Systems and methods for clustering user sessions using multi-modal information including proximal cue information
View Patent ↗Techniques for clustering user sessions using multi-modal information including proximal cue information are provided. The topology, content and usage of a document collection or web site are determined. User paths are then identified using longest repeating subsequence techniques. An information need feature vector is determined for each significant user path. Further, other feature vectors and proximal cue vectors for each document or web page in the significant path are determined. The other feature vectors include a content feature vector, a uniform resource locator feature vector, an inlink feature vector and an outlink feature vector, among others. The feature vectors and the proximal cue vectors are combined into a multi-modal vector that represents a user profile for each significant user path. The multi-modal vectors are clustered using a type of multi-modal clustering such as K-Means or Wavefront clustering.
1. A computer implemented method for clustering user sessions using multi-modal information and proximal information, comprising:
selecting a plurality of user paths in a collection of content portions, and for each user path:
determining multi-modal information for content portions associated with the user path;
determining proximal information for content portions associated with the user path;
combining the multi-modal information for content portions and the proximal information for content portions associated with the user path to form a user profile having a unified representation; and
clustering multi-modal information and proximal information of user profiles based on similarity.
2. The method of claim 1 , wherein determining proximal information includes;
analyzing portions of a text associated with a link between content portions along the user path.
3. The method of claim 1 , wherein determining proximal information includes;
analyzing portions of a text associated with an image link between content portions along the user path.
4. The method of claim 1 , further comprising:
using longest repeating subsequence to select the plurality of user paths.
5. The method of claim 1 , further comprising:
using vectors to represent multi-modal information and proximal information.
6. The method of claim 5 , further comprising:
weighting the vectors representing multi-modal information; and
weighting the vectors representing the proximal information.
7. The method of claim 1 , further comprising:
weighting content portion in the user path based on access frequency of the content portion.
8. The method of claim 1 , further comprising:
weighting content portion in the user path based on a position of the content portion in the user path.
9. The method of claim 1 , wherein clustering the multi-modal information and the proximal information is performed using at least one of K-Means clustering and Wavefront clustering.
10. A computer system to cluster user sessions using multi-modal information and proximal information, comprising:
a user path determining circuit that selects a plurality of user paths in a collection of content portions;
a multi-modal information determining circuit that determines multi-modal information for content portions associated with the user path;
a proximal information determining circuit that determines proximal information for content portions associated with the user path;
a similarity determining circuit that combines the multi-modal information for content portions and the proximal information for content portions associated with the user path to form a user profile having a unified representation; and
a cluster and similarity measure determining circuit that clusters multi-modal information and proximal information of user profiles based on similarity.
11. The system of claim 10 , wherein the proximal information determining circuit analyzes portions of a text associated with a link between content portions along the user path.
12. The system of claim 10 , wherein the proximal information determining circuit analyzes portions of a text associated with an image link between content portions along the user path.
13. The system of claim 10 , wherein the user path determining circuit selects user paths using the longest repeating subsequences.
14. The system of claim 10 , wherein the multi-modal information determining circuit includes at least one of;
a content feature determining circuit that determines contents of content portions associated with the user path;
a uniform resource locator feature determining circuit that determines uniform resource locators of content portions associated with the user path;
an inlink feature determining circuit that determines inlinks of content portions associated with the user path; and
an outlink feature determining circuit that determines outlinks of content portions associated with the user path.
15. The system of claim 10 , further comprising:
an information weighting circuit that weights the multi-modal information and the proximal information.
16. The system of claim 10 , further comprising:
a content portion access frequency weighting circuit that weights at least one content portion associated with the user path based on access frequency.
17. The system of claim 10 , further comprising:
a path position weighting circuit that weights at least one content portion associated with the user path based on the position of the content portion within the user path.
18. The system of claim 10 , where the cluster and similarity measure determining circuit clusters multi-modal information and proximal information using at least one of K-Means clustering and Wavefront clustering.