Tan Wentao / 谭文韬

Tan Wentao / 谭文韬

Focusing on information retrieval.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

General Method of Forum Information Extraction

Published in Journal of Chinese Computer Systems (Chinese), 1900

Abstract: The classification and information extraction of online forums are two important technologies of online data mining. The traditional web page classification methods do not take the structure features of their URLs into full account. They are often based on the characteristics of the content,therefore, they are susceptible to noise,of low efficiency and they can’t meet the needs of versatility. The traditional information extraction methods are based on text density and layout structure,ignoring the semantic information of the content. They are difficult to extract the content from a variety of forums effectively. This paper proposes a clustering method based on URLs’ structure (USC) and a filter method based on keyword scoring (KSF). Both methods only need to analyze a small number of samples in the data set and extract general rules to meet the demand of large-scale extraction. In the same data set,the F value of the USC method is 18.99% higher than that of the traditional classification method,and the accuracy of the KSF method is 18.46% higher than that of the traditional information extraction method.

Download here

Efficient Inter-image Relation Graph Neural Network Hashing for Scalable Image Retrieval

Published in ACM MMAsia, 1900

Abstract: Unsupervised deep hashing is a promising technique for largescale image retrieval, as it equips powerful deep neural networks and has advantage on label independence. However, the unsupervised deep hashing process needs to train a large amount of deep neural network parameters, which is hard to optimize when no labeled training samples are provided. How to maintain the well scalability of unsupervised hashing while exploiting the advantage of deep neural network is an interesting but challenging problem to investigate. With the motivation, in this paper, we propose a simple but effective Inter-image Relation Graph Neural Network Hashing (IRGNNH) method. Different from all existing complex models, we discover the latent inter-image semantic relations without any manual labels and exploit them further to assist the unsupervised deep hashing process. Specifically, we first parse the images to extract latent involved semantics. Then, relation graph convolutional network is constructed to model the inter-image semantic relations and visual similarity, which generates representation vectors for image relations and contents. Finally, adversarial learning is performed to seamlessly embed the constructed relations into the image hash learning process, and improve the discriminative capability of the hash codes. Experiments demonstrate that our method significantly outperforms the state-of-the-art unsupervised deep hashing methods on both retrieval accuracy and efficiency.

Download here

Integrated Prediction Method for Mental Illness with Multimodal Sleep Function Indicators

Published in ICA3PP, 1900

Abstract: Sleep quality has great effect on physical and mental health. Severe insomnia will cause autonomic neurological dysfunction. For making good clinical decisions, it is crucial to extract features of sleep quality and accurately predict the mental illness. Prior studies have a number of deficiencies to be overcome. On the one hand, the selected features for sleep quality are not good enough, as they do not account for multisource and heterogeneous features. On the other hand, the mental illness prediction model does not work well and thus needs to be enhanced and improved. This paper presents a multi-dimensional feature extraction method and an ensemble prediction model for mental illness. First, we do correlation analysis for every indicators and sleep quality, and further select the optimal heterogeneous features. Next, we propose a combinational model, which is integrated by basic modules according to their weights. Finally, we perform abundant experiments to test our method. Experimental results demonstrate that our approach outperforms many state-of-the-art approaches.

Download here

talks

teaching