Thematic clustering

Abstract

A data set is clustered into one or more initial clusters using a first term space. Initial themes for the initial clusters are determined. The first term space is reduced to create a reduced term space. At least a portion of the data set is reclustered into one or more baby clusters using the reduced term space. One or more singletons are reassigned to form one or more renovated clusters from the baby clusters. A renovated theme is determined for at least some of the renovated clusters. One or more of the renovated clusters and their respective themes are provided as output.

Description

BACKGROUND OF THE INVENTION

Information has traditionally been manually classified to aid potential readers in locating works of interest. For example, books are typically associated with categorization information (e.g., one or more Library of Congress classifications) and academic articles sometimes bear a list of keywords selected by their authors or editors. Unfortunately, while manual classification may be routinely performed for certain types of information such as books and academic papers, it is not performed (and may not be feasibly performed) for other types of information, such as the predominantly unstructured data found on the World Wide Web.

Attempts to automatically classify documents can also be problematic. For example, one technique for document classification is to designate as the topic of a given document the term occurring most frequently in that document. A problem with this approach is that in some cases, the most frequently occurring term in a document is not a meaningful description of the document itself. Another problem with this approach is that terms can have ambiguous meanings. For example, documents about Panthera onca, the British luxury car manufacturer, and the operating system might all be automatically classified using the term “Jaguar.” Unfortunately, a reader interested in one meaning of the term may wind up having to sift through documents pertaining to all of the other meanings as well.

BRIEF DESCRIPTION OF THE DRAWINGS

Various embodiments of the invention are disclosed in the following detailed description and the accompanying drawings.

FIG. 1 illustrates an example of an environment in which documents are clustered.

FIG. 2 illustrates an embodiment of a clustering system.

FIG. 3 illustrates an embodiment of a process for clustering documents.

FIG. 4A is a conceptual illustration of baby clusters and singletons.

FIG. 4B is a conceptual illustration of reclustering (after moving the singletons to the baby clusters).

FIG. 5 illustrates a representation of a data set comprising 25 documents that can be clustered into four clusters (with their topics) and five singletons.

FIG. 6 illustrates the data set after an embodiment of initial clustering has been performed.

FIG. 7 illustrates a graphical view of initial clustering of the data set.

FIG. 8 illustrates a graphical view of baby clusters and singletons.

FIG. 9A illustrates a graphical view of clusters after reassigning singletons is performed.

FIG. 9B illustrates documents in the data set as members of renovated clusters or as singletons after reclustering.

FIG. 9C illustrates example themes for each of the four clusters using the top shared terms of a given cluster.

DETAILED DESCRIPTION

The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and/or a processor, such as a processor configured to execute instructions stored on and/or provided by a memory coupled to the processor. In this specification, these implementations, or any other form that the invention may take, may be referred to as techniques. In general, the order of the steps of disclosed processes may be altered within the scope of the invention. Unless stated otherwise, a component such as a processor or a memory described as being configured to perform a task may be implemented as a general component that is temporarily configured to perform the task at a given time or a specific component that is manufactured to perform the task. As used herein, the term ‘processor’ refers to one or more devices, circuits, and/or processing cores configured to process data, such as computer program instructions.

A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims and the invention encompasses numerous alternatives, modifications and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example and the invention may be practiced according to the claims without some or all of these specific details. For the purpose of clarity, technical material that is known in the technical fields related to the invention has not been described in detail so that the invention is not unnecessarily obscured.

FIG. 1 illustrates an example of an environment in which documents are clustered. Corpus processor 106 is configured to collect (or otherwise receive) documents from a variety of data sources 110-114. Examples of such documents include news articles, forum messages, blog posts, and any other text (in formats such as HTML, TXT, PDF, etc.) as applicable. In the example shown, corpus processor 106 is configured to scrape content from external website 112 and to perform searches using an API made available by search engine 114, both of which are accessible via a network 108 (e.g., the Internet). Corpus processor 106 is also configured to receive documents from an internal source, such as repository 110.

In various embodiments, corpus processor 106 collects documents on demand. For example, a user of platform 116 (hereinafter “Benjamin Snyder,” an architect living in California) may initiate a request (via interface 118) for documents that pertain to him. In response to the request, corpus processor 106 obtains documents from one or more of the data sources. Corpus processor 106 can also be configured to store and periodically refresh the documents it collects regarding Benjamin Snyder, such as upon the request of Benjamin Snyder, or programmatically (e.g., once a month).

