Patent Grant
8312049
The present invention relates to organizing newsgroups via clustering. More particularly, the invention relates to a system and method which utilizes cross-posts to cluster newsgroups.
Self-organized structures (e.g., networks) have gained tremendous popularity over recent years, and such networks poses a significant challenge as to understanding respective constructs thereof. The networks (e.g., the Internet, world wide web, various social networks . . . ) can be very different from each other, but they all typically share a commonality in that their structures are not engineered, but rather are a result of dynamic non-Markovian processes of individual decisions. Understanding the respective structures of these networks can facilitate enabling modeling of as well as searching the networks effectively—the greatest success thus far has been in searching the world wide web, which has now become a primary repository of information and/or misinformation.
Internet newsgroups (e.g., Usenet) is another large network intermediate to the Internet and social networks. Usenet, the network of topic-oriented newsgroups on the Internet, comprises tens of thousands of newsgroups and hundreds of millions of postings by millions of authors throughout the world. Some basic workings of Usenet are as follows: each of over fifty thousand newsgroups has a unique name, with the name grouped into trees—some of the more common roots of these trees include alt., biz., and rec. Within a newsgroup, the messages are organized as threads. Each message is typically generated by a single author; individuals may author more than one message along a thread. Each thread originates in a single message with a subject heading usually reflecting the content of the message; later messages in the thread, of which there can be thousands, are posted as responses to the original message. Any message along a thread can be cross-posted (e.g., by its author) to any number of additional newsgroups.
Newsgroups facilitate transfer of information and ideas on topics specific to each individual newsgroup. For example, a user interested in gaining travel tips prior to traveling to a country (e.g., Africa) can enter a newsgroup specifically created and operated to discuss such information (rec.travel.africa). There are currently thousands of newsgroups that cover nearly every topic imaginable. When properly utilized, newsgroups can provide for a user to submit highly specific queries and obtain answers thereto, wherein oftentimes such answers are only available from a very limited set of disparate individuals or entities.
Presently, however, it is difficult to quickly locate a particular newsgroup relevant to a specific topic of desire. Even if a newsgroup, that appears to be on-point to a topic, is located by a user, such a newsgroup may not be currently active or may comprise information not germane to the particular topic of interest—thus, not really being on-point. Internet search engines are typically employed to locate newsgroup(s) which relate to information or discussion for a particular topic of interest. However, oftentimes, the search engines return newsgroups with low precision (e.g., non-relevant), due in part to such engines being primarily limited to employment of textual-based searching schemes. For example, if a user desires to join a newsgroup that discusses travel to Africa and enters keywords “travel” and “Africa” to a search engine, he/she may receive an exact desired newsgroup, but the search engine may just as likely return newsgroups focusing on irrelevant respective content (e.g., sports, politics . . . ).
Undesirable results can occur because a conventional search engine typically only ranks the newsgroups based on some pre-determined criteria. The search engine may rank newsgroups according to number of times that both “Africa” and “travel” appear in discussion of the newsgroups. Therefore, if the discussion in a sports newsgroup and a political newsgroup is focused on a sports figure traveling to Africa and/or a political figure traveling to Africa, respectively, they will be ranked highly when the search engine returns newsgroups from the search query, even though the user has no desire to discuss either sports or politics.
In view of at least the above, there exists a strong need in the art for a system and/or methodology to facilitate improved searching and organization of newsgroups.
The following presents a simplified summary of the invention in order to provide a basis understanding of some aspects of the invention. This summary is not an extensive overview of the invention. It is intended to neither identify key or critical elements of the invention nor delineate the scope of the invention. Its sole purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later.
The subject invention facilitates locating and/or organizing newsgroups, and overcomes many of the aforementioned deficiencies related to conventional systems for searching newsgroup-related information. The invention takes advantage of a realization by inventors of the subject invention that cross-postings between newsgroups can be indicative of degree of relatedness among respective newsgroups. As such, the subject invention provides for constructing a weighted graph of newsgroups utilizing cross-posting information. Such a weighted graph can be further refined via pre-processing, post-processing, segmenting schemes, and/or filtering schemes. The weighted graph can facilitate locating and/or understanding respective relationships of newsgroups.
