Construction and application of knowledge atlas de (IV): knowledge acquisition

This chapter first introduces the definition of knowledge acquisition related tasks 、 Important meetings and current mainstream methods , Then it introduces the knowledge acquisition tasks and methods involved in building e-commerce knowledge map in industrial application scenarios , Finally, it introduces the task of knowledge acquisition in low resource scenarios .

4.1 Overview of knowledge acquisition

Knowledge acquisition is an important part of building a large-scale knowledge map , It is the process of instantiating the knowledge elements defined by knowledge modeling , Based on the domain knowledge representation, it realizes from heterogeneous multi-source 、 Acquire knowledge from multi-scale data , For subsequent knowledge fusion 、 Knowledge reasoning calculation and knowledge empowerment provide knowledge support .

This chapter will first introduce some definitions of knowledge acquisition 、 Relevant subtasks and some influential international and domestic evaluation meetings ; Then for some important 、 The core subtask introduces the relevant scientific research progress ;

Finally, for Alibaba e-commerce commodity knowledge map , Combined with a specific business scenario of Alibaba , Introduce the specific practice and application of knowledge acquisition in the industry .

4.1.1 Knowledge acquisition related tasks

For knowledge acquisition , Popular understanding is to obtain all kinds of knowledge structurally from various multi-source heterogeneous data sources , The knowledge here refers to the second 2 Data after instantiation of knowledge elements defined by knowledge modeling in Chapter , The multi-source heterogeneous data source can be structured data （ Such as linked data 、 Database data ）、 Semi-structured data （ Such as web page HTML、XML、 Encyclopedia data, etc ） And unstructured data （ Massive text data , Even voice 、 Image data, etc ）. Instantiate knowledge through continuous learning , Continuously and reliably expand the knowledge map , Build a large-scale and high-quality knowledge map , It is the key technology in the process of map construction .

As a whole , The goal of knowledge acquisition is to use low cost 、 Build a knowledge map from multi-source data or iteratively expand the existing knowledge map , The knowledge map includes entities 、 Entity relations 、 Event relationship and attribute relationship .

Module input For large-scale data , Including structured data 、 Semi structured data and unstructured text data .

Output module </