Corpus processor 106 is configured to process the collected documents and make them available to clustering system 102 as an input data set (104). As one example, in some embodiments corpus processor 106 is configured to convert the documents it receives into plaintext, or otherwise extract text from those documents, as applicable. As will be described in more detail below, clustering system 102 is configured to discover clusters in the input data set and define the theme of each cluster.

FIG. 2 illustrates an embodiment of a clustering system. In the example shown in FIG. 2, system 102 comprises standard commercially available server hardware (e.g., having a multi-core processor 202, 8G+ of RAM 204, gigabit network interface adaptor(s) 206, and one or more hard drives 208) running a typical server-class operating system (e.g., Linux). In various embodiments, system 102 is implemented across a scalable infrastructure comprising multiple such servers, solid state drives, and/or other applicable high-performance hardware. Further, as illustrated in FIG. 1, clustering system 102 can be collocated on a platform 116 with other components, such as corpus processor 106. Clustering system 102 can also be configured to work with one or more third party elements. For example, the functionality of corpus processor 106 and/or interface 118 can be provided by a third party.

Whenever clustering system 102 is described as performing a task, either a single component or a subset of components or all components of system 102 may cooperate to perform the task. Similarly, whenever a component of system 102 is described as performing a task, a subcomponent may perform the task and/or the component may perform the task in conjunction with other components. In various embodiments, portions of system 102 are provided by one or more separate devices. As one example, the functionality of preprocessing engine 210 and clustering engine 212 may be provided by two different devices, rather than the functionality being provided by a single device. Also, in various embodiments, system 102 provides the functionality of corpus processor 106 and a separate corpus processor is omitted, as applicable.

Clustering system 102 is configured to perform a variety of tasks, including preprocessing (via preprocessing engine 210) and document clustering (via clustering engine 212).

Preprocessing

Clustering system 102 receives as input a set of documents 104 (e.g., from corpus processor 106). Preprocessing engine 210 uses the data set to create a term-by-document matrix M and a keyword list (ordered by column number of M).

A text object (i.e., a document or set of documents) is represented in Latent Semantic Indexing (“LSI”) by its term frequency (“TF”). The term frequency representations of two example documents d₁ and d₂, and of a collection C₁ comprising those two documents, are as follows:d₁=[2,0,3,0,0,5,0,1,4]d₂=[1,3,0,2,0,3,0,0,3]C₁=d₁+d₂=[3,3,3,2,0,8,0,1,7]

A concept representation of a text object can also be created, which shows whether the object contains the given term, but does not indicate the number of times the term appears in the document. Using the term frequencies for d₁, d₂, and C₁ above, the concept representations are as follows:Term Frequency Rep.→Concept Rep.d₁=[2,0,3,0,0,5,0,1,4]→b₁=[1,0,1,0,0,1,0,1,1]d₂=[1,3,0,2,0,3,0,0,3]→b₂=[1,1,0,1,0,1,0,0,1]C₁=[3,3,3,2,0,8,0,1,7]→B₁=[1,1,1,1,0,1,0,1,1]

Suppose the total data set (e.g., received by clustering system 102 as input 104) has n=N_d documents and a vocabulary of m=N_t keywords. Matrix M can be constructed as:

$M=\left[\begin{array}{ccc}{d}_{11}& \dots & d_{1m}\\ ⋮& \ddots & ⋮\\ d_{n1}& \dots & d_{\mathrm{nm}}\end{array}\right].$

The keyword list has N_t elements and the order of the keywords in the list matches the order of columns of the TF representation of the documents.

In some embodiments, Matrix M and the keyword list are created as follows: Suppose a total of 50 documents are received as a set 104 by clustering system 102. The documents are stored in a directory of storage 208 as individual files. Preprocessing engine 210 uses a Java program to construct a single file (“doc-file”) from the documents in the directory, with each line of the doc-file representing the content of one of the 50 documents. Various rules can be employed to construct the single doc-file, such as by removing email addresses in documents having more than 5,000 characters; truncating documents that are longer than 50,000 characters; and removing any HTML tags or other strings that may erroneously remain after the processing performed by corpus processor 106. Next, the doc-file is used to generate a “mat-file” and a “mat-file.clabel” file by using the doc2mat Perl program. The mat-file includes information about Matrix M and the mat-file.clabel file lists the keywords (e.g., after a tokenization using stemming and ignoring numerical terms is performed). Finally, the mat-file and mat-file.clabel files are read to create the Matrix M and the keyword list using an appropriate Java program. In various embodiments, additional processing is performed when constructing the matrix and keyword list from the mat-file and mat-file.clabel. For example, terms that appear in only a single one of the 50 documents can be omitted.