Newsgroup(s) cross-postings are in a sense analogs of hyperlinks of a webpage, which reflect a webpage author's judgment that additional web pages may be of interest to readership of an original webpage. Cross-posts are employed via the subject invention as an indication for similarity between topics discussed in newsgroups or an interaction between related communities. A graph in accordance with the subject invention that is defined in part via cross-posts provides a mechanism that greatly facilitates understanding newsgroup(s) structure as well as exploiting such structure to enhance navigation thereof.
One particular aspect of the invention provides for creation of the weighted graph. Newsgroup data is received, which includes but is not limited to data indicating identification of newsgroups, number of postings within newsgroups, and number of cross-posting between newsgroups. The data can be obtained from a database, hard drive, Internet search engine, etc. Cross-postings refer to the posting(s) of a same message to more than one newsgroup (e.g., similar to a group email). The number of cross-postings between newsgroups indicates the degree of relatedness between newsgroups, and thus provides structural constructs for generating a weighted graph. The newsgroup data can be organized in matrices and arrays, wherein an array can indicate the number of newsgroups available on a network at a given time, and a second symmetrical matrix can indicate the number of cross-postings between newsgroups. Matrices and arrays can comprise newsgroups and newsgroup related data in any arbitrary but fixed order, thus facilitating efficient creation of a weighted graph.
A weighted graph is created based upon the newsgroups cross-postings, wherein the newsgroups are represented as vertices and the cross-posts are represented as edges of the weighted graph. The graph can facilitate analysis of newsgroup clusters, as a user can visually determine from the graph which newsgroups are most related. Moreover, the weighted graph can be employed via a search engine to provide meaningful results in connection with a query for newsgroup(s) related to a particular item of interest.
With respect to another aspect of the present invention, the weighted graph can be partitioned into small clusters, wherein a cluster contains a plurality of newsgroups which are highly related to one another. In other words, the newsgroups of a first cluster can be related by a number of internal cross-postings among them that is greater than the number of cross-posting between newsgroups of the first cluster and newsgroups of another cluster. Thus, the degree of relatedness of newsgroups within the first cluster is higher than that for newsgroups outside the first cluster. The present invention thus overcomes some of the deficiencies of prior art systems that do not provide for grouping of newsgroups.
The weighted graph can be clustered via a spectral clustering algorithm, for example. The spectral clustering algorithm can be applied recursively to the weighted graph, to provide for modifying results without executing the algorithm over an entire data set repeatedly. Also, segmented clusters can be merged in accordance with another particular aspect of the present invention as part of a post-processing aspect of the invention. If weights of edges between two or more clusters are sufficiently large, the clusters can be merged into a larger cluster so as to correct for undesired segmentation resulting from the approximation-based nature of the clustering algorithm.
With respect to another particular aspect of the invention, the data can be filtered prior to being graphed, so as to facilitate removing irrelevant data prior to constructing the weighted graph thereby optimizing utilization of processing bandwidth as well as providing for a more useful weighted graph in that better respective data is employed to build the graph. Newsgroups can be filtered when they do not contain a threshold of postings, as newsgroups with low postings are indicative of a low level of interest in the topic. Furthermore, edges of low weight can be pared, as a small number of cross-postings between newsgroups indicate that the newsgroups are not related. Moreover, isolated newsgroups or subsets of newsgroups can be excluded from the weighted graph, as isolation would indicate a very specific interest or a low-interest newsgroup.
The results from the weighted cluster graph can either be returned as final output or can be used as input to other processing components such as search engines, newsgroup browsers or email programs. For example, if an email program were used to post a response to a newsgroup message, the program could suggest to the user that the message be cross-posted to all newsgroups in a cluster containing the original newsgroup. Or, as another example, a search engine could use the weighted cluster graph (e.g., represented as a matrix, database, vectors, . . . ) in connection with executing a newsgroup search and providing results in response thereto.
