position Describe the uncertainty of itself. The Point-Radius method belongs to the position calculation of the quantitative method, which can obtain the geographical coverage of all possible points in the target position, and is suitable for semi-quantitative text position description; the Probability method cannot quantitatively calculate the geographical coverage of the target position, but can give the target position. Probability distributions, suitable for qualitative textual location descriptions.

2. Implementation process.

(1) Check that the user enters the target location query words; search the query words in the location database, if the location does not exist or the geographical coverage of the location is missing, then actively conduct a query based on the network search mode, that is, use the network search engine with the target location as the theme Crawl information about the target location from the Internet;

(2) Identify and extract natural language location descriptions (including reference locations and spatial relationships) in network documents; natural language location description recognition mainly includes location name identification and spatial relationship identification, and the present invention uses a semantic-based multi-scale extraction method Extract natural language location descriptions. First, establish a corpus of location descriptions through artificial induction and machine learning. The corpus stores the characteristic vocabulary expressing location names and spatial relationships as well as the syntactic patterns of location descriptions; then, with the support of the corpus, pattern matching is performed on network texts. Obtain location descriptions; finally, geographically based and non-geographically semantically disambiguated place names;

(3) Location description classification; the premise of using reference location and spatial relationship to estimate the target location is that the reference location can obtain an accurate geographical range from the location database, and set a single location description to be expressed according to formula (1), and RO is the reference location Name, SR is the spatial relationship of the location, T is the occurrence time of the location description, C is the credibility of the location description, and S is the search reference of the reference object RO. Loc _i is described by the first K positions in the extraction results, and classified according to the preconditions. When Loc _{i.RO meets the preconditions, Loc i is} stored in the precise description set P, otherwise it is stored in the fuzzy description set A;

Loc = {RO, SR, T, C, S} (1)

(4) Calculate the search credibility rate C _s ; the credibility of the location description in the search results must reach a certain level before it can be used to estimate the target location. The present invention proposes the search credibility rate C _s as an evaluation index, and the search credibility rate is The ratio of the sum of the credibility of all location descriptions in P to the total number of location descriptions, as shown in formula (2), m is the number of location descriptions in P, K is the total number of location descriptions, Loc _i.C is a certain location description credibility.

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; Loc</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; C</span>}}{\mathrm{<span\; class="notranslate">\; K</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

The credibility of the location description is calculated according to formula (3), where ε is the attenuation parameter, n is the number of searches, and the credibility of the location description is set to 1 at the first search, and decays as the number of searches increases.

Loc _i .C=1*(ε) ⁿ (3)

When C _s satisfies the minimum credible threshold C _min , the precise description set P is directly output to estimate the target position; when C _s does not meet the conditions, the present invention adopts a method based on Internet multiple iterative search to ensure the search credibility, namely Take the fuzzy reference position in A to conduct a new round of Internet search, first estimate the geographical range of the reference position through network resources, and then use the reference position to estimate the target position;

(5) Iterative search of fuzzy reference position; according to the processing of step 3 and step 4, set the search result to be expressed by formula (4), n is the number of searches, m is the sequence number of the current search position, P is the precise description set, A is the set of fuzzy descriptions, and C _s is the search confidence rate.

WS[n][m] = {P, A, C _s } (4)

The iterative search process is as follows:

a). Store the fuzzy location description WS[0][0].A of the target location search result into the search set Q, set n=0, m=0;

b). Take the fuzzy description set WS[n][m].A in Q, and judge whether n+1 reaches the limit of search times, and if so, exit the search;

c). Take the location description Loc _i in WS[n][m].A in order to search for the n+1th time, obtain the search result WS[n+1][i], and associate it with the reference object RO search of the location description Reference, ie Loc _i.S = WS[n+1][i];

d). Remove from Q the fuzzy description set WS[n][m].A that completes the search, check

Whether WS[n+1][i].C _s satisfies the threshold C _min , if not, put WS[n+1][i].A into the search set Q;

e). Check whether there is a fuzzy description set in Q, and if so, repeat step b) to step d) for iterative search;

