Technology prediction

摘要

Embodiments of the invention relate to technology prediction. A technical dictionary of technical terms is constructed based on a collection of documents. The technical terms are partitioned into equivalence classes. A table is generated that correlates technical terms across equivalence classes based on temporal co-occurrence of the technical terms across the equivalence classes. For a given technical term the table is accessed to determine a first set of technical terms that correlate to the given technical term. The table is accessed again to determine a second set of technical terms that correlate to the first set of technical terms. It is predicted that the second set of technical terms will correlate to the given technical term in the future.

主权项

1. A method for technology prediction, the method comprising:
constructing a technical dictionary based on a collection of documents, the technical dictionary comprising technical terms; partitioning the technical terms into equivalence classes; generating a table that correlates technical terms across equivalence classes based on temporal co-occurrence of the technical terms across the equivalence classes; and for a given technical term:
accessing the table to determine a first set of technical terms that correlate to the given technical term;accessing the table to determine a second set of technical terms that correlate to the first set of technical terms, the second set of technical terms determined based on a computation of a co-occurrence score and a time-based correlation between the second set of technical terms and the first set of technical terms; andpredicting a candidate formation of new technology that merges a technology corresponding to the first set of technical terms with a second technology corresponding to the second set of technical terms, the predicting based on associations determined among the first set of technical terms and the second set of technical terms, the associations determined as a function of changes over time in the co-occurrence score and the time-based correlation.