|
|
| Line 1: |
Line 1: |
| {{orphan|date=July 2010}}
| | e - Shop Word - Press is a excellent cart for your on the web shopping organization. Affilo - Theme is the guaranteed mixing of wordpress theme that Mark Ling use for his internet marketing career. If you have any sort of questions concerning where and how you can make use of [http://GET7.pw/wordpress_backup_plugin_149100 wordpress backup], you can contact us at our own web-site. Step-4 Testing: It is the foremost important of your Plugin development process. Dead links are listed out simply because it will negatively have an influence on the website's search engine rating. In the most current edition you can customize your retailer layout and display hues and fonts similar to your site or blog. <br><br>You just download ready made templates to a separate directory and then choose a favorite one in the admin panel. While direct advertising is limited to few spots in your site and tied to fixed monthly payment by the advertisers, affiliate marketing can give you unlimited income as long as you can convert your traffic to sales. Which is perfect for building a mobile site for business use. You can add new functionalities and edit the existing ones to suit your changing business needs. Now a days it has since evolved into a fully capable CMS platform which make it, the best platform in the world for performing online business. <br><br>Just ensure that you hire experienced Word - Press CMS developer who is experienced enough to perform the task of Word - Press customization to get optimum benefits of Word - Press CMS. But if you are not willing to choose cost to the detriment of quality, originality and higher returns, then go for a self-hosted wordpress blog and increase the presence of your business in this new digital age. Setting Up Your Business Online Using Free Wordpress Websites. To turn the Word - Press Plugin on, click Activate on the far right side of the list. Socrates: (link to ) Originally developed for affiliate marketers, I've used this theme again and again to develop full-fledged web sites that include static pages, squeeze pages, and a blog. <br><br>You can add keywords but it is best to leave this alone. Russell HR Consulting provides expert knowledge in the practical application of employment law as well as providing employment law training and HR support services. Enterprise, when they plan to hire Word - Press developer resources still PHP, My - SQL and watch with great expertise in codebase. Fast Content Update - It's easy to edit or add posts with free Wordpress websites. The Pakistani culture is in demand of a main surgical treatment. <br><br>Under Settings —> Reading, determine if posts or a static page will be your home page, and if your home page is a static page, what page will contain blog posts. Here's a list of some exciting Word - Press features that have created waves in the web development industry:. It can be concluded that white label SEO comprise of a third party who resells a contract involving IT expert or consultant, SEO professional and end user. with posts or testimonials updated as they are uploaded to a particular section of the website. However, if you're just starting out your blog site or business site, you can still search for an ideal theme for it without breaking your bank account. |
| | |
| '''Cluster labeling''' is closely related to the concept of [[text clustering]]. This process tries to select descriptive labels for the clusters obtained through a clustering algorithm such as Flat Clustering and [[Hierarchical Clustering]]. For example, a cluster of documents that talks about various [[internet protocols]] can be best labeled as "Internet Protocols". Typically, the labels are obtained by examining the contents of the documents in a cluster. A good label not only summarizes the central concept of a cluster but also uniquely differentiates it from other clusters in the collection.
| |
| | |
| ==Differential Cluster Labeling==
| |
| Differential cluster labeling labels a cluster by comparing the terms in one cluster with the terms occurring in other clusters. The techniques used for [[feature selection]] in [[information retrieval]], such as [[mutual information]] and [[Pearson's chi-squared test|chi-squared feature selection]], can also be applied to differential cluster labeling. Terms having very low frequency are not the best in representing the whole cluster and can be omitted in labeling a cluster. By omitting those rare terms and using a differential test, one can achieve the best results with differential cluster labeling.<ref>Manning, Christopher D., Prabhakar Raghavan, and Hinrich Schutze. ''Introduction to Information Retrieval''. Cambridge: Cambridge UP, 2008. ''Cluster Labeling''. Stanford Natural Language Processing Group. Web. 25 Nov. 2009. <http://nlp.stanford.edu/IR-book/html/htmledition/cluster-labeling-1.html>.</ref>
| |
| | |
| ===Mutual Information===
| |
| | |
| {{Main|Mutual information}}
| |
| | |
| In the fields of [[probability theory]] and [[information theory]], mutual information measures the degree of dependence of two [[random variables]]. The mutual information of two variables X and Y is defined as:
| |
| | |
| <math>I(X, Y) = \sum_{x\in X}{ \sum_{y\in Y} {p(x, y)log_2\left(\frac{p(x, y)}{p_1(x)p_2(y)}\right)}}</math>
| |
| | |
| where ''p(x, y)'' is the [[joint probability|joint probability distribution]] of the two variables, ''p<sub>1</sub>(x)'' is the probability distribution of X, and ''p<sub>2</sub>(y)'' is the probability distribution of Y.