Moreover, final output results from the weighted cluster graph can be output to a printer, monitor, floppy disk, CD, DVD, hard drive, etc. The results can be displayed graphically or textually. For example, if an Internet search engine were utilized to locate a newsgroup, a weighted cluster graph created in accordance with the present invention could be displayed to the user to facilitate user analysis of the clusters.
To the accomplishment of the foregoing and related ends, the invention then comprises the features hereinafter fully described and particularly pointed out in the claims. The following description and the annexed drawings set forth in detail certain illustrative aspects of the invention. These aspects are indicative, however, of but a few of the various ways in which the principles of the invention may be employed and the present invention is intended to include all such aspects and their equivalents. Other objects, advantages and novel features of the invention will become apparent from the following detailed description of the invention when considered in conjunction with the drawings.
The present invention is now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It may be evident, however, that the present invention may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing the present invention.
As used in this application, the term “computer component” is intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a computer component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. By way of illustration, both an application running on a server and the server can be a computer component. One or more computer components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers.
Referring now to the drawings,
According to one particular aspect of the present invention, the data reception component 52 can receive matrices and arrays comprising newsgroup data. Alternatively, the data reception component 52 can facilitate organization of data into matrices, thus effecting efficient computation of data. For example, newsgroups can be organized in an array or matrix N, wherein the number of rows or columns of N corresponds to the number of newsgroups contained within a network at a given time. The entries of N can indicate the number of postings within a newsgroup, number of authors posting to a newsgroup, etc. corresponding to such entry N. A symmetrical matrix P can be created, wherein the number of rows and columns of P is equivalent to the number of newsgroups of N. Matrix P can contain data relating to the number of cross-postings between newsgroups.
In particular, let n=n(t) be a number of newsgroups at time t. The cross-post matrix P=P(t, δt) is a symmetric n×n matrix, with each row representing a different newsgroup, in some arbitrary but fixed order. Non-negative integer components Pij of P represent a total number of cross-posts between all messages on newsgroup i and newsgroups j over time interval δt before time t. This information can be represented either as a multigraph (i.e., a graph in which there may be multiple edges between vertices) or as a weighted graph. In the multigraph representation, the vertices represent distinct newsgroups and the number of edges between i and j represent the total number of cross-postings between newsgroup i and newsgroup j. In the weighted graph representation, the vertices again represent distinct newsgroups and the weight of the edge between i and j represents the number of cross-postings between newsgroup i and newsgroup j.
In accordance with one aspect of the present invention, the data reception component 52 can be effectively coupled to a data store 54 within the system 50—the data store 54 can be employed to retain data previously received by the data reception component 52 as well as various algorithms (e.g., clustering algorithms), historical data, profiles, contextual data, content-based data, and any other suitable information for carrying out the present invention. The data store 54 can be any storage component capable of retaining computer data, such as a CD, DVD, hard drive, floppy disk, server database, etc. The data store 54 can be further employed to store any suitable data type (e.g., XML, HTML, data tables . . . ), and any other data type or structure can be retained in the data store 54 in accordance with one aspect of present invention.
At least a subset of data received by the data reception component 52 and/or held in the data store 54 can be relayed to and/or from a graphing engine 56. The graphing engine 56 facilitates generation of a weighted graph relating to newsgroup data by utilizing data received by the data reception component 52 and/or contained within the data store 54, such that the graph depicts relatedness of two or more newsgroups. Specifically, the weighted graph is created based upon cross-postings between a plurality of newsgroups, as cross-postings between newsgroups are indicative of relatedness. For example, two newsgroups entitled “Travel to Africa” and “Travel by Air”, respectively, will be related, as cross-posts regarding travel by air to Africa will be shared between both newsgroups. As cross-posting between two or more newsgroups increases numerically (or a percentage of cross-postings to total posts increases), relatedness between such newsgroups likewise increases. The graphing engine 56 can depict degrees of relatedness by representing newsgroups as vertices and cross-posts as edges of a weighted graph. An exemplary weighted graph 58 in accordance with one aspect of the present invention is illustrated in
Turning briefly to
Increasing degrees of relatedness between vertices can be represented by increasing weights of edges connecting such vertices (e.g., thicker edges between vertices indicate more cross-posting between newsgroups). For example, more cross-posting exists between newsgroups 108 and 110 than between 108 and 112, visually illustrated by a thicker edge 114 between the newsgroups 108 and 110 as compared to edge 116 that connects newsgroups 108 and 112. No cross-posting exists between newsgroups 118 and 120, as no edge exists between such newsgroups.
