A Metalearning Approach to Processing the Scope of Negation

Proceedings of the Thirteenth Conference on Computational Natural Language Learning (CoNLL), pages 21–29, Boulder, Colorado, June 2009. c⃝2009 Association for Computational Linguistics A metalearning approach to processing the scope of negation Roser Morante, Walter Daelemans CNTS - Language Technology Group University of Antwerp Prinsstraat 13, B-2000 Antwerpen, Belgium {Roser.Morante,Walter.Daelemans}@ua.ac.be Abstract Finding negation signals and their scope in text is an important subtask in information ex- traction. In this paper we present a machine learning system that finds the scope of nega- tion in biomedical texts. The system combines several classifiers and works in two phases. To investigate the robustness of the approach, the system is tested on the three subcorpora of the BioScope corpus representing different text types. It achieves the best results to date for this task, with an error reduction of 32.07% compared to current state of the art results. 1 Introduction In this paper we present a machine learning system that finds the scope of negation in biomedical texts. The system works in two phases: in the first phase, negation signals are identified (i.e., words indicating negation), and in the second phase the full scope of these negation signals is determined. Although the system was developed and tested on biomedical text, the same approach can also be used for text from other domains. Finding the scope of a negation signal means de- termining at sentence level the sequence of words in the sentence that is affected by the negation. This task is different from determining whether a word is negated or not. For a sentence like the one in Exam- ple (1) taken from the BioScope corpus (Szarvas et al., 2008), the system detects that lack, neither, and nor are negation signals; that lack has as its scope lack of CD5 expression, and that the discontinuous negation signal neither ... nor has as its scope nei- ther to segregation of human autosome 11, on which the CD5 gene has been mapped, nor to deletion of the CD5 structural gene. (1) <sentence id=“S334.5”>Analysis at the phenotype and genetic level showed that <xcope id“X334.5.3”><cue type=“negation” ref=“X334.5.3”>lack</cue> of CD5 expression</xcope> was due <xcope id=“X334.5.1”> <cue type=“negation” ref=“X334.5.1”>neither</cue> to segregation of human autosome 11, on which the CD5 gene has been mapped, <cue type=“negation” ref=“X334.5.1”>nor</cue> to deletion of the CD5 structural gene</xcope>.</sentence> Predicting the scope of negation is relevant for text mining and information extraction purposes. As Vincze et al. (2008) put it, extracted information that falls in the scope of negation signals cannot be pre- sented as factual information. It should be discarded or presented separately. Szarvas et al. (2008) report that 13.45% of the sentences in the abstracts section of the BioScope corpus and 12.70% of the sentences in the full papers section contain negations. A sys- tem that does not deal with negation would treat the facts in these cases incorrectly as positives. Addi- tionally, information about the scope of negation is useful for entailment recognition purposes. The approach to the treatment of negation in NLP presented in …

Content truncated to 500 words