论文信息 - Using AOBP for Definitional Question Answering

Using AOBP for Definitional Question Answering

This paper presents an integrated system for the task of definitional question answering. Firstly, we extract question-related knowledge as much as possible which include 3 categories. The first is based on language model. We train our language model on four different corpora. The second resource is the syntax dependency relations, which are extracted by Minipar. And the third resource contains only one feature, the document score provided by Information Retrieval (IR) engine. After that, we use a novel Adaptive Optical Back-Propagation (AOBP) neural network to score candidate sentences by using above extracted knowledge. The topk candidates will be selected as the final question answers. We experiment on the task of definitional question answering in TREC2006&2005. The experimental results show that our method can greatly improve the performance.

Weihua Wang | Junkuo Cao | Yuanzhong Shu

[1] Xuanjing Huang,et al. FDUQA on TREC2005 QA Track , 2005 .

[2] Frank Fallside,et al. An adaptive training algorithm for back propagation networks , 1987 .

[3] Martin A. Riedmiller,et al. A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[4] George D. Magoulas,et al. Effective Backpropagation Training with Variable Stepsize , 1997, Neural Networks.

[5] Young-In Song,et al. Probabilistic model for definitional question answering , 2006, SIGIR '06.

[6] Geoffrey E. Hinton,et al. Learning internal representations by error propagation , 1986 .

[7] Robert A. Jacobs,et al. Increased rates of convergence through learning rate adaptation , 1987, Neural Networks.

[8] Roberto Battiti,et al. Accelerated Backpropagation Learning: Two Optimization Methods , 1989, Complex Syst..