The weighted graph 106 can be conveyed via any suitable manner (e.g., graphically displayed to a user, mathematically represented to a computing system, . . . ), or utilized for further modification and/or analysis thereof (e.g., in connection with a newsgroup search). It is to be understood that the illustrated weighted graph 106 is merely exemplary, and is not meant to be limiting in any manner as other embodiments of weighted graphs as well as other means for representing relatedness of newsgroups via utilization of cross-posting information are contemplated and intended to fall within the scope of the hereto appended claims. For example, an edge line can exist for each cross-posting (rather than increasing edge weights as cross-postings between newsgroups increase). Furthermore, vertices can be any suitable shape, and can be weighted to indicate a number of postings within each newsgroup (e.g., vertices representing newsgroups can be larger and/or varying in color as the number of newsgroup postings increase or decrease). Moreover, weighted graph 106 can be a 3-dimensional graph, resulting in a more visually appealing graph during instances that a large number of newsgroups are considered. The weighted graph can also be represented numerically as well if desired.
Returning to
In accordance with one aspect of the present invention, the system 50 can be part of a search engine that employs the weighted graph in connection with searching newsgroups. For example, a search for newsgroups discussing a particular topic can be initialized by entering search terms into the system 50. The data reception component 52 can facilitate deliverance of newsgroup data to the graphing engine 56 and/or data store 54, thereby in connection with generating a weighted graph having newsgroups as vertices and cross-posts as edges. The graphing engine 56 can be employed to relay such weighted graph to the data store 54 for storing and/or displaying. Furthermore, the graphing engine 56 can relay the weighted graph to the segmenting component 60, which facilitates segmentation of the graph into clusters of highly related newsgroups. The system 50 can return clusters of relevant newsgroups to a user as part of results to a query for newsgroups corresponding to topic(s) of interest. Furthermore, the system 50 can be employed in accordance with searching newsgroups relating to sale of goods and/or services (e.g., auction-related newsgroups, newsgroups containing classified listings, . . . ). For example, cross-postings between newsgroups employed to facilitate sale of goods and/or services can be indicative of type of goods and/or services being sold. The system 50 can facilitate clustering of such newsgroups, thereby enabling a buyer and/or seller of goods and/or services to quickly access most germane newsgroups.
In accordance with another aspect of the present invention, the data delivered to and/or relayed from the data reception component 52 can be filtered to decrease size of a data set relating to newsgroups, thus reducing computation time required to create a weighted graph. The data set can be filtered by removing newsgroups with insufficient postings, removing newsgroups with insufficient cross-postings, through artificial intelligence means, or any other suitable means for pruning a superset of data to more relevant subset(s) that can be employed by the graphing engine 56 to generate a meaningful weighted graph.
The system 50 can employ various inference schemes and/or techniques in connection with filtering undesirable newsgroups. As used herein, the term “inference” refers generally to the process of reasoning about or inferring states of the system, environment, and/or user from a set of observations as captured via events and/or data. Inference can be employed to identify a specific context or action, or can generate a probability distribution over states, for example. The inference can be probabilistic—that is, the computation of a probability distribution over states of interest based on a consideration of data and events. Inference can also refer to techniques employed for composing higher-level events from a set of events and/or data. Such inference results in the construction of new events or actions from a set of observed events and/or stored event data, whether or not the events are correlated in close temporal proximity, and whether the events and data come from one or several event and data sources. Various classification schemes and/or systems (e.g., support vector machines, neural networks, expert systems, Bayesian belief networks, fuzzy logic, data fusion engines . . . ) can be employed in connection with performing automatic and/or inferred action in connection with the subject invention.
