[Dry Store] Structural knowledge mining of large -scale text data

Source: Specialty

This article is a tutorial, it is recommended to read for 5 minutes

We have studied the principles and methods of excavating factual knowledge structures (for example, entities and their relationships) from a large number of non -structured text libraries.

Although the data of the real world is huge, it is unstructured to a large extent and exist in the form of natural language text. Digging the structure from a large amount of text data, without a large number of artificial annotations and marks, this is a challenge, but it is also very ideal. In this book, we have studied the principles and methods of digging factual knowledge structures (eg, entities and their relationships) from a large number of non -structured text fabrics. Different from many existing structure extraction methods, the existing methods are seriously dependent on artificial annotation data for model training. Our light work method is used as remote supervision by artificial management stored in external knowledge bases, and large -scale text corpus library is used. The rich data redundant in the context is understood. This light work -volume mining method brings a series of new principles and powerful methods to build a text corpus, including: (1) physical identification, typing, and synonymal discovery; (2) physical relationship extraction; (3) open domain attributes Value mining and information extraction. This book introduces this new research cutting edge and points out some promising research directions.

https://www.morganclaypool.com/doi/10.2200/s00860ed1v01y201806dmk015

- END -