Publications

Wang D., Lin C., Zhong L., and Wong KF.

NAACL 2021

We present a fast and scalable architecture, in which we incorporate both classification based and extraction-based approaches and then elaborately design four modules to jointly extract dialogue states. Experimental results on MultiWoz2.0 validate the superiority of our proposed model in terms of complexity and scalability when compared with state-of-the art methods, especially in multi-domain dialogues entangled with many turn utterances.

Peng X., Lin C., and Stevenson M.

NAACL 2021

Based on the robust ℓ1 norm goodness-of-fit criterion, we propose a simple post-processing step to improve CLWEs. It is fully agnostic to the training process of the original CLWEs and can therefore be applied widely. Extensive experimental results based on bilingual lexicon induction and cross-lingual transfer for natural language inference tasks show that our method substantially outperforms state-of-the-art baselines.

Peng X., Chen G., Lin C., and Stevenson M.

NAACL 2021

We propose a simple yet effective KGE framework which can reduce the training time and carbon footprints by orders of magnitudes compared with state-of-the-art approaches, while retaining competitive performance. We highlight three technical innovations, namely full batch learning via relational matrices, closed-form Orthogonal Procrustes Analysis for KGEs, and non-negative-sampling training. In addition, as the first KGE method whose entity embeddings also store full relation information, our trained models encode rich semantics and are highly interpretable.

Peng X., Zheng Y., Lin C., and Siddharthan A.

EACL 2021

We propose a hitherto unexplored and challenging task of historical text summarisation. To kickstart research in this field, we construct a high-quality summarisation corpus for historical German and Chinese, with modern German and Chinese summaries by experts. We also propose a model for historical text summarisation that does not require parallel supervision and provides a validated high-performing baseline for future studies.

Li X., Lin C., Li R., Wang C., and Guerin F.

ICML 2020

We tackle the problem disentangling the latent space of an autoencoder in order to separate labelled attribute information from other characteristic information. This then allows us to change selected attributes while preserving other information. Our method, matrix subspace projection, is much simpler than previous approaches to latent space factorisation, for example not requiring multiple discriminators or a careful weighting among their loss functions.

Li X., Chen G., Lin C. and Li R.

EMNLP 2020

We propose DGST, a novel and simple Dual-Generator network architecture for text Style Transfer. Our model employs two generators only, and does not rely on any discriminators or parallel corpus for training. Both quantitative and qualitative experiments on the Yelp and IMDb datasets show that our model gives competitive performance compared to several strong baselines with more complicated architecture designs.

Li R., Li X., Chen G. and Lin C.

COLING 2020

In this paper, we propose a simple, generic architecture called Timestep-Wise Regularisation VAE (TWR-VAE), which can effectively avoid posterior collapse and can be applied to any RNN-based VAE models. The effectiveness and versatility of our model are demonstrated in different tasks, including language modelling and dialogue response generation.

Mao R., Lin C. and Guerin F.

ACL 2019

End-to-end training with Deep Neural Networks (DNN) is a currently popular method for metaphor identification. However, standard sequence tagging models do not explicitly take advantage of linguistic theories of metaphor identification. We experiment with two DNN models which are inspired by two human metaphor identification procedures. By testing on three public datasets, we find that our models achieve state-of-the-art performance in end-to-end metaphor identification.

Li R., Lin C., Collinson M., Li X. and Chen G.

CoNLL 2019

Recognising dialogue acts (DA) is important for many natural language processing tasks such as dialogue generation and intention recognition. In this paper, we propose a dual-attention hierarchical recurrent neural network for DA classification. Our model is partially inspired by the observation that conversational utterances are normally associated with both a DA and a topic, where the former captures the social act and the latter describes the subject matter.

Li R., Li X., Lin C, Collinson M. and Mao R.

INLG 2019

In this paper, we present a new architecture called Full-Sampling-VAE-RNN, which can effectively avoid latent variable collapse. Compared to the general VAE-RNN architectures, we show that our model can achieve much more stable training process and can generate text with significantly better quality.

(Best Paper Award Runner-up)

Mao R., Lin C. and Guerin F.

ACL 2018

In this paper, we propose an unsupervised learning method that identifies and interprets metaphors at word-level without any preprocessing, outperforming strong baselines in the metaphor identification task. Our model extends to interpret the identified metaphors, paraphrasing them into their literal counterparts, so that they can be better translated by machines. We evaluated this with two popular translation systems for English to Chinese, showing that our model improved the systems significantly.