(6) Calculating the approximate geographic scope and credibility of all location descriptions; since the fuzzy location description of the kth search result needs to refer to the k+1th search result, the present invention uses a reverse calculation method, that is, starting from the last The search begins a geographic extent calculation. As shown in Figure 2, the calculation process is as follows:

a). Define the number of searches in the search result WS as n, and the number of search positions for the nth time as m,

m=WS[n-1].size; Define the geographical range set FC to store the geographical range of each search result;

b). Take the search result WS[n-1][m-1] of the nth search for the mth position;

c). Take the location Loc _y in WS[n-1][m-1].P in turn, query the reference location coordinates based on the location database, and use the Point-Radius algorithm to calculate the geographical coverage FP(y) and its credibility CP (y);

d). Take the position Loc _x in WS[n-1][m-1].A in turn, use Loc _x .S to query the reference position coordinates in the geographical range set FC, if the coordinates are successfully obtained, use Point-Radius The algorithm calculates the geographic coverage FA(y) and its credibility CA(y);

e). Merge the geographic range of all locations in P and A to obtain the geographic range FC(WS[n-1][m-1]) of the current search result;

f). Determine whether m-1 is greater than 0; if it is greater than 0, calculate the position of the next search result, set m=m-1, and skip to step b); if it is less than or equal to 0, proceed to the next step;

g). Determine whether n-1 is greater than 0; if it is greater than 0, calculate the position of the previous search result, set n=n-1, m=WS[n-1].size, and skip to step b); if If it is less than or equal to 0, proceed to the next step;

h).Output FC(WS[0][0]);

The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which the present invention belongs can make various modifications or supplements to the described specific embodiments or adopt similar methods to replace them, but they will not deviate from the spirit of the present invention or go beyond the definition of the appended claims range.

Claims (3)

1., based on a unknown position evaluation method for internet active iteration detection, it is characterized in that, comprise the following steps:

Step 1, checks user's input position query word; If position cannot obtain geographical covering from spatial database, then initiatively start internet iteration detection, be namely the theme with target location and utilize network search engines to crawl target location relevant information from internet;

Step 2, is the theme with position enquiring word and carries out initial probe, utilizes network engine to obtain from internet to comprise the collections of web pages that target location describes;

Step 3, the network documentation that the target location obtained for step 2 describes carries out geographic position parsing, and namely from network documentation, extract natural language location expression, described natural language location expression comprises reference position and spatial relationship;

Step 4, the natural language location expression adopting step 3 to obtain carries out location expression classification; If the reference position of location expression can obtain geographical covering from location database, location expression stored in accurate description collections P, otherwise stored in vague description set A;

Step 5, the credible rate Cs of assessment current search; If Cs is less than the credible threshold value C of search _min, be the theme with the reference position in vague description set A and carry out new round internet text search, if Cs is greater than or equal to the credible threshold value C of search _min, then skipping to the concrete grammar that step 7 assesses the credible rate Cs of current search is: the credible rate C of definition search _sas evaluation index, searching for credible rate is the confidence level sum of all location expressions in P and the ratio of location expression sum, and shown in two, m is location expression number in P, and K is location expression sum, Loc _i.C be the confidence level of certain location expression:

$C_s=\frac{\underset{i=0}{\overset{m-1}{\&Sigma;}}{\mathrm{Loc}}_i.C}{K}$ Formula two

The confidence level of location expression calculates according to formula three, and wherein ε is attenuation parameter, and n is searching times, and it is 1 when searching for first that desired location describes confidence level, and decays along with the increase of searching times;

Loc _i.C=1* (ε) ⁿformula three

Work as C _smeet minimum credible threshold value C _mintime, directly export accurate description collections P and carry out target location estimation; Work as C _swhen not satisfying condition, the method based on the search of internet successive ignition is adopted to ensure to search for credible rate, new round internet hunt is carried out in the fuzzy reference position of namely getting in A, first estimates reference position geographic range by Internet resources, and then utilizes estimation target location, reference position.

