A Multi-RNN Research Topic Prediction Model Based on Spatial Attention and Semantic Consistency-Based Scientific...

Area of Science:

• Computer Science
• Information Science
• Computational Linguistics

Background:

• Predicting research topics in interconnected fields like computer science presents challenges in fine-grained topic representation and cross-field semantic consistency.
• Existing research topic prediction methods struggle to address the complexities of interdisciplinary scientific influence.

Purpose of the Study:

• To develop an advanced model for predicting research topics within the computer science discipline.
• To address the limitations of current approaches in modeling fine-grained topics and maintaining semantic consistency across related fields.

Main Methods:

• Utilized multiple Recurrent Neural Network (RNN) chains to model research topics across different fields.
• Implemented a novel prediction model incorporating spatial attention for attribute importance and semantic consistency-based scientific influence modeling.
• Mapped research topics into a unified semantic space to capture interdisciplinary scientific influence.

Main Results:

• The proposed model demonstrated superior performance compared to state-of-the-art methods in extensive experiments.
• Achieved significant improvements in research topic prediction accuracy across five related computer science fields.
• Validated the effectiveness of spatial attention and semantic consistency in modeling cross-field influences.

Conclusions:

• The developed model effectively addresses the challenges of fine-grained topic representation and semantic consistency in interdisciplinary research.
• The approach offers a robust solution for predicting future research trends in computer science.
• Highlights the importance of considering scientific influence and semantic coherence for accurate topic prediction.