Elamur Research⌕
☰
home / papers / 2011 / Putting it Simply: a Context-Aware Approach to Lexical Simplification

Putting it Simply: a Context-Aware Approach to Lexical Simplification

Summary

The paper tackles lexical simplification: replacing difficult words in a sentence with easier equivalents without breaking grammar or changing meaning. The authors construct an unsupervised pipeline from two corpora that are comparable but not aligned: Simple English Wikipedia (SEW) and its regular counterpart, English Wikipedia (EW). They first harvest candidate word pairs. A pair is kept when WordNet links the two words as synonym or hypernym and context-vector similarity supports it. They retain only pairs where the replacement is less complex, scoring complexity by the word's frequency ratio across the two corpora multiplied by its length. MorphAdorner then expands each pair into matching inflected forms. At application time, a rule fires only when the sentence context agrees with it. This is checked both against the word itself, to avoid simplifying deliberately specific wording, and against the shared context of the pair, to avoid the wrong sense. Human judges compared the system with a most-frequent-synonym baseline. It simplified far more effectively and preserved meaning and grammar somewhat better. The work shows that simplification rules can be learned without parallel data, provided an in-domain simple corpus exists.

Key points

  • Data: an April 2010 snapshot with 3,266,245 EW articles and 60,100 SEW articles, meaning only about 2% of EW has simple counterparts. No article alignment or edit histories are used.

  • Context vectors: built from a 10-token co-occurrence window, with counts below 2 discarded. Rules fire only under two conditions:

    • word–sentence cosine is at or below 0.1;
    • cosine between the rule's shared context and the sentence context exceeds 0.01.

    Sentences with fewer than seven content words are skipped.

  • Evaluation set: 65 sentences, each simplified by both system and baseline (130 examples), rated by three native-speaker annotators. Inter-annotator agreement was moderate, with Cohen's Kappa ranging from about .35 to .53 depending on the criterion.

  • Overall results (system vs. baseline):

    CriterionSystemBaseline
    Simplification75.58%46.43%
    Meaning preserved62.79%55.95%
    Grammaticality ("good")77.91%70.23%

    Only the simplification gap is statistically significant (p < 0.001).

  • Frequency bands: meaning preservation peaks for medium-frequency words and falls for rare ones (48.15%), which the authors attribute to data sparsity.

Interesting ideas

  • I1. Complexity from corpus frequency ratios: comparing how often a word appears in a general corpus versus a simple one gives a cheap, domain-specific difficulty score. This could be reproduced for Ukrainian with any pair of "plain" and "standard" text collections.
  • I2. Shared-context vector for the rule: taking the element-wise minimum of the two words' context vectors keeps only the contexts both words share. Matching the sentence against this vector is a lightweight sense check that works without explicit word-sense disambiguation.
  • I3. Veto when the word fits its context too well: high similarity between a word and its sentence suggests the specific term was chosen on purpose. This "don't simplify" signal could transfer to paraphrasing or terminology normalization.
  • I4. Generating inflected rule pairs from lemmas: expanding each pair into aligned morphological forms rules out ungrammatical substitutions by construction. This is especially valuable for richly inflected languages like Ukrainian.
  • I5. Reporting results by frequency band: splitting the evaluation this way exposes where a data-driven method becomes unreliable. It also supports a practical policy of leaving rare words untouched.

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: Machine translation · Natural language processing

Cited by

87 papers of our library cite it (of the 140673 whose bibliographies we read); 8 of them have a page here, 79 not yet.

Cite

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

@inproceedings{biran2011putting,
  title = {{Putting it Simply: a Context-Aware Approach to Lexical Simplification}},
  author = {Biran, Or and Brody, Samuel and Elhadad, Noémie},
  booktitle = {Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies},
  year = {2011},
  doi = {10.7916/d8cv4s20},
  url = {https://doi.org/10.7916/d8cv4s20}
}

References

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

  1. Androutsopoulos, Ion and Prodromos Malakasiotis. 2010. A survey of paraphrasing and textual entailment methods. Journal of Artificial Intelligence Research 38:135–187. ↗
  2. Barzilay, Regina and Noemie Elhadad. 2003. Sentence alignment for monolingual comparable corpora. In Proc. EMNLP. pages 25–32. find
  3. Blake, Catherine, Julia Kampov, Andreas Orphanides, David West, and Cory Lown. 2007. Query expansion, lexical simplification, and sentence selection strategies for multi-document summarization. In Proc. DUC. find
  4. Carroll, John, Guido Minnen, Yvonne Canning, Siobhan Devlin, and John Tait. 1998. Practical simplication of english newspaper text to assist aphasic readers. In Proc. AAAI Workshop on Integrating Artificial Intelligence and Assistive Technology. find
  5. Chandrasekar, R., Christine Doran, and B. Srinivas. 1996. Motivations and methods for text simplification. In Proc. COLING. find our card
  6. Daelemans, Walter, Anja Hthker, and Erik Tjong Kim Sang. 2004. Automatic sentence simplification for subtitling in Dutch and English. In Proc. LREC. pages 1045–1048. find
  7. Deléger, Louise and Pierre Zweigenbaum. 2009. Extracting lay paraphrases of specialized expressions from monolingual comparable medical corpora. In Proc. Workshop on Building and Using Comparable Corpora. pages 2–10. find
  8. Devlin, Siobhan and Gary Unthank. 2006. Helping aphasic people process online information. In Proc. ASSETS. pages 225–226. find
  9. Elhadad, Noemie and Komal Sutaria. 2007. Mining a lexicon of technical terms and lay equivalents. In Proc. ACL BioNLP Workshop. pages 49–56. find
  10. Fellbaum, Christiane, editor. 1998. WordNet: An Electronic Database. MIT Press, Cambridge, MA. for text readability software for adults with intellectual disabilities. In Proc. ASSETS. pages 3–10. find
  11. Jonnalagadda, Siddhartha, Luis Tari, Jörg Hakenberg, Chitta Baral, and Graciela Gonzalez. 2009. Towards effective sentence simplification for automatic processing of biomedical text. In Proc. NAACL-HLT. pages 177–180. find
  12. McCarthy, Diana and Roberto Navigli. 2007. Semeval-2007 task 10: English lexical substitution task. In Proc. SemEval. pages 48–53. find
  13. Napoles, Courtney and Mark Dredze. 2010. Learning simple wikipedia: a cogitation in ascertaining abecedarian language. In Proc. of the NAACL-HLT Workshop on Computational Linguistics and Writing. pages 42–50. find
  14. Nelken, Rani and Stuart Shieber. 2006. Towards robust context-sensitive sentence alignment for monolingual corpora. In Proc. EACL. pages 161– 166. find
  15. Siddharthan, Advaith. 2004. Syntactic simplification and text cohesion. Technical Report UCAM-CL-TR-597, University of Cambridge, Computer Laboratory. find
  16. Vickrey, David and Daphne Koller. 2008. Applying sentence simplification to the CoNLL-2008 shared task. In Proc. CoNLL. pages 268–272. find
  17. Williams, Sandra and Ehud Reiter. 2005. Generating readable texts for readers with low basic skills. In Proc. ENLG. pages 127–132. find
  18. Yatskar, Mark, Bo Pang, Cristian Danescu-Niculescu-Mizil, and Lillian Lee. 2010. For the sake of simplicity: Unsupervised extraction of lexical simplifications from wikipedia. In Proc. NAACL-HLT. pages 365–368. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/biran-2011-putting-it-simply-context-aware/