Clustering

FIG. 3 illustrates an embodiment of a process for clustering documents. In various embodiments the process shown in FIG. 3 is performed by clustering engine 212. In some embodiments clustering engine 212 makes use of industry standard mathematical software tools such as MATLAB, which are installed on or otherwise in communication with system 102.

Initial Clustering

The process begins at 302 when a data set is clustered into one or more initial clusters using a first term set. A variety of clustering techniques can be used and the process described herein adapted as applicable. In some embodiments the clustering is performed using LSI-SVD, as follows. A Matrix Q_a is derived from Matrix M based on the following equation:

Q_a=√{square root over (c)}D_n⁻¹M, where constant c and matrix D_n are defined as follows:

$f_i=\sum _{j=1}^m{\left(M·M^T\right)}_{\mathrm{ij}},g_i=\frac{f_i}{\sum _j{f}_j},c=\frac{1}{\sum _j{f}_j},D_n=\mathrm{diag}\left(\sqrt{g}\right).$

Next, singular value decomposition ([U,S,V]=svd(Q_a,0)) is used to obtain and plot the coordinates of terms in a reduced two-dimensional space. Specifically, the coordinates can be derived from the m×m Matrix U:{right arrow over (t)}_i=(x_i,y_i),x_i=U(x_i,y_i),y_i=U(i,2),iε{1, 2, . . . , m=N_t}.

The angles between terms (term-term angles) are calculated on a unit sphere:

${\theta }_{i,j}={\cos }^{-1}\left(\frac{\stackrel{->}{t_i}·\stackrel{->}{t_j}}{\stackrel{->}{t_i}·\stackrel{->}{t_j}}\right).$

For each term t_i, a group of closely related terms is built, where an assumption is made that term t_i and t_j are closely related if θ is smaller than a threshold value, for example:

${\theta }_{i,j}<\frac{\pi }{4}.$ Each group is an m-dimensional concept representation.

Next, LSI-SVD is used to calculate a document-document distance matrix and plot it on a two-dimensional graph. Specifically, a weight Matrix W is built from Matrix M as follows:

$W_{\mu ,i}=\frac{M_{\mu ,i}{\mathrm{idf}}_i}{\sqrt{W_{\mu }}},W_{\mu }=\sum _{i=1}^m{\left(M_{\mu ,i}{\mathrm{idf}}_i\right)}^2,\mathrm{idf}=\log \left(\frac{n}{n_i}\right),n_i=\sum _{\mu -1}^n{M}_{\mu ,i}.$

Singular value decomposition ([U,S,V]=svd(W^T,0)) is used to obtain and plot the coordinates of documents in a reduced two-dimensional space. Specifically, the coordinates can be derived from the n×n Matrix V:{right arrow over (d)}_μ=(x_μ,y_μ),x_μ=V(μ,1),y_μ=V(μ,2)με{1, 2, . . . , n}.

The distances between documents (“ddDist”) are calculated as follows:dist(d_μ,d_v)=|{right arrow over (d)}_μ−{right arrow over (d)}_v|.

Finally, cluster membership is determined, using the assumption that a document belongs to a cluster (referred to as an “initial cluster” at this stage of processing) if its distance from at least one document in the cluster is less than a threshold value. In some embodiments the threshold value is an optimized radius. For example, it may be estimated as: ρ=κ×σ(min DistMin, max DistMin), where κ is a parameter based on the size of the dataset (for example, κ=3.5 for entity datasets), and σ(min DistMin, max DistMin) is an empirical function derived from the best fitting for several small datasets using the minimum and maximum of the minimum distances between documents. Documents that do not belong to any clusters are left as singletons.

At 304, a theme of each initial cluster is determined by a set of keywords. One way of determining themes is as follows. First, a set of core terms is calculated for each of the initial clusters, using the following three cases:

Case 1: Use the original terms shared by all documents in the cluster. (This case will likely apply to sets with hundreds of terms per document, or more.)

Case 2: If no terms are found in Case 1, add connected terms which are closely related to the terms in the documents, e.g., in accordance with the definition of “closely related” above. (This may be required for some small data sets.)

Case 3: If no terms are found in Case 2, use the union of the terms of all documents in the cluster. (This may be required for some very small data sets.)

Next, the concept of the cluster is calculated as the union of the core terms of the cluster and any terms closely related to those terms.

