Elamur Research
☰
home / papers / 2007 / Open Information Extraction from the Web

Open Information Extraction from the Web

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "© IJCAI; open access at ijcai.org — read only" does not allow redistributing the text itself. Original — ijcai.org/Proceedings/07/Papers/429.pdf

Summary

Classical information extraction (IE) systems require a human to name every relation of interest in advance and to prepare rules or labeled examples for it. On the web, where texts are highly diverse and the number of relations is unknown, this approach does not scale. The authors propose a new paradigm, Open IE: the system makes a single pass over the corpus and, without human involvement, collects triples of the form (entity, relation, entity) for arbitrary relations. The idea is implemented in TextRunner, a system with three modules. The self-supervised learner uses a syntactic parser only to label training examples automatically and trains a Naive Bayes classifier on them. The extractor does without a parser and needs only POS tags and a lightweight noun-phrase chunker. The assessor estimates the probability of each triple from how often it recurs across texts. The system was evaluated on 9 million web pages. Compared with KnowItAll, it made one third fewer errors and found orders of magnitude more relations. The work laid the foundation for open fact extraction at large scale.

Key points

  • The corpus contains about 9 000 000 web pages. On a comparable set of extractions, TextRunner makes 33% fewer errors than KnowItAll at similar recall.
  • Among the 11 million highest-probability triples, more than 1 million are concrete facts and more than 6.5 million are more abstract assertions. Roughly 92% of the triples express distinct assertions, although the authors acknowledge that this estimate is inflated.
  • The parser of Klein & Manning (2003) was used to label the training data. Maxent models from OpenNLP were used for POS tagging and chunking. A hand-built classifier achieved only one third of the precision of the trained one.
  • The running time is estimated at O(D + T log T). By comparison, the approach of Shinyama & Sekine requires O(D²) and discovered only 101 relations from 28 000 news articles.
  • As future work, the authors list detecting relation synonymy, resolving multiple mentions of the same entity, determining argument types, and merging triples into a graph.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: AI & machine learning

Topics: Part-of-speech tagging

Cited by

447 papers of our library cite it (of the 140673 whose bibliographies we read); 14 of them have a page here, 433 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{banko2007open,
  title = {{Open Information Extraction from the Web}},
  author = {Banko, Michele and Cafarella, Michael J. and Soderland, Stephen and Broadhead, Matt and Etzioni, Oren},
  booktitle = {IJCAI 2007},
  year = {2007},
  url = {https://www.ijcai.org/Proceedings/07/}
}

References

22 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. E. Agichtein and L. Gravano. Snowball: Extracting relations from large plain-text collections. In Proceedings of the Fifth ACM International Conference on Digital Libraries, 2000. find
  2. E. Brill and G. Ngai. Man (and woman) vs. machine: a case study in base noun phrase learning. In Proceedings of the ACL, pages 65–72, 1999. find
  3. S. Brin. Extracting Patterns and Relations from the World Wide Web. In WebDB Workshop at 6th International Conference on Extending Database Technology, EDBT’98, pages 172–183, Valencia, Spain, 1998. find
  4. R. Bunescu and R. Mooney. A shortest path dependency kernel for relation extraction. In Proc. of the HLT/EMLNP, 2005. find
  5. Michael J. Cafarella, Michele Banko, and Oren Etzioni. Relational web search. Technical Report 06-04-02, University of Washington, 2006. find
  6. M. Ciaramita, A. Gangemi, E. Ratsch, J. Saric, and I. Rojas. Unsupervised learning of semantic relations between concepts of a molecular biology ontology. In Proceedings of IJCAI, 2005. find
  7. M. Craven, D. DiPasquo, D. Freitag, A. McCallum, T. Mitchell, K. Nigam, and S. Slattery. Learning to construct knowledge bases from the world wide web. In Artificial Intelligence, 1999. find
  8. A. Culotta, A. McCallum, and J. Betz. Integrating probabilistic extraction models and relational data mining to discover relations and patterns in text. In Proceedings of HLT-NAACL, New York, NY, 2006. find
  9. D. Downey, O. Etzioni, and S. Soderland. A Probabilistic Model of Redundancy in Information Extraction. In Proc. of IJCAI, 2005. find
  10. D. Downey, M. Broadhead, and O. Etzioni. Locating Complex Named Entities in Web Text. In Proc. of IJCAI, 2007. find
  11. O. Etzioni, M. Cafarella, D. Downey, S. Kok, A. Popescu, T. Shaked, S. Soderland, D. Weld, and A. Yates. Unsupervised named-entity extraction from the web: An experimental study. Artificial Intelligence, 165(1):91–134, 2005. find
  12. N. Kambhatla. Combining lexical, syntactic and semantic features with maximum entropy models. In Proceedings of ACL, 2004. find
  13. Dan Klein and Christopher D. Manning. Accurate unlexicalized parsing. In Proceedings of the ACL, 2003. find our card
  14. D. Lin and P. Pantel. Discovery of inference rules from text. In Proceedings of KDD, 2001. find
  15. G. Ngai and R. Florian. Transformation-based learning in the fast lane. In Proceedings of the NAACL, pages 40–47, 2001. find
  16. M. Pasca, D. Lin, J. Bigham, A. Lif-chits, and A. Jain. Names and similarities on the web: Fact extraction in the fast lane. In (To appear) Proc. of ACL/COLING 2006, 2006. find
  17. A. Ratnaparkhi. Maximum Entropy Models for Natural Language Ambiguity Resolution. PhD thesis, University of Pennsylvania, 1998. find
  18. D. Ravichandran and D. Hovy. Learning surface text patterns for a question answering system. In Proceedings of the ACL, pages 41–47, Philadelphia, Pennsylvania, 2002. find
  19. E. Riloff. Automatically constructing extraction patterns from untagged text. In Proc. of AAAI, 1996. find
  20. B. Rosario and M. Hearst. Classifying semantic relations in bioscience text. In Proc. of ACL, 2004. find
  21. S. Sekine. On-demand information extraction. In Procs. of COLING, 2006. find
  22. Y. Shinyama and S. Sekine. Preemptive information extraction using unrestricted relation discovery. In Proc. of the HLT-NAACL, 2006. IJCAI-07 2676 find