Haochen TAN

I'm a PhD student in the Computer Science Department of City University of Hong Kong in HKSAR. My PhD supervisor is Prof. Linqi Song. Currently I am a research intern at Noah's Ark Lab under the supervision of Zhijiang Guo.

At CityU, I work on the sentence comprehension, long sequence compression and evaluation. Before joined Prof. Song's team, I was a research assistant under the supversion of Dr. Bernard Chiu. In my undergraduate years, I was advised by Prof. Cong Fan at UESTC.

Email  /  CV  /  Scholar  /  Github

Research

I'm interested in natural language processing, deep learning, generative AI. Most of my research is about sentence representation, lenghty sequence compression, long-form generation and evaluation.

In this paper, we propose PROXYQA, an innovative framework dedicated to the assessment of long-text generation. PROXYQA comprises in-depth human-curated metaquestions spanning various domains, each accompanied by specific proxy-questions with pre-annotated answers.

we propose a two-step framework, Reconstruct before Summarize (RbS), for effective and efficient meeting summarization. RbS first leverages a self-supervised paradigm to annotate essential contents by reconstructing the meeting transcripts. Secondly, we propose a relative positional bucketing (RPB) algorithm to equip (conventional) summarization models to generate the summary.

we identify two properties in dialogue modeling, i.e., locality and isotropy, and present a simple method for dialogue representation calibration, namely SimDRC, to build isotropic and conversational feature spaces. Experimental results show that our approach significantly outperforms the current state-of-the-art models on three dialogue tasks across the automatic and human evaluation metrics.

To avoid expensive data collection and error-propagation of translation-based methods, we present a simple but effective approach to perform zero-shot cross-lingual CSRL. Our model implicitly learns language-agnostic, conversational structure-aware and semantically rich representations with the hierarchical encoders and elaborately designed pre-training objectives.

In this paper, we propose a semantic-aware contrastive learning framework for sentence embeddings, termed Pseudo-Token BERT (PT-BERT), which is able to explore the pseudo-token space (i.e., latent semantic space) representation of a sentence while eliminating the impact of superficial features such as sentence length and syntax.

Existing attentive models attend to all words without prior focus, which results in inaccurate concentration on some dispensable words. In this paper, we propose to use semantic role labeling (SRL), which highlights the core semantic information of who did what to whom, to provide additional guidance for the rewriter model.

Miscellanea

Updated at July 2023, Thanks Jon Barron for this amazing template.