Finally, a theme representation of the entire data set is determined as follows. A term frequency representation (“clusterTF”) is constructed for each cluster in accordance with rules, such as the following: (1) the term is not shared by all cluster concepts; (2) the term is contained by at least one document in the clusters; and (3) the total term frequency of the term is not less than a threshold value “tfMin.” The concept shared by all clusters (i.e., the “theme” of the dataset) and the “theme” of each cluster, defined by the top “tfMax” (a threshold value, set to 12 by default) frequent terms based on the cluster term frequency representation are displayed.

For very small data sets, the processing performed at 302-304 of process 300 may be sufficient to achieve satisfactory clusters and theme representations. However, for more typical data sets, the existence of noise terms obscures clusters. As will now be described, renovated clusters, using a reduced term space in which noise terms are excluded, can be used to improve clustering results.

Reduced Term Space

At 306, the term space is reduced. One approach for reducing the term space is as follows. Suppose a term frequency representation for each cluster (“clusterTF”) has already been created, as described above. A reduced term space can be constructed by keeping in the term space only those terms in the clusterTFs.

Renovated Clustering

At 308, the data set is reclustered into one or more baby clusters using the reduced term space. In some embodiments, the same clustering techniques employed at 302 are used again at 308. In other embodiments, an alternate clustering technique is used. One example of such an alternate technique will now be described.

First, a new document term frequency Matrix M₂ and keyword list with a reduced term space of dimension m′<m are built. Instead of using singular value decomposition, for the second round of clustering a vector space model (VSM) is used to define relations of documents. The coordinates of documents are found this time using a weight Matrix W built from Matrix M₂ as follows:

$W_{\mu ,i}=\frac{{\mathrm{tf}}_{\mu ,i}{\mathrm{idf}}_i}{\sqrt{W_{\mu }}},W_{\mu }=\sum _{i=1}^{m^{\prime }}{\left({\mathrm{tf}}_{\mu ,i}\times {\mathrm{idf}}_i\right)}^2.$ A variety of different weight formulae can be used—the preceding is provided merely as an example. Next, a document-document relevance coefficient (“ddRC”) is calculated based on the calculated weights, together with a minimum and maximum:

$\mathrm{RC}\left(d_{\mu },d_v\right)=\sum _{i=1}^{m^{\prime }}{w}_{\mu ,i}{w}_{v,i}.$

A document is assumed to belong to a cluster if its relevance coefficient (“RC”) with one of the documents in the cluster is greater than the value of “rcMin.” In some embodiments, rcMin is 0.45. Using such a value for rcMin will typically result in small clusters (also referred to herein as “baby clusters”) and several singletons (illustrated in FIG. 4A).

Reassigning Singletons

At 310, an attempt is made to reassign singletons to the baby clusters. One approach to performing this assignment is as follows: A singleton will be assigned to a given cluster if (1) the RC with that cluster is greater than its RC with any other cluster; and (2) the RC is greater than a minimum threshold value “rcS.” In some embodiments, rcS=0.05 by default.

The RC between a document d_μ and a cluster c_α is calculated as:

$\mathrm{RC}\left(d_{\mu },c_{\alpha }\right)=\sum _{i=1}^{m^{\prime }}{w}_{\mu ,i}\frac{{\mathrm{tf}}_{\alpha ,i}}{W_{\alpha }},W_{\alpha }=\sqrt{\sum _{i=1}^{m^{\prime }}{\mathrm{tf}}_{\alpha ,i}^2}.$

The reassignment of singletons is conceptually illustrated in FIGS. 4A and 4B. Specifically, as shown in FIG. 4A, baby cluster 402 and baby cluster 404 each include two documents. Seven singletons (406-418) are present. The relevance (or closeness) of a singleton to each cluster is calculated. Singletons 410 and 408, and singletons 412 and 414 are respectively close enough to baby clusters 402 and 404 that they will be reassigned to those clusters (as indicated by arrows in FIG. 4A).

As illustrated in FIG. 4B, singletons 408, 410, 412, and 414 can be reassigned in compliance with the reassigning rules described above. The resulting clusters (i.e. baby clusters combined with any reassigned singletons) are referred to herein as “renovated clusters” 450 and 452.

At 312, renovated themes are determined for the renovated clusters. In various embodiments, the processing performed at 312 is the same as the processing preformed at 304.

Output

Finally, at 314 the renovated themes, renovated clusters (including any reassigned singletons), and/or any remaining singletons are provided as output 214, as applicable. The output of process 300 can be used in a variety of contexts, examples of which are described below.

