摘要：

Abstract:

【Purpose/significance】 The mining and presentation of rational knowledge can further realize the semantic understanding

of the text content, so as to contribute to the prediction of the later development of the event. However, with the explosive growth of net⁃

work text, the phenomenon of information complexity, heterogeneity and overload has seriously interfered with the audience's accurate

selection of relevant text information. Therefore, it is urgent to build a more scientific knowledge system to meet users' demands for ac⁃

curate knowledge.【Method/process】 Firstly, according to the characteristics of network text theory, The knowledge representation

method based on the theory graph is adopted to model the knowledge of network text theory, and the theory ontology ELO is con⁃

structed. Then the multi-source text data is collected based on crawler and the corpus is constructed. Meanwhile, the instance extrac⁃

tion method based on pattern matching is summarized and discussed under the guidance of the theory ontology. Finally, from the per⁃

spective of ontology representation, the concrete process and method of multi-source web text fusion are designed.【Result/conclu⁃

sion】 In the specific knowledge fusion framework, a strict knowledge organization system of web text is constructed to realize the effec⁃

tive utilization and value-added of web text. This framework is feasible and has theoretical guidance and reference value for the con⁃

struction and application of knowledge base.【Innovation/limitation】 This paper constructs the framework of knowledge fusion by syn⁃

thesizing three aspects, namely, the representation and extraction of information features, fusion algorithm and inference application of

online travelogue texts. The accuracy of knowledge extraction method and quality evaluation of fusion algorithm proposed in this paper

need to be further studied and refined.