Export
Minh C. Phan
1 previous affiliation
- Nanyang Technological University2015–2019
- PhD student · 2015–2019
Research Overview
Minh C. Phan focuses on natural language processing, information extraction, and representation learning. Early work addressed cross-device user linking and semantic matching for question-answering systems, while subsequent research explored collective named entity recognition and automated text scoring. The most recent available evidence indicates an emphasis on neural coherence features and entity linking within complex textual contexts.
Research Trajectory
The available evidence for Minh C. Phan's research spans a concentrated period between 2017 and 2018, primarily addressing challenges in applied machine learning, natural language processing, and information retrieval.
In 2017, Phan led research on user identification across heterogeneous hardware. As the first author of Cross-Device User Linking: URL, Session, Visiting Time, and Device-log Embedding, Phan proposed a learning-to-rank approach using gradient boosting and profile embeddings to match user identities based on browsing logs, URLs, and session data. This work framed the challenge as a pairwise classification problem, aligning with the broader research direction of Intra-Category Negative Sampling for Fine-Grained Representation Learning, where models are trained to distinguish between highly similar positive and negative pairs. During the same year, Phan contributed as a second author to Learning to Rank Question Answer Pairs with Holographic Dual LSTM Architecture. This research introduced the HD-LSTM, an architecture that integrated holographic composition into long short-term memory networks to improve semantic matching and deep sentence modeling without incurring prohibitive parameter costs.
Moving into 2018, Phan's focus shifted toward document-level natural language processing and information extraction. As the first author of CoNEREL: Collective Information Extraction in News Articles, Phan developed a system for collective named entity recognition and entity linking. By processing news articles and their associated reader comments in batch mode, the system leveraged shared contextual information and co-reference resolution to refine entity class labels. This approach contributed to the broader area of Mapping Lexical Variation to Standardized Domain Concepts, addressing the linguistic complexities of linking varied textual mentions to standardized knowledge bases like Wikipedia.
Additionally, Phan contributed as a second author to SkipFlow: Incorporating Neural Coherence Features for End-to-End Automatic Text Scoring. This work addressed the limitations of standard LSTM networks in processing long essay sequences by introducing a "SkipFlow" mechanism. The architecture modeled semantic relationships between snapshots of hidden representations, utilizing these relationships as auxiliary neural coherence features to improve automated text scoring.
Current Directions
The most recent papers in the provided evidence are from 2018. Based on these publications, Phan's latest documented research investigates document-level semantic modeling and collective information extraction. This involves developing neural architectures capable of maintaining coherence over long text sequences, as well as designing systems that leverage shared contexts across related documents to improve entity recognition and linking.
Profile updated
Loading new papers…
Other researchers whose papers cluster into the same research directions — a proxy for overlapping technical interests and shared intellectual neighborhoods.
See something wrong? Report a mistake