CN103729466A - Name country identification method based on WEB and GBBoosting algorithms - Google Patents

Abstract

The invention discloses a name and country identification method based on WEB and GBBoosting algorithm, which belongs to the technical field of WEB data mining. The method includes the following steps: Step 1: Extract the names of academics in colleges and universities through WEB data extraction technology; Step 2: Construct GBBoosting algorithm: Construct weak classifiers, each weak classifier outputs a weak classification hypothesis for the input sample, and pass all weak classifiers The fusion of weights constitutes a strong classifier; Step 3: Identify the country to which it belongs through the GBBoosting algorithm. The national name recognition method based on WEB and GBBoosting algorithm of the present invention effectively solves the problem that the names of two countries cannot be classified under the similar spelling mode; at the same time, this method is easier to implement than other existing classification methods, and can It is better used in engineering practices such as semantic labeling of names and cities or cities and countries.

Description

Recognition Method of Person's Name and Country Based on WEB and GBBoosting Algorithm

technical field

The invention belongs to the technical field of WEB data mining, and in particular relates to a name and country recognition method based on WEB and GBBoosting algorithm.

Background technique

With the rapid development of the Internet and the increasing abundance of WEB resources, in order to quickly and accurately mine necessary and meaningful data from massive data information, in recent years, WEB semantic analysis technology and text classification technology have been widely used in the field of WEB data mining. Applications, WEB-based applications have changed users' living habits and working styles to some extent, and are also welcomed and appreciated by more and more users.

Classification methods such as KNN and Bayesian have achieved good classification results in many classification fields. For example, Jiemei et al. applied KNN to the field of image processing, and proposed a KNN-based classification algorithm to detect the gray level inhomogeneity of MR images. Correction segmentation method (Patent No.: 201010583560.6, publication date: 2011.07.27); Yang Liu et al. applied Bayesian to the field of computer software, and proposed an intelligent SMS classification and search method based on improved Bayesian classification (Patent No. : 201310356056.6, public date: 2013.12.04). However, the classification accuracy of the above-mentioned classification method in the scene of classification of personal names by country needs to be further improved, especially when the spelling of personal names in two countries is similar, the classification accuracy is only higher than random guessing. It can be seen that the above classification algorithm has great limitations in the application of the classification of personal names and countries.

Based on the deficiencies of the above-mentioned classification methods in the classification of personal names and countries, the present invention proposes a Boosting-based GBBoosting algorithm, which aims to solve the problems existing in the scene of classification of personal names and countries. Compared with other classification algorithms, its classification The accuracy and recall rate have been greatly improved, especially when the spelling of names in two countries is similar, the performance is excellent. It has very important practical significance and broad application prospects to apply the GBBoosting algorithm to recognition scenarios such as person names and cities, etc., to carry out semantic annotation of names or cities by country, and then apply it to the hot social field.

Contents of the invention

In view of this, the object of the present invention is to provide a method for identifying national names based on WEB and GBBoosting algorithm. The method uses WEB data extraction technology to extract the names of academics in colleges and universities, and constructs weak classifiers. Output a weak classification hypothesis, form a strong classifier through the weight fusion of all weak classifiers, and finally use the GBBoosting algorithm to identify the country to which the name belongs.

To achieve the above object, the present invention provides the following technical solutions:

A name and country recognition method based on WEB and GBBoosting algorithm, including the following steps: Step 1: Extracting the names of academics in universities through WEB data extraction technology; Step 2: Constructing GBBoosting algorithm: Constructing weak classifiers, each weak classifier pairs input The sample outputs a weak classification hypothesis, and a strong classifier is formed by merging the weights of all weak classifiers; Step 3: Identify the country to which it belongs through the GBBoosting algorithm.

Further, in step 1, the college page of the university is obtained through the GOOGLE search engine interface, and then semantic analysis is performed on the college page to obtain the page where the scholars of the college are located, and finally the scholar information in the extracted page is obtained through named entity recognition technology and semantic analysis technology.

Further, in step 2, the construction steps of the weak classifier specifically include:

1) Represent the two types of training text as vectors ${\stackrel{\&Right\; Arrow;}{V}}_1=(x_1,x_2,...,x_i,...,x_{\mathrm{no}}),{\stackrel{\&Right\; Arrow;}{V}}_2=({\mathrm{the\; y}}_1,{\mathrm{the\; <figure-callout\; id="2"\; label="y"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...,{\mathrm{the\; y}}_i,...,{\mathrm{the\; y}}_{\mathrm{no}});$

的中间向量 2) According to the formula Two training texts are calculated

The intermediate vector of

${\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

计算出中间向量

的垂直向量

对于任意一个测试向量a_i，如果(w_i·a_i)＞0，则将a_i的标签标记为+1，如果(w_i·a_i)＜0，则将a_i的标签标记为-1；3) According to the formula

