Jiahao Li
Research Subjects
Research Interests
Jiahao Li has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Chao Ni
Innovation and Management Center (Ningbo) for School of Software Technology, Zhejiang University; The State Key Laboratory of Blockchain and Data Security, Zhejiang University, China
1 shared publication - Xinrui Li
Innovation and Management Center (Ningbo) for School of Software Technology, Zhejiang University; The State Key Laboratory of Blockchain and Data Security, Zhejiang University, China
1 shared publication
Latest Publications
- Open Access
- 12 February 2026
- Published by Elsevier in SSRN Electronic Journal
MoPRec: Multi-modal Fusion with Developer Preference Modeling for Enhanced GitHub Topic Recommendations
Topic tagging is crucial for enhancing the discoverability and comprehensibility of open-source projects. However, existing recommendation systems often overlook developer-specific interests, struggle with noisy textual data, and lack effective multi-modal fusion. To address these limitations, we propose a novel multi-modal topic recommendation system for the GitHub platform. Our approach introduces a developer-centric model that quantifies developers' technical preferences via a developer-topic interaction matrix, modeled using explicit statistics, Non-negative Matrix Factorization (NMF), and Graph Attention Networks (GATs). To combat noise in project documentation, we design a semantic distillation framework leveraging a Large Language Model (LLM) with tailored prompts to extract high-fidelity textual features. Furthermore, a hierarchical attention mechanism is developed to effectively integrate heterogeneous data modalities. Comprehensive experiments on a dataset of 5,590 projects demonstrate that our system achieves an F1-score of 0.4546, surpassing the best-performing baseline (IR-TFIDF) by 23.40\%. Case studies further validate its superiority in recognizing technical terms, offering new insights for software comprehension on knowledge-intensive platforms.
- 12 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMoPRec: Multi-modal Fusion with Developer Preference Modeling for Enhanced GitHub Topic Recommendations
Topic tagging is crucial for enhancing the discoverability and comprehensibility of open-source projects. However, existing recommendation systems often overlook developer-specific interests, struggle with noisy textual data, and lack effective multi-modal fusion. To address these limitations, we propose a novel multi-modal topic recommendation system for the GitHub platform. Our approach introduces a developer-centric model that quantifies developers' technical preferences via a developer-topic interaction matrix, modeled using explicit statistics, Non-negative Matrix Factorization (NMF), and Graph Attention Networks (GATs). To combat noise in project documentation, we design a semantic distillation framework leveraging a Large Language Model (LLM) with tailored prompts to extract high-fidelity textual features. Furthermore, a hierarchical attention mechanism is developed to effectively integrate heterogeneous data modalities. Comprehensive experiments on a dataset of 5,590 projects demonstrate that our system achieves an F1-score of 0.4546, surpassing the best-performing baseline (IR-TFIDF) by 23.40\%. Case studies further validate its superiority in recognizing technical terms, offering new insights for software comprehension on knowledge-intensive platforms.