Method and apparatus for tracking the provenance of information

Inventors

Cohen, Jonathan D.

Abstract

A method for determining a flow of information within a plurality of documents is provided. The method may include receiving the plurality of documents with each document having document content and a document timestamp. Based on the documents, the method may include constructing an inferred document causality forest by determining causal edges of the inferred document causality forest based on the document contents and the document timestamps. The inferred document causality forest may indicate a temporal flow of information from at least one earliest causal document of the plurality of documents. Further, each causal edge may extend from a node for an antecedent document to a node for a subsequent document that has a similarity above a similarity threshold and a document timestamp of the antecedent document may be earlier in time than a document timestamp of the subsequent document.

Core Innovation

A method for determining a flow of information within a plurality of documents receives the plurality of documents, where each document includes document content and a document timestamp. The method constructs an inferred document causality forest based on the plurality of documents by determining causal edges between documents based on the document contents and the document timestamps, and the inferred document causality forest indicates a temporal flow of information from at least one earliest causal document of the plurality of documents.

Each causal edge extends from a node for an antecedent document to a node for a subsequent document that has a similarity above a similarity threshold, where the document timestamp of the antecedent document is earlier in time than the document timestamp of the subsequent document. The antecedent document and the subsequent document are each one of the documents of the plurality of documents, and the temporal flow is represented by causal edges linking earlier, highly similar documents to later documents.

For each causal edge, each document is associated with a channel, and the method determines an influence score for each channel represented within the inferred document causality forest. The influence score is based on a quantity of subsequent documents of other channels linked to each channel in the inferred document causality forest, supporting channel-level analysis of temporal information flow by quantifying cross-channel linking behavior derived from the inferred document causality forest.

Claims Coverage

The independent claims are directed to a method and an apparatus for determining a temporal flow of information among documents by constructing an inferred document causality forest, defining causal edges using content similarity and timestamps, and computing channel influence scores from cross-channel link quantities. The claim set further includes dependent claim variations that specify similarity computation and antecedent selection, and at least one extension that constructs an inferred channel causality graph from the inferred document causality forest.

Across the independent claims, the core coverage is directed to building an inferred document causality forest with content- and timestamp-based causal edges using a similarity threshold, and to deriving channel influence scores from the resulting cross-channel links. Dependent coverage includes specific ways of determining causal edges and an extension that constructs an inferred channel causality graph from the inferred document causality forest.

Stated Advantages

Documented Applications