Gene

Theoretical computer science

Co-occurrence

Bipartite graph

Complement (music)

Set (abstract data type)

Biochemistry

Data mining

Computer science

Natural language processing

Thermodynamics

Physics

Diffusion

Complementation

Graph

Mathematics

Artificial intelligence

Geometry

Cluster analysis

Information retrieval

Chemistry

Programming language

Thesaurus

Word (group theory)

Phenotype

Scientometrics

The performance of keyword expansion in prior methods is often enhanced by adopting external knowledge. Given a set of initial keywords, this paper is motivated to propose a novel method to expand semantically or conceptually related keywords from domain corpus by employing mass diffusion. A bipartite word network is thus constructed based on co-occurrence relations between initial keywords and candidate words. The expanded keywords are identified via two-step mass diffusion which is carried out in the bipartite network. Experimental results prove that the proposed method outperforms both the typical statistical-based approach and graph-based approach. Our research is expected to complement the theoretical framework of keyword expansion and is applicable to the scenarios of query expansion, thesaurus construction, and text clustering.

A co-occurrence based approach of automatic keyword expansion using mass diffusion