Jointly Identifying Entities and Extracting Relations in Encyclopedia Text via A Graphical Model Approach
Xiaofeng Yu, Wai Pang Lam · 2010
In this paper, we investigate the problem of entity identification and relation extraction from encyclopedia articles, and we propose a joint discriminative probabilistic model with arbitrary graphical structure to optimize all relevant subtasks simultaneously. This modeling offers a natural formalism for exploiting rich dependencies and interactions between relevant subtasks to capture mutual benefits, as well as a great flexibility to incorporate a large collection of arbitrary, overlapping and nonindependent features. We show the parameter estimation algorithm of this model. Moreover, we propose a new inference method, namely collective iterative classification (CIC), to find the most likely assignments for both entities and relations. We evaluate our model on real-world data from Wikipedia for this task, and compare with current state-of-the-art pipeline and joint models, demonstrating the effectiveness and feasibility of our approach. 1