Framework for large-scale multi-label classification
AssigneeLinedIn Corporation
InventorsLiyun Li, Hyun Chul Lee, Shuai Li
Life of this patent
Estimating the term…
Abstract
A framework for large-scale multi-label classification of an electronic document is described. An example multi-label classification system is configured to first apply weak classifiers and identify seed labels, then determine additional labels based on the seed labels and label correlation data derived from member profiles maintained by an on-line social network system. The multi-label classification system then constructs an s-t graph comprising nodes that correspond to the seed labels and the additional labels. A clustering algorithm is then applied to the constructed graph to arrive to a labels graph. The labels graph is deemed to include nodes that correspond to topics discussed or referenced in the electronic document.
Claims19
A method comprising:
- applying, one or more weak classifiers to contents of an electronic document to identify one or more seed labels, the one or more seed labels representing respective one or more preliminary content topics associated with the electronic document;
- generating a seed graph, nodes of the seed graph representing the one or more seed labels;
- based on co-occurrence of labels in member profiles of an on-line social networking system, deriving one or more additional labels, utilizing the one or more seed labels;
- generating, using at least one processor, augmenting the seed graph utilizing the one or more additional labels to generate an expanded graph comprising nodes and edges, the nodes representing the one or more additional labels and the one or more seed labels, each node from the expanded graph that represents an additional node from the one or more additional labels does not also represent a seed label from the one or more seed labels, an edge in the expanded graph having an edge weight constructed as correlation between labels represented by nodes connected to the edge;
- applying a clustering algorithm to the expanded graph to generate a labels graph;
- and
- identifying nodes of the labels graph, as a set of resolved content topics associated with the electronic document.
A machine-readable non-transitory storage medium having instruction data to cause a machine to:
- apply one or more weak classifiers to contents of an electronic document to identify one or more seed labels, the one or more seed labels representing respective one or more preliminary content topics associated with the electronic document;
- generate a seed graph, nodes of the seed graph representing the one or more seed labels;
- derive one or more additional labels based on co-occurrence of labels in member profiles of an on-line social networking system, utilizing the one or more seed labels;
- generate augment the seed graph utilizing the one or more additional labels to generate an expanded graph comprising nodes and edges, the nodes representing the one or more additional labels and the one or more seed labels, each node from the expanded graph that represents an additional node from the one or more additional labels does not also represent a seed label from the one or more seed labels, an edge in the expanded graph having an edge weight constructed as correlation between labels represented by nodes connected to the edge;
- apply a clustering algorithm to the expanded graph to generate a labels graph;
- and
- identify nodes of the labels graph as a set of resolved content topics associated with the electronic document.
Dependencies are inferred from the claim text.