Using Word Semantics To Assist English as a Second Language Learners
Summary
The authors present a tool that helps learners of English as a second or foreign language (ESL/EFL) understand difficult words while reading. The tool is built on their lexical substitution system SALSA, which selects synonyms based on context. It works in two stages. First, it gathers all possible candidates from five lexical resources without distinguishing between word senses. Then it scores how well each candidate fits the specific sentence, using features built on frequencies from Google N-grams. On the SemEval-2007 data, the system ranks the synonym chosen by the annotators first in 21.3% of cases and places it in the top ten in 64.7%. The system was first used to build a web interface with a side panel showing synonyms and examples from WordNet. Later, a browser extension was developed: it automatically highlights difficult words on any page and shows a hint in a pop-up window. The tool also collects user feedback. This feedback can be used to build gold-standard data and, in the future, to adapt the tool to individual readers.
Key points
- SALSA draws candidates from Roget, WordNet, TransGraph, Lin and Encarta. Among the ranking methods tested, both supervised and unsupervised, features from Google N-grams performed best.
- On the SemEval-2007 lexical sample (1,700 instances for 170 words), the system achieves 21.3% first-choice hits and 64.7% in the top-10. The authors consider this on par with the best results of the competition.
- On a set of 550 words in running text, the results are 29.9% as the first choice and 73.7% in the top-10.
- The static version of the interface was previously tested on ESL students reading articles from the BBC learning website and on school students who are native English speakers. A regression model on features of interaction with the tool predicted post-reading test results well (Hokamp et al., 2014).
- The browser extension also works with dynamically generated content. It collects explicit feedback (upvotes or downvotes) and implicit feedback: which words were clicked, the word's IDF, and the number of syllables.
Cite
@inproceedings{azab2015using,
title = {{Using Word Semantics To Assist English as a Second Language Learners}},
author = {Azab, Mahmoud and Hokamp, Chris and Mihalcea, Rada},
booktitle = {Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations},
year = {2015},
doi = {10.3115/v1/n15-3024},
url = {https://doi.org/10.3115/v1/n15-3024}
}References
- T. Brants and A. Franz. 2006. Web 1T 5-gram version 1. Linguistic Data Consortium. ↗
- J. Burstein, J. Shore, J. Sabatini, and Y. Lee. 2007. Developing a reading support tool for English language learners. In Demo proceedings of the the annual conference of the North American chapter of the Association for Computational Linguistics (NAACL-HLT 2007), Rochester, NY. find
- Y. Canning and J. Tait. 1999. Syntactic simplification of newspaper text for aphasic readers. In Proceedings of the ACM SIGIR’99 Workshop on Customised Information Delivery, Berkeley, California. find
- M.S. Carlo, D. August, B. McLaughlin, C.E. Snow, C. Dressler, D. Lippman, T. Lively, and C. White. 2004. Closing the gap: Addressing the vocabulary needs of english language learners in bilingual and mainstream classrooms. Reading Research Quarterly, 39(2). ↗
- J. Carroll, G. Minnen, Y. Canning, S. Devlin, and J. Tait. 1998. Practical simplification of English newspaper text to assist aphasic readers. In Proceedings of the AAAI-98 Workshop on Integrating Artificial Intelligence and Assistive Technology, Madison, Wisconsin. find
- J. Carroll, G. Minnen, D. Pearce, Y. Canning, S. Devlin, and J. Tait. 1999. Simplifying text for language-impaired readers. In Proceedings of the Conference of the European Chapter of the ACL (EACL 1999), Bergen, Norway. find our card
- D. Crystal. 1997. English as a global language. find
- Cambridge University Press. 2004. http://nces.ed.gov/fastfacts/display.asp?id=96. ↗
- S. Devlin, J. Tait, J. Carroll, G. Minnen, and D. Pearce. 2000. Making accessible international communication for people with language comprehension difficulties. In Proceedings of the Conference of Computers Helping People with Special Needs. find
- N. Elhadad. 2006a. Comprehending technical texts: Predicting and defining unfamiliar terms. In Proceedings of the Annual Symposium of the American Medical Informatics Association, Washington. find
- N. Elhadad. 2006b. User-Sensitive Text Summarization: Application to the Medical Domain. Ph.D. thesis, Columbia University. find
- G. Grefenstette and F. Segond, 2003. Multilingual On-Line Natural Language Processing, chapter 38. find
- C. Hokamp, R. Mihalcea, and P. Schuelke. 2014. Modeling language proficiency using implicit feedback. In Proceedings of the Conference on Language Resources and Evaluations (LREC 2014), Reykjavik, Iceland, May. find
- D. McCarthy and R. Navigli. 2007. The semeval English lexical substitution task. In Proceedings of the ACL Semeval workshop. find
- A. Siddharthan, A. Nenkova, and K. McKeown. 2004. Syntactic simplification for improving content selection in multi-document summarization. In Proceedings of the 20th international conference on Computational Linguistics. ↗
- R. Sinha and R. Mihalcea. 2012. Explorations in lexical-sample and all-words lexical substitution. Journal of Natural Language Engineering. ↗
- Ravi Som Sinha and Rada Mihalcea. 2014. Explorations in lexical sample and all-words lexical substitution. Natural Language Engineering, 20(1):99–129. find
- Y. Yano, M. Long, and S. Ross. 1994. The effects of simplified and elaborated texts on foreign language tool’s utility and effectiveness in terms of students’ reading comprehension. Language Learning, 44. find