System 102 can be used to help the Benjamin Snyder mentioned in conjunction with FIG. 1 to differentiate documents that are about him from other documents about other Benjamin Snyders. Documents about the user Benjamin Snyder will be clustered together and have as a theme terms such as “architect” and “California.” Documents about the other Benjamin Snyders will also be in their own respective clusters and have other themes, such as “basketball player”/“Kentucky” or “dentist”/“New York,” which do not describe the user.

As another example, a celebrity that may be commonly mentioned in news articles can use system 102 to receive news alerts about one aspect of the celebrity's life (e.g., film-making or parenting) and selectively ignore information about other aspects (e.g., a failed musical career). In this scenario, the input set could be generated as a result of a search for the celebrity on site 114, with the results being automatically clustered into the various themes.

As yet another example, a user of platform 116 can perform a search for documents residing on data source 110 pertaining to “jaguars,” and receive back results clustered based on the various disambiguations of the term.

In various embodiments, output 214 is provided to other processes, in addition to or instead of being displayed to a human user. For example, in the case of the celebrity described above, sentiment analysis can be performed by another component of platform 116 to assign scores to the distinct aspects of the celebrity's life (as determined from the clusters). In such a scenario, the celebrity might learn that articles mentioning the celebrity's film-making attempts are 70% positive, while articles mentioning the musical career are 25% positive.

Example

“Star” Data Set

FIGS. 5-9C and the accompanying text illustrate portions of process 300 as applied to an example set of data. Specifically, suppose a search of data source 114 is performed by corpus processor 106 using the term “star” and that the first 25 results are processed by corpus processor 106 and provided to clustering system 102. The 25 documents collectively include a total of 4215 terms. Four different topics are represented in the data (each having five documents) and five documents are singletons, as illustrated in FIG. 5. When terms that occur in only a single one of the 25 documents are removed, a total of 1074 terms remain as the initial term space.

Initial Clustering

Using singular value decomposition for term-term interaction and LSI-SVD for document-document interaction, with κ=4.5, the initial clusters (and no singletons) depicted in FIG. 6 are obtained. The theme of the data set is determined to be “star,” as it is the shared, connected term with the highest frequency. When a tfMin=8 and tfMax=10 are used to build a term frequency representation and generate themes, it can be seen that a total of eleven of the documents are misplaced. Some of the clusters are virtually correct (e.g., cluster 602 compared to 502). Other clusters are over inclusive (604) with no clear theme, and yet others are under inclusive (e.g., cluster 606 compared to 504). Moreover, there are no singletons. FIG. 7 illustrates the star data set, in two-dimensions, after initial clustering has been performed.

Reducing Term Space and Re-Clustering

Using the clustering term frequency representations obtained in initial clustering, a new matrix and keyword list using the reduced term space is generated. At this point, the dimension of the keyword list is now 93. Reclustering is performed using VSM with an rcMin=0.45 and, as illustrated in FIG. 8, four baby clusters (with lines connecting their member documents) and 16 singletons are found. The baby clusters are small, but are well defined and usable as cluster seeds when reassigning singletons.

Reassigning Singletons

Finally, singletons are reassigned by calculating its RC with each cluster (as in VSM) and keeping the document as a singleton if the max RC<rcS. (Shown in FIG. 9A). In the final output, only three of the 25 documents are misplaced (indicated with underlining) in FIG. 9B. In addition, each cluster has a clear theme, as shown in FIG. 9C.

Although the foregoing embodiments have been described in some detail for purposes of clarity of understanding, the invention is not limited to the details provided. There are many alternative ways of implementing the invention. The disclosed embodiments are illustrative and not restrictive.

Claims

1. A system, comprising: a processor configured to: cluster a data set into one or more initial clusters using a first term space;determine an initial theme for each initial cluster;reduce the first term space to create a reduced term space, wherein reducing the first term space comprises removing at least one noise term from the first term space, wherein the noise term is a term that is determined to be present in at least two of the initial themes determined for two respective initial clusters, and wherein a term frequency of the noise term at least meets a predetermined threshold;recluster at least a portion of the data set into one or more baby clusters using the reduced term space, wherein after reclustering, at least one singleton is present, wherein a singleton is an element from the data set that was not assigned to any baby clusters during the reclustering;assign at least one singleton to at least one baby cluster to form one or more renovated clusters;determine a renovated theme for at least one of the renovated clusters; andprovide as output one or more of the renovated clusters and their respective themes; anda memory coupled to the processor and configured to provide the processor with instructions.