| |
| | |
| In the case of cluster labeling, the variable X is associated with membership in a cluster, and the variable Y is associated with the presence of a term.<ref>Manning, Christopher D., Prabhakar Raghavan, and Hinrich Schutze. ''Introduction to Information Retrieval''. Cambridge: Cambridge UP, 2008. ''Mutual Information''. Stanford Natural Language Processing Group. Web. 25 Nov. 2009. <http://nlp.stanford.edu/IR-book/html/htmledition/mutual-information-1.html>.</ref> Both variables can have values of 0 or 1, so the equation can be rewritten as follows: | |
| | |
| <math>I(C, T) = \sum_{c\in {0, 1}}{ \sum_{t\in {0, 1}} {p(C = c, T = t)log_2\left(\frac{p(C = c, T = t)}{p(C = c)p(T = t)}\right)}}</math>
| |
| | |
| In this case, ''p(C = 1)'' represents the probability that a randomly selected document is a member of a particular cluster, and ''p(C = 0)'' represents the probability that it isn't. Similarly, ''p(T = 1)'' represents the probability that a randomly selected document contains a given term, and ''p(T = 0)'' represents the probability that it doesn't. The [[joint probability|joint probability distribution function]] ''p(C, T)'' represents the probability that two events occur simultaneously. For example, ''p(0, 0)'' is the probability that a document isn't a member of cluster ''c'' and doesn't contain term ''t''; ''p(0, 1)'' is the probability that a document isn't a member of cluster ''c'' and does contain term ''t''; and so on.
| |
| | |
| ====Example====
| |
| The following example calculates the mutual information between the cluster "commerce" and the term "tariff":
| |
| {| class="wikitable"
| |
| |-
| |
| !
| |
| ! Documents in "commerce"
| |
| ! Documents not in "commerce"
| |
| |-
| |
| | Documents containing "tariff"
| |
| | 60
| |
| | 10,000
| |
| |-
| |
| | Documents not containing "tariff"
| |
| | 200
| |
| | 500,000
| |
| |}
| |
| Total number of documents = (60 + 200 + 10,000 + 500,000) = 510,260
| |
| | |
| P (C = 1, T = 1) = 60/510,260
| |
| | |
| P (C = 1, T = 0) = 200/510,260
| |
| | |
| P (C = 0, T = 1) = 10,000/510,260
| |
| | |
| P (C = 0, T = 0) = 500,000/510,260
| |
| | |
| P (C = 1) = (# of documents in the cluster) / (total number of documents) = (60 + 200) / 510,260 = 260/510,260
| |
| | |
| P (C = 0) = (# of documents not in the cluster) / (total number of documents) = (10,000 + 500,000) / 510,260 = 510,000/510,260
| |
| | |
| P (T = 1) = (# of documents containing the term) / (total number of documents) = (60 + 10,000) / 510,260 = 10,060/510,260
| |
| | |
| P (T = 0) = (# of documents not containing the term) / (total number of documents) = (200 + 500,000) / 510,260 = 500,200/510,260
| |
| | |
| Plugging these probabilities into the above equation gives us the following MI value:
| |
| | |
| <math>I(C, T) = \frac{60}{510,260} log_2\left(\frac{60/510,260}{260/510,260 * 10,060/510,260}\right)</math>
| |
| <math>+ \frac{200}{510,260} log_2\left(\frac{200/510,260}{260/510,260 * 500,200/510,260}\right)</math>
| |
| <math>+ \frac{10,000}{510,260} log_2\left(\frac{10,000/510,260}{510,000/510,260 * 10,060/510,260}\right)</math>
| |
| <math>+ \frac{500,000}{510,260} log_2\left(\frac{500,000/510,260}{510,000/510,260 * 500,200/510,260}\right)</math>
| |
| <math>= \frac{60}{510,260} log_2\left(\frac{60*510,260}{260 * 10,060}\right) + \frac{200}{510,260} log_2\left(\frac{200*510,260}{260 * 500,200}\right)</math>
| |
| <math>+ \frac{10,000}{510,260} log_2\left(\frac{10,000*510,260}{510,000 * 10,060}\right) + \frac{500,000}{510,260} log_2\left(\frac{500,000*510,260}{510,000 * 500,200}\right)</math>
| |
| | |
| = 0.000417322 - 0.000137100 - 0.000154725 + 0.000155158
| |
| | |
| = 0.000280655
| |
| | |
| Therefore, the mutual information between the cluster "commerce" and the term "tariff" is 0.000280655. We can create a label for the cluster "commerce" by calculating the mutual information of each term in the cluster, and selected the k terms with the highest MI value.
