In the experiment, concept pairs in Chinese ConceptNet are applied as seeds after la-beled with the correctness while sentences in Sinica Corpus serve as the source. Although not all relations could be efficiently extracted at the first iteration, the wrong output could be revised by manually labeling and fed as the seed in next iteration. The mistake could be corrected promptly and enhance the performance of relation extraction.
We also provide the brief comparison of different MIL algorithms. stMIL algorithm models the situation that positive bags contain few positive instances, which fits our data and problem. Among all MIL algorithms we have mentioned, stMIL performs the best in our problem.
To sum up, we proposed a iteratively learning framework, which requires little human efforts and generates nearly correct new pairs related to several relations from a corpus.
Bibliography
[1] Y. S. Abu-Mostafa, M. Magdon-Ismail, and H.-T. Lin. Learning from data. AML-Book, 2012.
[2] J. Amores. Multiple instance classification: Review, taxonomy and comparative study. Artificial Intelligence, 201:81–105, 2013.
[3] S. Andrews, I. Tsochantaridis, and T. Hofmann. Support vector machines for multiple-instance learning. In S. Becker, S. Thrun, and K. Obermayer, editors, Ad-vances in Neural Information Processing Systems 15, pages 577–584. MIT Press, 2003.
[4] N. Bach and S. Badaskar. A survey on relation extraction. Literature review for Language and Statistics II, 2007.
[5] T. Berners-Lee, J. Hendler, O. Lassila, et al. The semantic web. Scientific american, 284(5):28–37, 2001.
[6] K. Bollacker, C. Evans, P. Paritosh, T. Sturge, and J. Taylor. Freebase: a collabo-ratively created graph database for structuring human knowledge. In Proceedings of the 2008 ACM SIGMOD international conference on Management of data, pages 1247–1250. ACM, 2008.
[7] S. Brin. Extracting patterns and relations from the world wide web. In The World Wide Web and Databases, pages 172–183. Springer, 1999.
[8] R. Bunescu and R. Mooney. Learning to extract relations from the web using mini-mal supervision. In Annual meeting-association for Computational Linguistics, vol-ume 45, page 576, 2007.
[9] R. C. Bunescu and R. J. Mooney. Multiple instance learning for sparse positive bags. In Proceedings of the 24th international conference on Machine learning, pages 105–112. ACM, 2007.
[10] W. Che, J. Jiang, Z. Su, Y. Pan, and T. Liu. Improved-edit-distance kernel for chinese relation extraction. In Proceedings of IJCNLP, pages 132–137, 2005.
[11] Y. Chen, Q. Zheng, and W. Zhang. Omni-word feature and soft constraint for chinese relation extraction. In Proceedings of the 52nd Annual Meeting of the Associ-ation for Computational Linguistics, pages 572–581, 2014.
[12] R. Davis, H. Shrobe, and P. Szolovits. What is a knowledge representation? AI magazine, 14(1):17, 1993.
[13] G. Doran and S. Ray. A theoretical and empirical analysis of support vector machine methods for multiple-instance classification. Machine Learning, 97(1-2):79–102, 2014.
[14] R. Hoffmann, C. Zhang, X. Ling, L. Zettlemoyer, and D. S. Weld. Knowledge-based weak supervision for information extraction of overlapping relations. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies-Volume 1, pages 541–550. Association for Computational Linguistics, 2011.
[15] R. Huang, L. Sun, and Y. Feng. Study of kernel-based methods for chinese relation extraction. Information Retrieval Technology, pages 598–604, 2008.
[16] Y.-L. Kuo and J. Y.-j. Hsu. Resource-bounded crowd-sourcing of commonsense knowledge. In IJCAI Proceedings-International Joint Conference on Artificial In-telligence, volume 22, page 2470, 2011.
[17] W. Li, P. Zhang, F. Wei, Y. Hou, and Q. Lu. A novel feature-based approach to chinese entity relation extraction. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies:
Short Papers, pages 89–92. Association for Computational Linguistics, 2008.
[18] D. Liu, Z. Zhao, Y. Hu, and L. Qian. Incorporating lexical semantic similarity to tree kernel-based chinese relation extraction. In Chinese Lexical Semantics, pages 11–21. Springer, 2013.
[19] H. Liu and P. Singh. Conceptnet—a practical commonsense reasoning tool-kit. BT technology journal, 22(4):211–226, 2004.
[20] O. L. Mangasarian and E. W. Wild. Multiple instance classification via successive linear programming. Journal of Optimization Theory and Applications, 137(3):555–
568, 2008.
[21] M. Minsky. A framework for representing knowledge. The Psychology of Computer Vision, pages 211–277, 1975.
[22] M. Mintz, S. Bills, R. Snow, and D. Jurafsky. Distant supervision for relation extrac-tion without labeled data. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volume 2 - Volume 2, pages 1003–1011. Association for Computational Linguistics, 2009.
[23] D. L. Poole and A. K. Mackworth. Artificial Intelligence: foundations of computa-tional agents. Cambridge University Press, 2010.
[24] L. Qiu and Y. Zhang. Zore: A syntax-based system for chinese open relation ex-traction. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, pages 1870–1880. Association for Computational Linguistics, 2014.
[25] S. Ray and M. Craven. Supervised versus multiple instance learning: An empiri-cal comparison. In Proceedings of the 22nd international conference on Machine learning, pages 697–704. ACM, 2005.
[26] S. Riedel, L. Yao, and A. McCallum. Modeling relations and their mentions without labeled text. In Machine Learning and Knowledge Discovery in Databases, pages 148–163. Springer, 2010.
[27] B. Schölkopf and A. J. Smola. Learning with kernels: Support vector machines, regularization, optimization, and beyond. MIT press, 2002.