2. The system of claim 1 wherein a first renovated theme for a first renovated cluster comprises one or more frequent terms that are uniquely associated with the first renovated cluster.

3. The system of claim 1 wherein a first initial theme comprises one or more frequent terms.

4. The system of claim 1 wherein the processor is configured to cluster the data set into the one or more initial clusters at least in part by using singular value decomposition on a term-by-document matrix and wherein the processor is configured to recluster at least in part by using a vector space model.

5. The system of claim 1 wherein the data set includes entity data.

6. The system of claim 1 wherein the processor is configured to recluster at least in part by: using a first relevance criterion to generate a set of baby clusters and a set of singletons; andusing a second relevance criterion to assign at least one singleton included in the set of singletons to a baby cluster included in the set of baby clusters, wherein the first and second relevance criteria are different.

7. The system of claim 1 wherein the processor is further configured to preprocess the data set at least in part by generating a term-by-document matrix.

8. The system of claim 1 wherein the processor is further configured to preprocess the data set at least in part by generating the first term space.

9. The system of claim 8 wherein the data set comprises a plurality of documents and wherein the processor is configured to generate the first term space at least in part by evaluating terms present in the data set and not include in the first term space a solo term that occurs solely in a single document included in the plurality of documents.

10. The system of claim 8 wherein the processor is configured to generate the first term space at least in part by evaluating terms present in the data set and not include in the first term space one or more terms associated with numbers.

11. The system of claim 1 wherein at least some of the data included in the data set describes a first individual, wherein at least some of the data included in the data set describes a second individual, and wherein the output is provided to the first individual.

12. A method, comprising: clustering a data set into one or more initial clusters using a first term space;determining an initial theme for each initial cluster;reducing the first term space to create a reduced term space, wherein reducing the first term space comprises removing at least one noise term from the first term space, wherein the noise term is a term that is determined to be present in at least two of the initial themes determined for two respective initial clusters, and wherein a term frequency of the noise term at least meets a predetermined threshold;reclustering at least a portion of the data set into one or more baby clusters using the reduced term space, wherein after reclustering, at least one singleton is present, wherein a singleton is an element from the data set that was not assigned to any baby clusters during the reclustering;assigning at least one singleton to at least one baby cluster to form one or more renovated clusters;determining a renovated theme for at least one of the renovated clusters; andproviding as output one or more of the renovated clusters with their respective themes.

13. A computer program product embodied in a non-transitory computer readable storage medium and comprising computer instructions for: clustering a data set into one or more initial clusters using a first term space;determining an initial theme for each initial cluster;reducing the first term space to create a reduced term space, wherein reducing the first term space comprises removing at least one noise term from the first term space, wherein the noise term is a term that is determined to be present in at least two of the initial themes determined for two respective initial clusters, and wherein a term frequency of the noise term at least meets a predetermined threshold;reclustering at least a portion of the data set into one or more baby clusters using the reduced term space, wherein after reclustering, at least one singleton is present wherein a singleton is an element from the data set that was not assigned to any baby clusters during the reclustering;assigning at least one singleton to at least one baby cluster to form one or more renovated clusters;determining a renovated theme for at least one of the renovated clusters; andproviding as output one or more of the renovated clusters and their respective themes.

14. The system of claim 1 wherein the processor is further configured to: determine a theme for at least one baby cluster that did not form a renovated cluster during the assignment; andprovide as output the baby cluster that did not form a renovated cluster during the assignment and the theme.

15. The system of claim 1 wherein the processor is further configured to: determine a theme for at least one singleton that was not included in a renovated cluster during the assignment; andprovide as output the singleton that was not included in a renovated cluster during the assignment and the theme.

16. The method of claim 12 wherein clustering the data set into the one or more initial clusters includes using singular value decomposition on a term-by-document matrix and wherein reclustering includes using a vector space model.

17. The method of claim 12 wherein reclustering includes: using a first relevance criterion to generate a set of baby clusters and a set of singletons; andusing a second relevance criterion to assign at least one singleton included in the set of singletons to a baby cluster included in the set of baby clusters, wherein the first and second relevance criteria are different.

18. The method of claim 12 further comprising preprocessing the data set at least in part by generating a term-by-document matrix.

19. The method of claim 12 further comprising preprocessing the data set at least in part by generating the first term space.

20. The method of claim 19 wherein the data set comprises a plurality of documents and wherein the method further includes generating the first term space at least in part by evaluating terms present in the data set and not include in the first term space a solo term that occurs solely in a single document included in the plurality of documents.