Bipartite Network of Interest (BNOI): Extending Co-Word Network with Interest of Researchers Using Sensor Data and...

Area of Science:

• Information Science
• Computer Science
• Sensor Technology

Background:

• Traditional co-word networks often yield overly general results, failing to capture specific researcher interests.
• Existing methods lack the granularity to distinguish between general keywords and those reflecting genuine researcher inquiry.
• Identifying specific researcher interests is crucial for domain experts seeking targeted knowledge.

Purpose of the Study:

• To develop a knowledge network that specifically addresses researcher interests in sensor applications.
• To answer the question: "what sensors can be used for what kind of applications?"
• To create a more domain-specific and insightful knowledge representation than traditional co-word networks.

Main Methods:

• Constructed a bipartite network of interest (BNOI) by generalizing specific researcher questions.
• Employed classification models trained on nine feature extraction methods (N-grams, Word2Vec, BERT).
• Utilized a multi-feature fusion strategy and a voting principle (VP) method for model assembly.

Main Results:

• The BNOI model achieved high F-measures: 93.2% for "sensors" and 85.5% for "applications" after bias removal and cross-validation.
• Demonstrated improved accuracy in identifying entities of interest compared to traditional co-word approaches.
• Successfully generalized specific research questions into a structured knowledge network.

Conclusions:

• The bipartite network of interest (BNOI) effectively captures researcher interests in sensor technology.
• BNOI provides a more targeted and valuable knowledge discovery tool for domain experts.
• This approach enhances the ability to answer specific research questions within specialized fields.