| |
| | |
| ===Chi-Squared Selection===
| |
| {{Main|Pearson's chi-squared test}}
| |
| The Pearson's chi-squared test can be used to calculate the probability that the occurrence of an event matches the initial expectations. In particular, it can be used to determine whether two events, A and B, are [[statistically independent]]. The value of the chi-squared statistic is:
| |
| | |
| <math>X^2 = \sum_{a \in A}{\sum_{b \in B}{\frac{(O_{a,b} - E_{a, b})^2}{E_{a, b}}}}</math>
| |
| | |
| where ''O<sub>a,b</sub>'' is the ''observed'' frequency of a and b co-occurring, and ''E<sub>a,b</sub>'' is the ''expected'' frequency of co-occurrence.
| |
| | |
| In the case of cluster labeling, the variable A is associated with membership in a cluster, and the variable B is associated with the presence of a term. Both variables can have values of 0 or 1, so the equation can be rewritten as follows:
| |
| | |
| <math>X^2 = \sum_{a \in {0,1}}{\sum_{b \in {0,1}}{\frac{(O_{a,b} - E_{a, b})^2}{E_{a, b}}}}</math>
| |
| | |
| For example, ''O<sub>1,0</sub>'' is the observed number of documents that are in a particular cluster but don't contain a certain term, and ''E<sub>1,0</sub>'' is the expected number of documents that are in a particular cluster but don't contain a certain term.
| |
| Our initial assumption is that the two events are independent, so the expected probabilities of co-occurrence can be calculated by multiplying individual probabilities:<ref>Manning, Christopher D., Prabhakar Raghavan, and Hinrich Schutze. ''Introduction to Information Retrieval''. Cambridge: Cambridge UP, 2008. ''Chi2 Feature Selection''. Stanford Natural Language Processing Group. Web. 25 Nov. 2009. <http://nlp.stanford.edu/IR-book/html/htmledition/feature-selectionchi2-feature-selection-1.html>.</ref>
| |
| | |
| ''E<sub>1,0</sub> = N * P(C = 1) * P(T = 0)''
| |
| | |
| where N is the total number of documents in the collection.
| |
| | |
| ====Example====
| |
| Using the same data for the mutual information example, we can calculate the expected probabilities and plug them into the equation to calculate the chi-squared statistic:
| |
| {| class="wikitable"
| |
| |-
| |
| !
| |
| ! Documents in "commerce"
| |
| ! Documents not in "commerce"
| |
| |-
| |
| | Documents containing "tariff"
| |
| | 60
| |
| | 10,000
| |
| |-
| |
| | Documents not containing "tariff"
| |
| | 200
| |
| | 500,000
| |
| |}
| |
| | |
| P (C = 1) = (# of documents in the cluster) / (total number of documents) = (60 + 200) / 510,260 = 260/510,260
| |
| | |
| P (C = 0) = (# of documents not in the cluster) / (total number of documents) = (10,000 + 500,000) / 510,260 = 510,000/510,260
| |
| | |
| P (T = 1) = (# of documents containing the term) / (total number of documents) = (60 + 10,000) / 510,260 = 10,060/510,260
| |
| | |
| P (T = 0) = (# of documents not containing the term) / (total number of documents) = (200 + 500,000) / 510,260 = 500,200/510,260
| |
| | |
| <math>E_{0, 0} = 510,260 * \frac {510,000}{510,260} * \frac {500,200}{510,260}</math>
| |
| | |
| = 499,945
| |
| | |
| <math>E_{0,1}= 510,260 * \frac {510,000}{510,260} * \frac {10.060}{510,260}</math>
| |
| | |
| = 10,055
| |
| | |
| <math>E_{1,0}= 510,260 * \frac {260}{510,260} * \frac {500,200}{510,260}</math>
| |
| | |
| = 254.9
| |
| | |
| <math>E_{1,1}= 510,260 * \frac {260}{510,260} * \frac {10,060}{510,260}</math>
| |
| | |
| = 5.13
| |
| | |
| <math>X^2 = \frac{(500,000 - 499,945)^2}{499,945} + \frac{(10,000 - 10,055)^2}{10,055} + \frac{(200 - 254.9)^2}{254.9} + \frac{(60 - 5.13)^2}{5.13}</math> | |
| | |
| = 599
| |
| | |
| Since each variable can have two values, the number of [[degrees of freedom (statistics)|degrees of freedom]] is (2 - 1)(2 - 1) = 1. The [[chi-squared distribution]] for one degree of freedom states that the probability of observing a value greater than 10.83 is less than 0.001. Therefore, we can reject the [[null hypothesis]], which states that the two events are independent. Since the term "tariff" and the cluster "commerce" are dependent, we can assume that the term is a good label for the cluster.
| |
| | |
| ==Cluster-Internal Labeling==
| |
| Cluster-internal labeling selects labels that only depend on the contents of the cluster of interest. No comparison is made with the other clusters.
| |
| Cluster-internal labeling can use a variety of methods, such as finding terms that occur frequently in the centroid or finding the document that lies closest to the centroid.
