Keywords: Relation Recognition from Text Algorithm, ACE Benchmark Task
Notes
- They observe that the addition of bigrams is helpful, but that the further addition of trigrams is not significantly helpful. A simple analysis shows that the lift in F-measure from unigram to bigram on average is x ~1.09 while the lift from the further addition of trigram is x ~1.02.
Quotes
Abstract
- "Relation extraction is the task of finding semantic relations between entities from text. The state-of-the-art methods for relation extraction are mostly based on statistical learning, and thus all have to deal with feature selection, which can significantly affect the classification performance. In this paper, we systematically explore a large space of features for relation extraction and evaluate the effectiveness of different feature subspaces. We present a general definition of feature spaces based on a graphic representation of relation instances, and explore three different representations of relation instances and features of different complexities within this framework. Our experiments show that using only basic unit features is generally sufficient to achieve state-of-the-art performance, while overinclusion of complex features may hurt the performance. A combination of features of different levels of complexity and from different sentence representations, coupled with task-oriented feature pruning, gives the best performance."
5 Experiments
5.1 Data Set and Experiment Setup
- "As in most existing work, instead of using the entire sentence, we used only the sequence of tokens that are inside the minimum complete subtree covering the two arguments.
5.2 General Search in the Feature Subspace
- "First, within each feature subspace, while bigram features improved the performance significantly over unigrams, trigrams did not improve the performance very much."
- "In the case of the syntactic parse tree subspace, adding production features even hurt the performance. This suggests that inclusion of complex features is not guaranteed to improve the performance."
- "However, the difference in performance between the syntactic parse tree subspace and the other two subspaces is not very large. This suggests that each feature subspace alone already captures most of the useful structural information between tokens for relation extraction.
- "The reason why the sequence feature subspace gave good performance although it contained the least structural information is probably that many relations defined in the ACE corpus are short-range relations, some within single noun phrases. For such kind of relations, sequence information may be even more reliable than syntactic or dependency information, which may not be accurate due to parsing errors.
References
- [Bunescu and Mooney, 2005] => R. Bunescu and R. J. Mooney. (2005). A Shortest Path Dependency Kernel for Relation Extraction. In Proc. of HLT/EMNLP-2005.
- Razvan C. Bunescu and Raymond J. Mooney. 2005b. Subsequence kernels for relation extraction. In Proceedings of NIPS.
- [Culotta and Sorensen, 2004] => A. Culotta and J. S. Sorensen. (2004). Dependency Tree Kernels for Relation Extraction. In Proc. of ACL-2004.
- Chad Cumby and Dan Roth. 2003. On kernel methods for relational learning. In Proceedings of ICML.
- [HasegawaSG, 1994] => T. Hasegawa, S. Sekine, and R. Grishman. (2004). Discovering Relations among Named Entities from Large Corpora. In Proc. of ACL-2004.
- [Kambhatla, 2004] => N. Kambhatla. (2004). Combining lexical, syntactic, and semantic features with maximum entropy models for extracting relations. Poster in Proc. ACL-2004
- [Moschitti, 2004]] => A. Moschitti. (2004). A study on Convolution Kernels for Shallow Semantic Parsing. In proceedings of the 42-th Conference on Association for Computational Linguistic (ACL-2004), Barcelona, Spain, 2004.
- Jun Suzuki, Tsutomu Hirao, Yutaka Sasaki, and Eisaku Maeda. 2003. Hierarchical directed acyclic graph kernel: Methods for structured natural language data. In Proceedings of ACL.
- Dmitry Zelenko, Chinatsu Aone, and Anthony Richardella. 2003. Kernel methods for relation extraction. Journal of Machine Learning Research, 3:1083–1106.
- [ZhangZS, 2006] => M. Zhang, J. Zhang, and J. Su. (2006). Exploring Syntactic Features for Relation Extraction using a Convolution Tree Kernel. In Proceedings of HLT-2006.
- [Zhang et al, 2006] => M. Zhang, J. Zhang, J. Su, and G. Zhou. (2006). A Composite Kernel to Extract Relations between Entities with Both Flat and Structured Features. In Proc. COLING-ACL 2006.
- [Zhao and Grishman, 2005] => S. Zhao and R. Grishman. (2005). Extracting Relations with Integrated Information Using Kernel Methods. In Proc. or ACL-2005.
- [Zhou et al, 2005] => G. Zhou, J. Su, J. Zhang, M. Zhang (2005) http://www.lsi.upc.edu/~surdeanu/seminars/guodong_acl05.pdf">Exploring Various Knowledge in Relation Extraction, In Proc. of ACL-2005.
BibTex
@InProceedings{jiang-zhai:2007:MainConf,
author = {Jiang, Jing and Zhai, ChengXiang},
title = {A Systematic Exploration of the Feature Space for Relation Extraction},
booktitle = {Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceedings of the Main Conference},
month = {April},
year = {2007},
address = {Rochester, New York},
publisher = {Association for Computational Linguistics},
pages = {113--120},
url = {http://www.aclweb.org/anthology/N/N07/N07-0115}}