System and Method for Extracting Deep Learning Based Causal Relation with Expansion of Training Data

摘要

The present invention relates to a device and method for extracting a causal relationship based on deep learning through the expansion of learning data, capable of efficiently building a causal relationship model in a new domain while minimizing issues about data building expenditures incurred by domain specialization when a causal relationship needs to be extracted from a sentence. The device includes: a data analysis and expansion part performing labeling for the extraction of a causal relationship through data analysis, applying boot strapping, and applying bagging and active learning for the relief of error propagation to expand data; and a deep learning-based causal relationship extraction part using a whole sentence as input, deriving a correct answer label corresponding to each morpheme lastly by adding up forward and backward output values of each sequence, reflecting word class information by performing word embedding by morpheme and embedding by syllable, and using a clue word dictionary quality reflecting a causal relationship characteristic to perform the extraction of a causal relationship.