| |
| | |
| ===Centroid Labels===
| |
| {{Main|Vector space model}}
| |
| A frequently used model in the field of [[information retrieval]] is the vector space model, which represents documents as vectors. The entries in the vector correspond to terms in the [[vocabulary]]. Binary vectors have a value of 1 if the term is present within a particular document and 0 if it is absent. Many vectors make use of weights that reflect the importance of a term in a document, and/or the importance of the term in a document collection. For a particular cluster of documents, we can calculate the [[centroid]] by finding the [[arithmetic mean]] of all the document vectors. If an entry in the centroid vector has a high value, then the corresponding term occurs frequently within the cluster. These terms can be used as a label for the cluster.
| |
| One downside to using centroid labeling is that it can pick up words like "place" and "word" that have a high frequency in written text, but have little relevance to the contents of the particular cluster.
| |
| | |
| ===Title Labels===
| |
| An alternative to centroid labeling is title labeling. Here, we find the document within the cluster that has the smallest [[Euclidean distance]] to the centroid, and use its title as a label for the cluster. One advantage to using document titles is that they provide additional information that would not be present in a list of terms. However, they also have the potential to mislead the user, since one document might not be representative of the entire cluster.
| |
| | |
| ===External knowledge labels===
| |
| Cluster labeling can be done indirectly using external knowledge such as pre-categorized knowledge such as the one of Wikipedia.<ref>David Carmel, Haggai Roitman, Naama Zwerdling. [http://portal.acm.org/citation.cfm?doid=1571941.1571967 Enhancing cluster labeling using wikipedia.] SIGIR 2009: 139-146</ref> In such methods, a set of important cluster text features are first extracted from the cluster documents. These features then can be used to retrieve the (weighted) K-nearest categorized documents from which candidates for cluster labels can be extracted. The final step involves the ranking of such candidates. Suitable methods are such that are based on a voting or a fusion process which is determined using the set of categorized documents and the original cluster features.
| |
| | |
| ==External links==
| |
| * [http://nlp.stanford.edu/IR-book/html/htmledition/hierarchical-clustering-1.html Hierarchical Clustering]
| |
| * [http://erulemaking.ucsur.pitt.edu/doc/papers/dgo06-labeling.pdf Automatically Labeling Hierarchical Clusters]
| |
| | |
| ==References==
| |
| <references/>
| |
| | |
| {{DEFAULTSORT:Cluster Labeling}}
| |
| [[Category:Information retrieval]]
| |
e - Shop Word - Press is a excellent cart for your on the web shopping organization. Affilo - Theme is the guaranteed mixing of wordpress theme that Mark Ling use for his internet marketing career. If you have any sort of questions concerning where and how you can make use of wordpress backup, you can contact us at our own web-site. Step-4 Testing: It is the foremost important of your Plugin development process. Dead links are listed out simply because it will negatively have an influence on the website's search engine rating. In the most current edition you can customize your retailer layout and display hues and fonts similar to your site or blog.
You just download ready made templates to a separate directory and then choose a favorite one in the admin panel. While direct advertising is limited to few spots in your site and tied to fixed monthly payment by the advertisers, affiliate marketing can give you unlimited income as long as you can convert your traffic to sales. Which is perfect for building a mobile site for business use. You can add new functionalities and edit the existing ones to suit your changing business needs. Now a days it has since evolved into a fully capable CMS platform which make it, the best platform in the world for performing online business.
Just ensure that you hire experienced Word - Press CMS developer who is experienced enough to perform the task of Word - Press customization to get optimum benefits of Word - Press CMS. But if you are not willing to choose cost to the detriment of quality, originality and higher returns, then go for a self-hosted wordpress blog and increase the presence of your business in this new digital age. Setting Up Your Business Online Using Free Wordpress Websites. To turn the Word - Press Plugin on, click Activate on the far right side of the list. Socrates: (link to ) Originally developed for affiliate marketers, I've used this theme again and again to develop full-fledged web sites that include static pages, squeeze pages, and a blog.
You can add keywords but it is best to leave this alone. Russell HR Consulting provides expert knowledge in the practical application of employment law as well as providing employment law training and HR support services. Enterprise, when they plan to hire Word - Press developer resources still PHP, My - SQL and watch with great expertise in codebase. Fast Content Update - It's easy to edit or add posts with free Wordpress websites. The Pakistani culture is in demand of a main surgical treatment.
Under Settings —> Reading, determine if posts or a static page will be your home page, and if your home page is a static page, what page will contain blog posts. Here's a list of some exciting Word - Press features that have created waves in the web development industry:. It can be concluded that white label SEO comprise of a third party who resells a contract involving IT expert or consultant, SEO professional and end user. with posts or testimonials updated as they are uploaded to a particular section of the website. However, if you're just starting out your blog site or business site, you can still search for an ideal theme for it without breaking your bank account.