Step 6, repeats step 1 to step 5, till often the credible rate of wheel Search Results meets threshold value or reaches searching times restriction;

Step 7, calculates approximate geographic range and the confidence level thereof of all location expressions;

Step 8, the multiple location expression geography of integrated and refinement covers, and obtains the geographic range of target location;

Described step 6 is fuzzy reference position iterative search; The process of foundation step 4 and step 5, setting search result adopts formula four to express, and n is searching times, and m is that P is accurate description collections, and A is vague description set, C when time position number of search _sthe credible rate of search:

WS [n] [m]={ P, A, C _sformula four

Described iterative search procedures comprises following sub-step:

Step 4.1, describes WS [0] [0] .A stored in search set Q, if n=0, m=0 by the ambiguous location of target location Search Results;

Step 4.2, gets vague description set WS [n] [m] .A in Q, judges whether n+1 reaches searching times restriction, if it is exits search;

Step 4.3, gets location expression Loc in WS [n] [m] .A successively _icarry out (n+1)th search, obtain Search Results WS [n+1] [i], and the references object RO search being associated with location expression is quoted, be i.e. Loc _i.S=WS [n+1] [i];

Step 4.4, has removed vague description set WS [n] [m] .A of search from Q, checks

WS [n+1] [i] .C _swhether meet threshold value C _minif do not meet, WS [n+1] [i] .A is put into search set Q;

Step 4.5, checks in Q whether there is vague description set, if had, repeats step 4.2 and carries out iterative search to step 4.4;

Described step 7, the ambiguous location due to kth Search Results describes to be needed, with reference to kth+1 Search Results, to adopt the mode that backward calculates, and namely from last search, carries out geographic range calculating, specifically comprises following sub-step:

Step 5.1, in definition Search Results WS, searching times is n, and n-th searching position number is m, m=WS [n-1] .size; Definition geographic range set FC stores the geographic range of each Search Results;

Step 5.2, gets Search Results WS [n-1] [m-1] of n-th search m position;

Step 5.3, gets the position Loc in WS [n-1] [m-1] .P successively _y, position-based data base querying reference position coordinate, utilizes Point-Radius algorithm to calculate geographical covering FP (y) and confidence level CP (y) thereof;

Step 5.4, gets the position Loc in WS [n-1] [m-1] .A successively _x, utilize Loc _x.S in geographic range set FC, inquire about reference position coordinate, if successfully obtain coordinate, then utilize Point-Radius algorithm to calculate geographical covering FA (y) and confidence level CA (y) thereof;

Step 5.5, merges the geographic range of all positions in P and A, obtains the geographic range FC (WS [n-1] [m-1]) when time Search Results;

Step 5.6, judges whether m-1 is greater than 0; If be greater than 0, then carry out the position calculation of next Search Results, make m=m-1, skip to step 5.2; If be less than or equal to 0, then carry out next step;

Step 5.7, judges whether n-1 is greater than 0; If be greater than 0, then carry out the position calculation of a front Search Results, make n=n-1, m=WS [n-1] .size, skips to step 5.2; If be less than or equal to 0, then carry out next step;

Step 5.8, exports FC (WS [0] [0]);

2. a kind of unknown position evaluation method based on the detection of internet active iteration according to claim 1, it is characterized in that, in described step 3, the identification of natural language location expression mainly comprises the identification of location name identification and spatial relationship, adopt the multiple dimensioned extracting method based on semanteme to extract natural language location expression, specifically comprise following sub-step:

Step 3.1, sets up the corpus of location expression, stores and express location name and the feature vocabulary of spatial relationship and the syntactic pattern of location expression in corpus;

Step 3.2, under the support of corpus, carries out pattern match to network text, obtains location expression;

Step 3.3, eliminates place name ambiguity based on geography with the semanteme of non-geographic.

3. a kind of unknown position evaluation method based on the detection of internet active iteration according to claim 1, it is characterized in that, in described step 3, the prerequisite utilizing reference position and spatial relationship estimation target location is that reference position can obtain accurate geographic range from location database, set single location expression to express according to formula one, RO is reference position title, SR is locational space relation, T is the time of origin of location expression, C is the confidence level that location expression has, and S is the searching for reference of references object RO; Extract K location expression Loc before in result _i, and classify according to precondition, work as Loc _iwhen .RO meeting precondition, Loc _istored in accurate description collections P, otherwise stored in vague description set A;

Loc={RO, SR, T, C, S} formula one.