Main    Products    Solutions    Technology    Pricing    Partners    Support    Corporate Info    Feedback    Contact Us
Technology :

Search Science:

Pattern Profiling Search (a results distinction engine)
PatternScape enhances conventional key word search with profiling and clustering methods. These methods present contour maps of search results with distinction across the resulting clusters. Rather than relying on cleverly contrived filtering mathematics to limit results, PS employs math to provide results in structured cluster trees. Thus, one can better navigate vast quantities of results, which have been transformed into distinct cluster groupings. Distribution vectors and power values further illuminate results and assist in user navigation.

For example, the system matches behavior traits and markers (clusters) of criminals in a crime database and not simply present huge lists of indistinct criminal reports. When the user examines a mining survey database for instance, KS endeavors to identify drill targets with inter-relating metrics and not just millions of documents of surveys and compound mixtures.

PS can be referred to as a first pass analysis tool for professional and scientific data. PS structures information to assist users in decisions to determine the next steps in an analytic problem and is enhanced by cluster based pattern profiling search results.

In short, PS is a pattern profiling system focused on providing ClusterViews to help users navigate and drill down to pertinent results through a logical tree and branch presentation. Distribution vectors and power values assist in quantitative and qualitative analysis for the extreme user.

PS math
PS math further intensifies the researchers analysis potential. Whether the research is for market adoption of a taste in a food, for example, or the feature in a mechanical product or its expected demographic user, PatternScape helps identify weaknesses and strengths through a pattern clustering presentation of variants and their vector power and distribution. This math will simply substantiate choices based on objective metrics.

Information contours are made visible through the process of profiling and mapping symptoms, behavior, traits, and markers. Similarities and differences, and their inter-related patterns, are indicated for a variety of uses.

Some examples:
  • Finding a mortgage fraud applicant who is operating cross-institutionally is an example of pattern profiling. Identification of behavior, traits, and patterns, contribute to a subject profile. No matter how skilled the subject, some trait elements begin to propagate as patterns across many different incidences and crimes. We simply map all these incidences to the subject whether they are using aliases or not and cluster them based on the crime, case and incident reports.
  • A DNA base pair (gene) cluster or a medical symptom cluster can present disease in similar ways.
  • The same can be applied to a security person who records an observation of suspicious activity, for example. PS simply maps the converted recording to the intelligence (security) repository and identifies other known incidences. It is not difficult to imagine how a security repository for all casinos could track aliases and repeat behavior, for instance.
  • A resume can be mapped to a job database in the same way thereby identifying skill-to-job matches (information convergence). This is done with PatternScape mathematical clustering thus obtaining scientific precision.

Document mapping
Document mapping is an analytic science used to map the content within a document and present it visually for examining content clustering. Using similar principles as frequency of occurrence to attribute significance, document mapping creates a map of inter-related complex clusters from the document. The user might then navigate the cluster distribution and send the desired line of enquiry to the search engine for matching documents. Thus a mapped resume will find matching job assignments that most closely relate to the patterns in the resume. A mapped observation will find related observations. And a mapped technical paper will find diagnosis for engine repair, for example.

DocMap is really quite flexible and works for many different professional needs: the imagination is its limit. Therefore, lets imagine a difficult example: a technical paper on the cause of a disease based on gene analysis. By submitting this paper one might find an inter-relationship of genotypes and mutants across global genome repositories which suggests the notion and importance of co-existence, similarity and inter-relationship. Regardless the need, PS is useful for complex or ordinary analytic needs.

Co-Existence and terms and phrase convergence
Co-x, as it is known in PatternScape, is a mathematical technology, which like DocMap, analyses content and creates a ClusterView of the topology of content. The key difference of Co-x and DocMap, however, is that Co-x operates across the entire intelligence repository (index), which is referred to as the total information space. The resulting map shows the clustering, co-clustering, and co-existence of content in the total space, thus providing a complex inter-relationship and clusters of terms, phrases and codes.

If the space were an alphanumeric DNA sequence array, Co-x would simply map clustering sequences. A simple example of its use: Enter the following phrase in the Co-x pad: “Acute Pharyngitis”. The resulting Co-x tree will cluster the most frequent co-existing phrases with “Acute Pharyngitis”. One could easily guess that “sore throat” might well be the predominant co-existing symptom followed by a sub relating “runny nose”. Conversely, enter sore nose and watch for the results - it might surprise. Indeed, Vicodin Lortab is recommended pain treatment and Co-x presented it as co-existing with sore throat.

Often users don’t quite know what to enter in a search query, therefore let Co-x help you identify co-relating elements that in turn might aid to identify lines of interest.

Cluster views (variant maps) created from the matrix search pad contain valuable information about frequency, structure and occurrence in the data set and their vector relationships. The vector information maps the complex relationships of terms with one another throughout the document space and within each document. The effect is a power value (eValue) and a distribution vector. The user, analyzing this kind of information is able to differentiate findings with distribution distinctions and preponderance. A market and competitor study, terrorist profile, medical symptoms, and mine survey (compound and geophysical clustering) are examples for math use.

Link views are conventional result lists of target documents and links. They present after one selects a ClusterView branch, the results of which are matching documents and links. Each document and link result in LinkView has a power value and distribution vector to show the balance of content (contour at the target level) within each document. Consequently, preponderance and order in LinkView is governed by supporting metrics, the use and significance of which is placed in the researchers' hands and not some cleverly contrived ranking system. Consequently, LinkView and ClusterView results might be pertinent, while each individual branch and related documents can be examined for utmost pertinence. In a set of criminal reports, one could easily imagine knowing which document or group of cases apply most usefully and why to the investigation.

Copyright © 2006 KnowledgeShape. All rights reserved.