Calculate the intermediate vector

The vertical vector of

For any test vector a _i , if (w _i ·a _i )>0, mark the label of a _i as +1, and if (w _i ·a _i )<0, mark the label of a _i as - 1;

The weak classifiers are iterated, and their weights are fused to form a strong classifier. The specific steps are as follows:

First, given two training sets D ₁ =(x ₁ ,x ₂ ,..., _xi ,...,x _n ), D ₂ =(y ₁ ,y ₂ ,...,y _i , ...,y _n ), a test set D _Test ＝(z ₁ ,z ₂ ,..., _zi ,...,z _n ), the training sets D1, D2, and the test set D _Test represent respectively In vector form: ${\mathrm{D.}}_1=(\stackrel{\&Right\; Arrow;}{x_1},\stackrel{\&Right\; Arrow;}{x_2},...,\stackrel{\&Right\; Arrow;}{x_i},...,\stackrel{\&Right\; Arrow;}{x_{\mathrm{no}}}),{\mathrm{D.}}_2=(\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_1},\stackrel{\&Right\; Arrow;}{{\mathrm{the\; <figure-callout\; id="2"\; label="y"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">y</figure-callout>}}_2},...,\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_i},...,\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_{\mathrm{no}}}),{\mathrm{D.}}_{\mathrm{test}}=(\stackrel{\&Right\; Arrow;}{z_1},\stackrel{\&Right\; Arrow;}{{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">z</figure-callout>}}_2},...,\stackrel{\&Right\; Arrow;}{z_i},...,\stackrel{\&Right\; Arrow;}{z_{\mathrm{no}}}),$ And initialize D ₁ , D ₂ , and the sample weights in D _Test respectively;

2)根据线性分类器的构造过程，得到与两个向量

的中间向量

垂直的向量

生成弱分类器H(x)₁；经过p次循环，得到p个不同的垂直向量

p个弱分类器h(x)₁,h(x)₂,...,h(x)_p；最终H(x)＝h(x)₁+h(x)₂+...+h(x)_p，即

Secondly, 1) Randomly select M(N/5<M<N) samples from D ₁ and D ₂ to form subsets D ₁₁ and D ₂₁ , add correspondingly to the vectors in subsets D ₁₁ and D ₂₁ and Normalize to get two vectors

2) According to the construction process of the linear classifier, the two vectors

The intermediate vector of

vertical vector

Generate a weak classifier H(x) ₁ ; after p cycles, get p different vertical vectors

p weak classifiers h(x) ₁ ,h(x) ₂ ,...,h(x) _p ; final H(x)=h(x) ₁ +h(x) ₂ +...+h (x) _p , ie

Further, in Step 3, the name of the university scholar is identified by the GBBoosting algorithm to identify the country to which the scholar belongs.

The beneficial effect of the present invention is that: the present invention provides a method for identifying country names based on WEB and GBBoosting algorithm, which effectively solves the problem that the names of people in two countries have similar spellings and cannot be classified; at the same time, this method is better than the existing Other classification methods are easier to implement and can be better applied to engineering practices such as semantic labeling of person names or cities.

Description of drawings

In order to make the purpose, technical scheme and beneficial effect of the present invention clearer, the present invention provides the following drawings for illustration:

Fig. 1 is the macro flow chart of method for the present invention;

Figure 2 is a vector similarity calculation diagram;

Figure 3 is a structural diagram of a weak classifier;

Fig. 4 is the micro flow chart of this method.

Detailed ways

The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Fig. 1 is the macro-flow chart of the method of the present invention, as shown in the figure, this method comprises the following steps: Step 1: Extract the names of academics in universities through WEB data extraction technology; Step 2: Construct GBBoosting algorithm: Construct a weak classifier, each A weak classifier outputs a weak classification hypothesis for the input sample, and forms a strong classifier through the weight fusion of all weak classifiers; Step 3: Identify the country to which it belongs through the GBBoosting algorithm.

Fig. 4 is a micro-flow chart of the method, and the specific implementation steps of the method will now be described in conjunction with Fig. 4 .

1. Extract the names of university scholars through WEB data extraction technology

1) Search "university+computerscience" through the GOOGLE search engine to find the homepage of the college; 2) Find the information page of all scholars in the college through the homepage of the college. The names of the scholars in the school generally exist in the corresponding college (department). As long as you find the URL of the corresponding college (department), you can get the names and home page addresses of all the scholars in the school. In step 2, find the URL of the computer school (department) corresponding to the university. After observing the URL addresses of the two pages of the school (department) and school scholars, you can get two rules:

①The latter address contains the former address.

②The latter address also includes the features of "people, faculty, faculty&Advisors".

