System and method for linguistic feature generation across multiple layer word representations

摘要

A computer implemented method, system and computer program for linguistic feature generation across multiple layer word representations. A computer-implemented method is provided for outputting one or more cross-layer patterns that identify target semantic phenomena in text, the method being designated as representing target semantic phenomena. For each word of at least some words of each training text fragment of the training text fragment, extracting the feature value defined by the respective layer and statistically analyzing the feature value identified for the training text fragment Identifying one or more cross layer patterns comprising layers representing a common pattern for training text fragments, the common cross layer pattern comprising layers of one or more words Define at least another feature value of one or more feature values of another and another respective layer of another word And analyzing, and outputting the identified cross layer pattern to identify text fragments that represent the target semantic phenomenon. [Selected figure] Figure 3