[28] R. Snow, D. Jurafsky, and A. Y. Ng. Learning syntactic patterns for automatic hy-pernym discovery. Advances in Neural Information Processing Systems 17, 2004.
[29] M. Surdeanu, J. Tibshirani, R. Nallapati, and C. D. Manning. Multi-instance multi-label learning for relation extraction. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natu-ral Language Learning, pages 455–465. Association for Computational Linguistics, 2012.
[30] L. Yao, S. Riedel, and A. McCallum. Collective cross-document relation extraction without labelled data. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1013–1023. Association for Computational Linguistics, 2010.
[31] D. Zelenko, C. Aone, and A. Richardella. Kernel methods for relation extraction.
The Journal of Machine Learning Research, 3:1083–1106, 2003.
[32] G. Zhou, J. Su, J. Zhang, and M. Zhang. Exploring various knowledge in relation extraction. In Proceedings of the 43rd annual meeting on association for computa-tional linguistics, pages 427–434. Association for Computacomputa-tional Linguistics, 2005.
[33] Z.-H. Zhou and J.-M. Xu. On the relation between multi-instance learning and semi-supervised learning. In Proceedings of the 24th international conference on Machine learning, pages 1167–1174. ACM, 2007.
Appendix A
List of CKIP Part-Of-Speech Tag
The CKIP (Chinese Knowledge and Information Processing) part-of-speech tags are de-fined by CKIP team, Academia Sinica.1 The original document contains the simplified tag and syntactical functions. We explain each tag with English description as the third column in the following table.
Simplified Tag syntactical Function English Description
A 非謂形容詞 adjective, which could not be used
as a predicate
Caa 對等連接詞，如：和、跟 coordinating conjunction (e.g. and)
Cab 連接詞，如：等等 enumeration conjunction (e.g. and
so on)
Cba 連接詞，如：的話 conjunction of conditional clause
Cbb 關聯連接詞 relational conjunction (e.g.
al-though, but)
Da 數量副詞 adverb of amount
Dfa 動詞前程度副詞 adverb of degree, used before verb
Dfb 動詞後程度副詞 adverb of degree, used after verb
Di 時態標記 tense
Dk 句副詞 adverb on the start of sentence (e.g.
however)
D 副詞 adverb
Na 普通名詞 common noun
1中研院平衡語料庫詞類標記集: http://ckipsvr.iis.sinica.edu.tw/papers/category_list.doc
Nb 專有名稱 proper noun
Nc 地方詞 location
Ncd 位置詞 position
Nd 時間詞 time
Neu 數詞定詞 number modifier
Nes 特指定詞 particularly referring modifier,
modifying noun or number
Nep 指代定詞 referring modifier
Neqa 數量定詞 amount modifier
Neqb 後置數量定詞 amount modifier, used after the
amount term
Nf 量詞 classifier
Ng 後置詞 post-position
Nh 代名詞 pronoun
Nv 名物化動詞 nominalization
I 感嘆詞 interjection
P 介詞 preposition
T 語助詞 particle
VA 動作不及物動詞 intransitive verb of action
VAC 動作使動動詞 causative verb of action
VB 動作類及物動詞 intransitive-like verb of action,
which requires a preposition before the object
VC 動作及物動詞 transitive verb of action
VCL 動作接地方賓語動詞 verb connecting action and location
VD 雙賓動詞 verb referring to double objects
VE 動作句賓動詞 action verb referring to a clause
ob-ject
VF 動作謂賓動詞 action verb referring to a phrase
VG 分類動詞 classifying verb, connecting subject
and object
VH 狀態不及物動詞 intransitive verb of status
VHC 狀態使動動詞 causative verb of status
VI 狀態類及物動詞 intransitive-like verb of status, which requires a preposition before the object
VJ 狀態及物動詞 transitive verb of status
VK 狀態句賓動詞 status verb referring to a clause
ob-ject
VL 狀態謂賓動詞 status verb referring to a phrase
V_2 有 have
DE 的, 之, 得, 地 syntactic expletive, performing
syn-tactic role but conveying no mean-ing
SHI 是 be verb
FW 外文標記 foreign term
COLONCATEGORY 冒號 colon
COMMACATEGORY 逗號 comma
DASHCATEGORY 破折號 dash
ETCCATEGORY 刪節號 ellipsis
EXCLAMATIONCATEGORY 驚嘆號 exclamation mark
PARENTHESISCATEGORY 括弧 parenthesis
PAUSECATEGORY 頓號 pause mark
PERIODCATEGORY 句號 period mark
QUESTIONCATEGORY 問號 question mark
SEMICOLONCATEGORY 分號 semi-colon
SPCHANGECATEGORY 雙直線 double line
Appendix B
List of Relations in Chinese ConceptNet
A set of relations are defined in ConceptNet 51, along with the descriptions. In the table below, we illustrate the relations used in Chinese ConceptNet. To make clear, each relation is followed by an example.
Relation Type Description Example
MotivatedByGoal
Someone does A because they want result B; A is a step toward accomplishing the goal B.
工作, 賺錢
work, make money
HasProperty A has B as a property; A can be described as B.
身體, 健康 body, health CapableOf Something that A can typically do is B.
學生, 學習 student, learn HasFirstSubevent A is an event that begins with subevent B.
回家, 搭車
go home, take a bus
Desires
A is a conscious entity that typically wants B. Many assertions of this type use the ap-propriate language’s word for “person” as A.
學生, 放假 student, vacation
HasSubevent
A and B are events, and B happens as a subevent of A.
進修, 出國
further education, go abroad
1Relations in ConceptNet 5: https://github.com/commonsense/conceptnet5/wiki/Relations