It is only necessary to traverse all the links in the School of Computer Science (Department), and filter out the URLs in the links that meet the above two rules and whose corresponding text is "faculty or people". Through experiments, it is found that generally two URL addresses can be filtered out. The two URLs are because the college generally contains the people menu, and faculty belongs to the submenu link of people, the second URL is the required link, so when there are two URLs, select the second URL address, otherwise select the first URL address. Finally, enter the filtered URL to get the names and corresponding personal homepages of all scholars. 3) Extract the names and homepages of all scholars through the faculty page of the School of Computer Science (Department). Extract all the links on the faculty page of the School of Computer Science (Department), find the text corresponding to the link, and analyze whether the text is a person's name through named entity technology.

2. Implement the GBBoosting algorithm: construct weak classifiers, each weak classifier outputs a weak classification hypothesis for the input sample, and form a strong classifier through the weight fusion of all weak classifiers.

The construction of the weak classifier is to judge the size of the vector inner product of two types of texts through the simple space vector similarity, that is, to find the angle between the two vectors. As shown in Figure 2, the more similar the two texts are, the smaller the angle between the corresponding vectors is, and the larger the cosine of the angle is. As shown in Figure 3, the weak classifier is improved on the basis of the similarity of simple space vectors to construct a simple linear classifier. The specific steps are as follows:

计算出两种训练文本

的中间向量

2)根据公式

计算出中间向量的垂直向量

Step 1: Given two types of training text vector representations ${\stackrel{\&Right\; Arrow;}{V}}_1=(x_1,x_2,...,x_i,...,x_{\mathrm{no}}),{\stackrel{\&Right\; Arrow;}{V}}_2=({\mathrm{the\; y}}_1,{\mathrm{the\; <figure-callout\; id="2"\; label="y"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...,{\mathrm{the\; y}}_i,...,{\mathrm{the\; y}}_{\mathrm{no}});$ Step 2: 1) According to the formula

Two training texts are calculated

The intermediate vector of

2) According to the formula

Calculate the intermediate vector The vertical vector of

$\stackrel{<span\; class="notranslate">\; \&Right\; Arrow;</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="m"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">m</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

和门限值0，对于任意一个测试向量a_i，如果(w_i·a_i)＞0，则将a_i的标签标记为+1，如果(w_i·a_i)＜0，则将a_i的标签标记为-1。Step 3: There is a d-dimensional vector

and threshold value 0, for any test vector a _i , if (w _i ·a _i )>0, mark the label of a _i as +1, if (w _i ·a _i )<0, set a The label of _i is marked as -1.

通过计算得到与两个向量的中间向量

垂直的向量将测试集D_Test＝(z₁,z₂,...,z_i,...,z_n)中的每个样本与向量V做点积，通过点积结果的正负判断样本的分类，其具体步骤如下：Weak classifiers are the basis for implementing the GBBoosting algorithm. Each weak classifier outputs a weak classification hypothesis for the input sample, and a strong classifier is formed by merging the weights of all weak classifiers. Given two training sets D ₁ =(x ₁ ,x ₂ ,..., _xi ,...,x _n ), D ₂ =(y ₁ ,y ₂ ,...,y _i ,.. .,y _n ). Randomly select M samples from D ₁ and D ₂ respectively, and generate two vectors

Calculate the intermediate vector with two vectors

vertical vector Do a dot product of each sample in the test set D _Test = (z ₁ ,z ₂ ,..., _zi ,...,z _n ) with the vector V, and judge the classification of the sample by the positive or negative of the dot product result , the specific steps are as follows:

Step 1: Two training sets D ₁ =(x ₁ ,x ₂ ,..., _xi ,...,x _n ), D ₂ =(y ₁ ,y ₂ ,...,y _i ,. ..,y _n ), a test set D _Test =(z ₁ ,z ₂ ,..., _zi ,...,z _n ), the training sets D1, D2, and the test set D _Test are expressed as Vector form: ${\mathrm{D.}}_1=(\stackrel{\&Right\; Arrow;}{x_1},\stackrel{\&Right\; Arrow;}{x_2},...,\stackrel{\&Right\; Arrow;}{x_i},...,\stackrel{\&Right\; Arrow;}{x_{\mathrm{no}}}),{\mathrm{D.}}_2=(\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_1},\stackrel{\&Right\; Arrow;}{{\mathrm{the\; <figure-callout\; id="2"\; label="y"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">y</figure-callout>}}_2},...,\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_i},...,\stackrel{\&Right\; Arrow;}{{\mathrm{the\; y}}_{\mathrm{no}}}),{\mathrm{D.}}_{\mathrm{test}}=(\stackrel{\&Right\; Arrow;}{z_1},\stackrel{\&Right\; Arrow;}{{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="HDA0000457910160000012.png"\; state="\{\{state\}\}">z</figure-callout>}}_2},...,\stackrel{\&Right\; Arrow;}{z_i},...,\stackrel{\&Right\; Arrow;}{z_{\mathrm{no}}}),$ And initialize D ₁ , D ₂ , and sample weights in D _Test respectively.

