Yongkun Wang

Honglun Zhang, Liqiang Xiao, Wenqing Chen et al.

Multi-task learning in text classification leverages implicit correlations among related tasks to extract common features and yield performance gains. However, most previous works treat labels of each task as independent and meaningless one-hot vectors, which cause a loss of potential information and makes it difficult for these models to jointly learn three or more tasks. In this paper, we propose Multi-Task Label Embedding to convert labels in text classification into semantic vectors, thereby turning the original tasks into vector matching tasks. We implement unsupervised, supervised and semi-supervised models of Multi-Task Label Embedding, all utilizing semantic correlations among tasks and making it particularly convenient to scale and transfer as more tasks are involved. Extensive experiments on five benchmark datasets for text classification show that our models can effectively improve performances of related tasks with semantic representations of labels and additional information from each other.

Siwei Qiang, Yongkun Wang, Yaohui Jin

Micro-blogging services can track users' geo-locations when users check-in their places or use geo-tagging which implicitly reveals locations. This "geo tracking" can help to find topics triggered by some events in certain regions. However, discovering such topics is very challenging because of the large amount of noisy messages (e.g. daily conversations). This paper proposes a method to model geographical topics, which can filter out irrelevant words by different weights in the local and global contexts. Our method is based on the Latent Dirichlet Allocation (LDA) model but each word is generated from either a local or a global topic distribution by its generation probabilities. We evaluated our model with data collected from Weibo, which is currently the most popular micro-blogging service for Chinese. The evaluation results demonstrate that our method outperforms other baseline methods in several metrics such as model perplexity, two kinds of entropies and KL-divergence of discovered topics.