Text Clustering		 Updated: Nov 29, 2004


What Clustering is:
A tool that uses linguistic and relational technologies to analyze the patent documents in your Result Set or Work File, and groups them based on the key words found in the analysis. The linguistic analysis is applied to the title and abstract.

[back to top]


What to use Clustering for:
Use Clustering to:
  • Identify and focus on the most relevant patents
  • Examine relationships between patents
  • Create a more targeted analysis of patent data
  • Analyze similarities to focus on what's most relevant
 [back to top]


Clustering Derwent data:
Derwent gives you patent titles and abstracts rewritten in clear and consistent industry-specific terms. When you use Clustering on Derwent data, your results will be even more meaningful than those you get when clustering original patent data because original patents are sometimes written to conceal the real value of an innovation.

[back to top]


Where to access Clustering:
Access Clustering from the Clustering tab on your Result Set or Work File.

[back to top]


How Clustering works:
Documents are analyzed and assigned uniquely to one and only one cluster. The list of clusters is displayed, along with the keywords that characterize the cluster and the number of patent documents that are contained in the cluster. You can drill down into any cluster and view individual documents — or you can focus on the most relevant clusters and explore newly exposed relationships. You can also view the relationships between clusters based on common terms found in the documents in each cluster.

Results can also be visualized graphically with a map that provides an overview of the clusters and an indication of the relationship among them.

[back to top]


Overview of the Clustering control panel:
Following is a high-level overview of the Clustering control panel:

Screen capture of the Clustering control panel

[back to top]


How to begin:

1.   From the Current Results display for a query Result Set or a Work File, click the Clustering tab.
2.   Delphion displays your Result Set or Work File with the Clustering control panel (shown following).

Screen capture of the Clustering control panel

3.   Choose the number of Items to Cluster.
4.   Choose the Number of clusters you want created in your analysis.
5.   Click ANALYZE.
6.   The Cluster Overview page displays.

[back to top]


The Cluster Overview page:
The Cluster Overview page (shown following) shows you the Cluster number that has been assigned to each cluster, the number of patents in each cluster, and the common keywords (Descriptive words) that describe the cluster.

Screen capture of Clustering Overview page

Click a Cluster number to go to the Current Cluster page for that cluster. From the Current Cluster page, you can choose to create a Result Set from that individual cluster and then, from the Result Set, use any of the other available analytical or productivity tools like Work Files or Snapshot. See Creating a Result Set from a cluster for more information.

OR

Click the Show Visual Map link to go to the Visual Clustering Map for the whole group of clusters. See Using the Visual Clustering Map for more information.

[back to top]


Creating a Result Set from a cluster:
The Current Cluster page (shown following) lets you drill down into a specific cluster, view the Delphion Integrated Views for individual patents in that cluster, and create a Result Set that includes all of the patents in that specific cluster.

Screen Capture of Current Cluster page

In the Cluster Content section, click View as Result Set to see that cluster in the Delphion Result Set format.

Once the patents for the cluster are in a Result Set, you can:
  • change the display by showing thumbnails, abstracts, or a variety of other attributes
  • create a Work File
  • download individual PDFs
  • use PDF Express or Data Extract
  • create a Snapshot
  • click to see details on individual patents
Screen capture of the Custer Results Set

[back to top]


Using the Visual Clustering Map:
  1. Follow steps 1-5 above to view Cluster Overview page.
  2. Click the Show Visual Map link on the top of the Cluster Overview page.
  3. Delphion returns Visual Clustering Map. Each cluster is drawn on the map with:
    • A rectangle with a size proportional to the number of patents in the cluster
    • Each rectangle labeled with two numbers, the first being the cluster number and the second being the number of patents in the cluster
    • As many keywords for that cluster displayed based on the size of the rectangle
  4. To view the Cluster Content page for a cluster, double-click on the cluster.
  5. To move a cluster on the map, left-click on the cluster and drag it to a new location.
  6. To view all of the keywords that define a cluster, right-click on the cluster.
  7. Use controls across the top of the map to manipulate the map.
  8. Controls on the Visual Clustering Map allow you to:
    • Stop and Start the clustering process
    • Size (- or +) the entire map
    • Change the Font Size of the text
    • Show or hide the similarity percent between clusters by checking and un-checking Link Values check box
    • Show or hide the relationships between clusters by changing the Min Link value which establishes the minimum similarity percent between clusters required for the relationship to be displayed
    • Make text boxes Uniform Sizes
 Note: Lines shown between clusters, indicate where there are relationships between the clusters. A red line indicates a high percentage of similarity, a green line indicates a medium percentage and the blue line indicates a low percentage.

[back to top]
Thomson Reuters Copyright © 1997-2013 Thomson Reuters 
Subscriptions  |  Web Seminars  |  Privacy  |  Terms & Conditions  |  Site Map  |  Contact Us  |  Help