Turning now to
The discussing of the methodology is supplemented with source code in accordance with one particular implementation of the present method so as to facilitate understanding of the subject method. Any suitable program/language can be employed in connection with the invention—accordingly, the invention is not limited to any particular program and/or language (e.g., Matlab). It is to be appreciated that the method of
N is an n×3 array, wherein each row of N contains three numbers indicating identification of a newsgroup, number of posts to each newsgroup over a certain period of time, and number of authors who have posted a message to such newsgroup over such period of time.
E is an m×3 array. Each row of E contains identification of two newsgroups and number of cross-posts there between.
id and name are two n×1 arrays. Entry i of array id is the identification of a newsgroup, whose name is stored in entry i of array name.
At 302, newsgroup data is received (e.g., via a data storage device such as a server database, compact disk, DVD, floppy disk, hard drive, etc., by manually entering data, or any other method of retrieving and/or receiving data). The newsgroup data can include but is not limited to the number of newsgroups, a way to identify each newsgroup, number of posts in each newsgroup, number of cross-posts in each newsgroup, data identifying all newsgroups which contain each cross-post, and number of authors posting to each newsgroup. In the exemplary source code, all necessary data has been previously retrieved and stored in arrays N, E, id, and name.
At 304 the newsgroup data retrieved is filtered to exclude newsgroups and corresponding data that contain too few posts, posts from too few authors, or are irrelevant to the user given a particular user's identity, state, and context. Inference techniques can be utilized to determine whether newsgroup data received is relevant given a particular user. Furthermore, edges are removed if edge weight does not meet a certain threshold (e.g., if there are not enough cross-posts between two newsgroups, corresponding edges connecting vertices representing such two newsgroups are eliminated). It is to be understood that although the exemplary code filters newsgroups and eliminates edges adjacently, the present invention can be modified to filter undesirable newsgroups prior to reception of data and to pare insufficient edges after creation of a weighted graph. In the exemplary code in accordance with one aspect of the present invention, the newsgroups are filtered which contain less than 100 posts over a period of one year, and edges between vertices are pared when less than five cross-posts are shared between corresponding news groups over a period of one year.
At 306 a weighted graph is created, with newsgroups represented as vertices and cross-posts represented as edges. Weights of edges connecting vertices are indicative of the number of cross-posts between newsgroups that such vertices represent. Moreover, a weighted graph can include weighted vertices, wherein such vertex weights indicate the number of postings within a newsgroup (e.g., newsgroups with several hundred postings will be represented by larger vertices when compared to vertices representing newsgroups with fewer postings). The source code in accordance with one aspect of the present invention simulates generation of a graph by constructing an adjacency matrix of the graph, thus indicating the weight of edges between all vertices.
A1=sparse(E(:,1), E(:,2), E(:,3));
A=A1 (giantcomp,giantcomp);
Then at 308 a weighted graph is recursively segmented into clusters of desirable size.
At 310, two or more resulting clusters after segmentation are merged if such clusters are substantially related to one another. Relatedness of two or more clusters can be determined by reviewing weighted edges between two clusters. Merging two separate clusters when summation of weights of edges between a first cluster and a second cluster is greater than half of the summation of weights of edges adjacent to such first cluster is one exemplary manner facilitating determination of instances that merging two clusters is appropriate. The method can merge two clusters A and B when
where α(A, B)=ΣiεAΣjεBαij is the partition between A and B, αA=ΣiεAΣjαij is the total number of edges beginning in A (regardless of whether they end inside or outside of A), and αij are the entries of the adjacency matrix constructed at 306. The below code illustrates one particular implementation for such merging.
