Building information networks using semantic based techniques to avoid tedious work and to achieve high efficiency has been a long-term goal in the information management world. A great volume of research has focused on developing large scale information networks for general domains to pursue the comprehensiveness and integrity of the information. However, constructing customised information networks containing subject-specific knowledge has been neglected. Such research can potentially return high value in terms of both theoretical and practical contribution. In this paper, a new type of network, solution-oriented information network, is coined that includes research problems and proposed techniques as nodes, and the relationship between them. A lightweight Semantic-based Knowledge Fusion Model (SKFM) is proposed leveraging the power of Natural Language Processing (NLP) and Crowdsourcing to construct the proposed information networks using academic papers (knowledge) from Scopus. SKFM relies on NLP in terms of automatic components while Crowdsourcing is initiated when uncertain cases arise. Applying the NLP technique assists to develop a semi-automatic knowledge fusion method for saving effort and time in extracting information from academic papers. Leveraging human power in uncertain cases is to make sure the essential concepts for developing the information networks are extracted reliably and connected correctly. SKFM shows a theoretical contribution in terms of lightweight knowledge extraction and reconstruction framework, as well as practical value by providing solutions proposed in academic papers to address corresponding research issues in subject-specific areas. Experiments have been implemented which have shown promising results. In the research field of intrusion detection, the information of attack types and proposed solutions has been extracted and integrated in a graphic manner with high accuracy and efficiency.