Yueran Zu’s research while affiliated with Beihang University and other places

Keyphrase Generation compresses a document into some highly-summative phrases, which is an important task in natural language processing. Most state-of-the-art adopt greedy search or beam search decoding methods. These two decoding methods generate a large number of duplicated keyphrases and are time-consuming. Moreover, beam search only predicts a fixed number of keyphrases for different documents. In this paper, we propose an adaptive generation model-AdaGM, which is mainly inspired by the importance of the first words in keyphrase generation. In AdaGM, a novel reset state training mechanism is proposed to maximize the difference in the predicted first words. To ensure the discreteness and get an appropriate number of keyphrases according to the content of the document adaptively, we equip beam search with a highly effective filter mechanism. Experiments on five public datasets demonstrate the proposed model can generate marginally less duplicated and more accurate keyphrases. The codes of AdaGM are available at: https://github.com/huangxiaolist/adaGM.

Recently proposed models such as BERT, perform well in many text processing tasks. They get context-sensitive features, which is a good semantic for word sense disambiguation, through deeper layer and a large number of texts. But, for Chinese text classification, majority of datasets are crawled from social networking sites, these datasets are semantically complex and variable. How much data is needed to pre-train these models in order for them to grasp semantic features and understand context is a question. In this paper, we propose a novel shallow layer language model, which uses sememe information to guide model to grasp semantic information without a large number of pre-trained data. Then, we use the Chinese character representations generated from this model to do text classification. Furthermore, in order to make Chinese as easy to initialize as English, we employ convolution neural networks over Chinese strokes to get Chinese character structure initialization for our model. This model pre-trains on a part of the Chinese Wikipedia dataset, and we use the representations generated by this pre-trained model to do text classification. Experiments on text classification datasets show our model outperforms other state-of-arts models by a large margin. Also, our model is superior in terms of interpretability due to the introduction of semantic and morphological information.

Modern sparse-to-dense optical flow estimation algorithms usually achieve state-of-art performance. Those algorithms need two steps: matching and interpolation. Matching is often unreliable for very large displacement optical flow due to illumination changes, deformations and occlusion etc. Moreover, conspicuous errors around motion discontinuities still keep serious as most methods consider edge only at interpolation step. The context-adaptive matching (CAM) is proposed for optical flow which is better at large displacement and edge preserving. The CAM is selective in feature extraction, adaptive in flow propagation and search radius adjusting. Selective features are proposed to consider edge preserving in matching step. Except for the usually used SIFT descriptor, the local directional pattern flow (LDPF) is introduced to keep more edge structure, and the oriented fast and rotated brief (ORB) is utilized to select out several most similar candidates. Unlike coarse-to-fine matching, which proposed a propagation step with only neighbors, we propose adaptive propagation to extend the matching candidates in order to improve the possibility of getting right correspondences. Furthermore, guided by prior knowledge and taking advantage of upper layers results, adaptive radius instead of constrained radius are proposed at finer layers. The CAM interpolated by EpicFlow is fast and robust for large displacements especially for fast moving objects and also preserves the edge structure well. Extensive experiments show that our algorithm is on par with the state-of-art optical flow methods on MPI-Sintel, KITTI and Middlebury.

Optical flow estimation is still an open question in computer vision. Matching is the initialization of the final optical flow results. A good matching is important for the flow. In this paper, a combined local-global matching method is proposed. The local matching method and the global method are integrated together to make a trade-off between the large displacement and local consistency of optical flow. Extensive experiments on state-of-art challenging datasets MPI-Sintel show that the proposed method is efficient and effective.