Then, at 312 a weighted graph or substantially equivalent results are output to a user or relayed to a data store. Graphical results can be output to a monitor, printer, disk, hard drive, etc. Furthermore a resulting weighted cluster graph can be output to a user graphically or textually. In an instance of a search engine utilizing method 300, output newsgroup clusters can be displayed in hyperlink format, thus directing a user to a selected newsgroup or cluster of newsgroups. The method can store output results in a file on a hard drive, floppy disk, DVD, such as for example via the code below. Alternatively, the method can use the results as input to another processing component, such as a search engine, newsgroup browser, email program, etc.
Now turning to
At 406 a weighted graph is generated—it is to be appreciated that formatting data in such a manner to substantially create a weighted graph is sufficient. For example, one could choose to create an adjacency matrix and still fall within the scope of the invention. At 408, size of a generated weighted graph is compared with a threshold. If the size of a weighted graph is greater than such threshold, the weighted graph can be subjected to further filtering and paring to facilitate reduction of size of such weighted graph to reach a suitable size. It should be noted that filtering and paring an extensive amount of newsgroup data could possibly bias remaining data during weighted graph segmentation. Moreover, the present invention can be employed to determine whether newsgroup data is to be additionally filtered and pared prior to generation of a weighted graph. For example, directly after initially filtering newsgroup data, the size of remaining newsgroup data following filtering can be compared with a threshold. If the size of the remaining subset of newsgroup data is greater than such threshold, more filtering can be done prior to generation of a weighted graph.
At 410, a weighted graph is recursively segmented into clusters of suitable size. Segmenting a weighted graph is discussed in greater detail infra. At 412, clusters created via segmentation are reviewed to determine whether merging two cluster would be beneficial. At 414, remaining clusters are reviewed to determine if cluster size and quality are acceptable. If one or more clusters are not of a sufficient size, they can be merged into another cluster with which they have substantial relation based upon the number of cross-postings between such two clusters. At 416, results of a weighted graph or substantial equivalent thereof are output.
Turning now to
Regarding segmentation of a weighted graph (for example, exemplary weighted cluster graph 508), such graphs can ideally be partitioned by segmenting between vertices to substantially minimize edges between clusters while concurrently creating clusters of desirable size. Defining different tradeoffs between these two objectives can result in different optimization problems. In accordance with one particular aspect of the present invention, clusters of substantially equal size (e.g., clusters comprising an approximately equal number of newsgroups) should result from partitioning a weighted graph. For sake of brevity, the following algorithm functions are defined to partition each segment into two new segments, S and
Several traditional objective functions exist for determining where to segment between vertices in a weighted graph. A few examples are:
where A=(αij) is the adjacency matrix, cut=ΣiεS,jε
Spectral analysis is a method mainly based on analysis of eigenvectors of a normalization of the adjacency matrix of the graph. Determining a Laplacian of a matrix is one common method for normalizing such matrix (e.g., the Laplacian L=D−A, where D is a diagonal matrix with dii=Σjαij). A Fiedler vector v of adjacency matrix A is an eigenvector corresponding to a second smallest eigenvalue of the Laplacian L, which can be determined using equation Lv=λv, where λ is a second smallest eigenvalue of L. In practice, however, determining a vector corresponding to a second smallest solution of a generalized eigenvalue equation Lv=λDv performs better than determining the Fiedler vector. This vector v is equal to the eigenvector corresponding to the second largest eigenvalue of D−1/2AD−1/2 multiplied to D−1/2.
Prior to performing any segmentation, a segmentation value s is defined and vertices of a weighted graph are divided into at least two segments based upon whether or not values assigned to each segment by vector v are greater than the segmentation value s. There are several different manners to define the segmenting value s, including equating s to the median of values assigned to the vertices by v; defining s as being 0; assigning s as the value in the larges gap in the sorted list of Fiedler vector values; and/or utilizing one of the objective functions to find a value of s.