2)根据线性分类器的构造过程，得到与两个向量

的中间向量

垂直的向量

生成弱分类器H(x)₁。经过p次循环，得到p个不同的垂直向量

p个弱分类器h(x)₁,h(x)₂,...,h(x)_p。Step 2: 1) Randomly select M (N/5<M<N) samples from D ₁ and D ₂ to form subsets D ₁₁ and D ₂₁ , and add correspondingly to the vectors in subsets D ₁₁ and D ₂₁ and normalize to get two vectors

2) According to the construction process of the linear classifier, the two vectors

The intermediate vector of

vertical vector

Generate a weak classifier H(x) ₁ . After p cycles, get p different vertical vectors

p weak classifiers h(x) ₁ ,h(x) ₂ ,...,h(x) _p .

Step 3: H(x)＝h(x) ₁ +h(x) ₂ +...+h(x) _p , namely

Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that it can be described in terms of form and Various changes may be made in the details without departing from the scope of the invention defined by the claims.

Claims (4)

1. the name country origin recognition methods based on WEB and GBBoosting algorithm, is characterized in that: comprise the following steps: step 1: by WEB Data Extraction Technology, extract the scholar of colleges and universities name;

Step 2: structure GBBoosting algorithm: structure Weak Classifier, each Weak Classifier, to weak typing hypothesis of input sample output, is merged and is formed a strong classifier by the weight of all Weak Classifiers;

Step 3: by the country origin under GBBoosting algorithm identified.

2. the name country origin recognition methods based on WEB and GBBoosting algorithm according to claim 1, it is characterized in that: in step 1, by GOOGLE search engine interface, obtain institute of the colleges and universities page, then at institute's page, carry out semantic analysis and obtain the scholar of the institute place page, finally by named entity recognition technology and semantic analysis technology, obtain extracting the scholar's information in the page.

3. the name country origin recognition methods based on WEB and GBBoosting algorithm according to claim 1, is characterized in that: in step 2, the constitution step of Weak Classifier specifically comprises:

1) by the training text vector representation of two types, be ${\stackrel{\&RightArrow;}{V}}_1=(x_1,x_2,...,x_i,...,x_n),{\stackrel{\&RightArrow;}{V}}_2=(y_1,y_2,...,y_i,...,y_n);$

2) according to formula

calculate two kinds of training texts

intermediate vector

${\stackrel{\&RightArrow;}{V}}_3=(z_1{,z}_2,...,z_i,...,z_n);$

3) according to formula

calculate intermediate vector

vertical vector

for any one test vector a _iif, (w _ia _i) > 0, by a _ilabel be+1, if (w _ia _i) < 0, by a _ilabel be-1;

Iteration Weak Classifier, its weights merge and form strong classifier, and its concrete steps are as follows:

First, given two training set D ₁=(x ₁, x ₂..., x _i..., x _n), D ₂=(y ₁, y ₂..., y _i..., y _n), a test set D _test=(z ₁, z ₂..., z _i..., z _n), by training set D1, D2, test set D _test, be expressed as vector form: $D_1=(\stackrel{\&RightArrow;}{x_1},\stackrel{\&RightArrow;}{x_2},...,\stackrel{\&RightArrow;}{x_i},...,\stackrel{\&RightArrow;}{x_n}),D_2=(\stackrel{\&RightArrow;}{y_1},\stackrel{\&RightArrow;}{y_2},...,\stackrel{\&RightArrow;}{y_i},...,\stackrel{\&RightArrow;}{y_n}),D_{\mathrm{Test}}=(\stackrel{\&RightArrow;}{z_1},\stackrel{\&RightArrow;}{z_2},...,\stackrel{\&RightArrow;}{z_i},...,\stackrel{\&RightArrow;}{z_n}),$ And difference initialization D ₁, D ₂, D _testin sample weights;

Secondly, 1) from D ₁, D ₂in choose at random the individual sample of M (N/5<M<N) composition subset D ₁₁, D ₂₁, respectively to subset D ₁₁, D ₂₁in corresponding be added and unit obtains two vectors of vector 2), according to the construction process of linear classifier, obtain and two vectors

intermediate vector

vertical vector

generate Weak Classifier H (x) ₁; Through p circulation, obtain p different vertical vector p Weak Classifier h (x) ₁, h (x) ₂..., h (x) _p; Final H (x)=h (x) ₁+ h (x) ₂+ ...+h (x) _p,

4. the name country origin recognition methods based on WEB and GBBoosting algorithm according to claim 1, is characterized in that: in step 3, the scholar of colleges and universities name is gone out to scholar belonging country by GBBoosting algorithm identified.

Images