In order to partition a weighted graph into more than two clusters, the segmenting component 60 (shown in
Now turning briefly to
At 608, a segmentation position on the weighted graph is proposed. The minimum Mcut ratio of the Q intervals can be determined by calculating an Mcut ratio at endpoints of each of the Q intervals. After the minimum Mcut value is found from such Q intervals, then the Mcut is calculated at P values above and below the endpoint of the minimizing interval. P should be a value facilitating attainment of a desired accuracy and reliability in determining an appropriate segmentation of a weighted graph, while not being so large as to require unreasonable computation. Of these 2P eigenvector entries, the eigenvector entry where the Mcut ratio is lowest will be the proposed location of segmentation on the graph.
At 610, the Mcut ratio of the proposed segmenting position is compared to a pre-defined threshold t. If the Mcut ratio is less than t, at 612 the weighted graph is segmented at the proposed segmenting position, and acts 602-610 are recursively applied to both created segments. If the Mcut ratio is greater than threshold t, no segmentation occurs, and at 614 the algorithm determines whether any segments exist on which acts 602-610 have not been executed. If such acts 602-610 have not been executed on a segment, acts 602-610 are again applied to such segment. If all existing segments are of a desired size and/or quality, method 600 completes at 616.
The threshold t should be chosen to create final clusters of a desired size and/or quality, wherein quality is determined by some measure of relatedness within each segment (e.g., a sufficient degree of cross-posting exists between newsgroups within each cluster). In one particular aspect of the present invention, t can be defined as an increasing function of the number of nodes in a segment on which method 600 is currently being executed. For example,
are two examples wherein a number of nodes in a segment can be utilized to define t. Alternatively, t can be defined as any number between 0 and 1, depending on a how many newsgroups in a final cluster are desirable.
is displayed.
Returning to
Turning briefly to
Regarding
Returning now to
To facilitate further reduction of size in a data set, the paring component 406 can eliminate cross-posts between newsgroups (and thus edges between vertices) when the number of cross-posts between two or more newsgroups is below a pre-defined threshold. Alternatively, the threshold can vary with the size of a data set contained in data reception component 902 by defining the threshold as an increasing function of number(s) of newsgroups contained in the data store 910. The paring component 906 can further be employed to eliminate newsgroups that do not contain a sufficient amount of cross-posting with other newsgroups. Thus, size of a cluster graph can be effectively reduced. Moreover, when the filter component 904 and/or the paring component 906 are used in conjunction with a segmenting component (not shown), the computational cost of segmenting a weighted graph into small clusters can be greatly reduced.
Turning now to
The paring component 1124 removed from the data set edges between vertices 1136 and 1138, 1140 and 1142, and 1142 and 1144 respectively. An insubstantial number of cross-posts between a plurality of newsgroups is indicative of a minimal degree of relatedness between such newsgroups, which in turn insinuates that elimination of edges representing such cross-posts is desirable.
Turning now to
where α(A,B)=ΣiεAΣjεBαij is the cut between A and B, and αA=ΣiεAΣjαij is the total number of edges beginning in A (regardless of whether they end inside or outside of A). The post-processing component 1208, however, is capable of merging two or more segmented clusters using any algorithm or threshold of relatedness.
Turning now to
Turning now to
Now turning to
Regarding
Turning now to
Turning to
Regarding
In order to provide additional context for various aspects of the present invention,
With reference to
The system bus 1718 can be any of several types of bus structure(s) including the memory bus or memory controller, a peripheral bus or external bus, and/or a local bus using any variety of available bus architectures including, but not limited to, an 8-bit bus, Industrial Standard Architecture (ISA), Micro-Channel Architecture (MSA), Extended ISA (EISA), Intelligent Drive Electronics (IDE), VESA Local Bus (VLB), Peripheral Component Interconnect (PCI), Universal Serial Bus (USB), Advanced Graphics Port (AGP), Personal Computer Memory Card International Association bus (PCMCIA), and Small Computer Systems Interface (SCSI).
The system memory 1716 includes volatile memory 1720 and nonvolatile memory 1722. The basic input/output system (BIOS), containing the basic routines to transfer information between elements within the computer 1712, such as during start-up, is stored in nonvolatile memory 1722. By way of illustration, and not limitation, nonvolatile memory 1722 can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory 1720 includes random access memory (RAM), which acts as external cache memory. By way of illustration and not limitation, RAM is available in many forms such as synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM).
Computer 1712 also includes removable/nonremovable, volatile/nonvolatile computer storage media.
It is to be appreciated that
A user enters commands or information into the computer 1712 through input device(s) 1736. Input devices 1736 include, but are not limited to, a pointing device such as a mouse, trackball, stylus, touch pad, keyboard, microphone, joystick, game pad, satellite dish, scanner, TV tuner card, digital camera, digital video camera, web camera, and the like. These and other input devices connect to the processing unit 1714 through the system bus 1718 via interface port(s) 1738. Interface port(s) 1738 include, for example, a serial port, a parallel port, a game port, and a universal serial bus (USB). Output device(s) 1740 use some of the same type of ports as input device(s) 1736. Thus, for example, a USB port may be used to provide input to computer 1712, and to output information from computer 1712 to an output device 1740. Output adapter 1742 is provided to illustrate that there are some output devices 1740 like monitors, speakers, and printers among other output devices 1740 that require special adapters. The output adapters 1742 include, by way of illustration and not limitation, video and sound cards that provide a means of connection between the output device 1740 and the system bus 1718. It should be noted that other devices and/or systems of devices provide both input and output capabilities such as remote computer(s) 1744.
Computer 1712 can operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1744. The remote computer(s) 1744 can be a personal computer, a server, a router, a network PC, a workstation, a microprocessor based appliance, a peer device or other common network node and the like, and typically includes many or all of the elements described relative to computer 1712. For purposes of brevity, only a memory storage device 1746 is illustrated with remote computer(s) 1744. Remote computer(s) 1744 is logically connected to computer 1712 through a network interface 1748 and then physically connected via communication connection 1750. Network interface 1748 encompasses communication networks such as local-area networks (LAN) and wide-area networks (WAN). LAN technologies include Fiber Distributed Data Interface (FDDI), Copper Distributed Data Interface (CDDI), Ethernet/IEEE 802.3, Token Ring/IEEE 802.5 and the like. WAN technologies include, but are not limited to, point-to-point links, circuit switching networks like Integrated Services Digital Networks (ISDN) and variations thereon, packet switching networks, and Digital Subscriber Lines (DSL).
Communication connection(s) 1750 refers to the hardware/software employed to connect the network interface 1748 to the bus 1718. While communication connection 1550 is shown for illustrative clarity inside computer 1712, it can also be external to computer 1712. The hardware/software necessary for connection to the network interface 1748 includes, for exemplary purposes only, internal and external technologies such as, modems including regular telephone grade modems, cable modems and DSL modems, ISDN adapters, and Ethernet cards. What has been described above includes examples of the present invention. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the present invention, but one of ordinary skill in the art may recognize that many further combinations and permutations of the present invention are possible. Accordingly, the present invention is intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims. Furthermore, to the extent that the term “includes” is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term “comprising” as “comprising” is interpreted when employed as a transitional word in a claim.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5511158 | Sims | Apr 1996 | A |
| 5819269 | Uomini | Oct 1998 | A |
| 5923846 | Gage et al. | Jul 1999 | A |
| 6064971 | Hartnett | May 2000 | A |
| 6341364 | Wever et al. | Jan 2002 | B1 |
| 6438580 | Mears et al. | Aug 2002 | B1 |
| 6477582 | Luo et al. | Nov 2002 | B1 |
| 6594673 | Smith et al. | Jul 2003 | B1 |
| 6604105 | Roth | Aug 2003 | B1 |
| 6636849 | Tang et al. | Oct 2003 | B1 |
| 6952700 | Modha et al. | Oct 2005 | B2 |
| Number | Date | Country |
|---|---|---|
| 11096134 | Apr 1999 | JP |
| Number | Date | Country | |
|---|---|---|---|
| 20040267686 A1 | Dec 2004 | US |
