Unsupervised discovery of morphologically related words based on orthographic and semantic similarity

What it is about

The algorithm takes an unannotated corpus and returns a ranked list of probably related word pairs, such as a verb and its past form or a noun and its derived adjective. Two words are paired when they are close both in spelling and in meaning. Spelling is measured by minimum edit distance, and meaning by mutual information between the words in the corpus. The method does not model morphemes as concatenation, and it does not use affix frequencies. The authors test German and English. Precision, the share of correct pairs above a cut-off on the ranked list, is high at the top of the list: about 97% for the first 500 German pairs, out of 5,279 in total. The paper also discusses the types of morphological patterns that appear in the output.

Ideas

• I1. en/src/lexgen.rs: rank candidate pairs of English words in data/forms.tsv that are not yet linked, by edit distance and by a co-occurrence score from a raw corpus. Measure precision at 500 and 1,000 pairs against the existing links in the dictionary.
• I2. Add the paper's transitivity idea: if a is related to b and b to c, then a and c form a pair. Check on the held-out links of data/lemmas.tsv whether the transitive pairs are right more often than the ranking score alone.
• I3. Report precision at fixed cut-offs in the output of the candidate list, as the paper does, so that the ranking can be compared run to run before any use in the rules of en/src/morph.rs.

For Mova

The code is en/src/lexgen.rs, with en lexicon. The chapters are 16. Derivation I: prefixes and 17. Derivation II: suffixes. The paper's German data is not English, and the English results are only qualitative in the text, so Mova needs its own numbers. Measure first on English held-out links, twice, with the ranking fixed before any change.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3115/1118647.1118653

@inproceedings{baroni2002unsupervised,
  title = {{Unsupervised discovery of morphologically related words based on orthographic and semantic similarity}},
  author = {Baroni, Marco and Matiasek, Johannes and Trost, Harald},
  booktitle = {Proceedings of the ACL-02 Workshop on Morphological and Phonological Learning},
  year = {2002},
  doi = {10.3115/1118647.1118653},
  url = {https://doi.org/10.3115/1118647.1118653}
}

References

16 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. A. Albright and B. Hayes. 1999. An automated learner for phonology and morphology. UCLA manuscript. find
2. M. Baroni. 2000. Distributional cues in morpheme discovery: A computational model and empirical evidence. Ph.D. dissertation, UCLA. find
3. M. Baroni, J. Matiasek and H. Trost. 2002. Wordform-and class-based prediction of the components of German nominal compounds in an AAC system. To appear in Proceedings of COLING 2002. find
4. P. Brown, P. Della Pietra, P. DeSouza, J. Lai, and R. Mercer. 1990. Class-based n-gram models of natural language. Computational Linguistics, 18:467-479. find
5. K. Church and P. Hanks. 1989. Word association norms, mutual information, and lexicography. Proceedings of ACL 27, 76–83. find
6. J. Goldsmith. 2001. Unsupervised learning of the morphology of a natural language. Computational Linguistics, 27:153-198. find
7. C. Jacquemin. 1997. Guessing morphology from terms and corpora. Proceedings of SIGIR 97, 156–265. find
8. D. Jurafsky and J. Martin. 2000. Speech and Language Processing. Prentice-Hall, Upper Saddle River, NJ. find
9. L. Karttunen, K. Gaál, and A. Kempe. 1997. Xerox Finite-State Tool Xerox Research Centre Europe, Grenoble. find
10. H. Kučera and N. Francis. 1967. Computational analysis of present-day American English. Brown University Press, Providence, RI. find
11. C. Manning and H. Schütze. 1999. Foundations of statistical natural language processing. MIT Press, Cambridge, MASS. find
12. S. Neuvel. 2002. Whole word morphologizer. Expanding the word-based lexicon: A non-stochastic computational approach. Brain and Language, in press. find
13. R. Rosenfeld. 1996. A maximum entropy approach to adaptive statistical language modeling. Computer Speech and Language, 10:187–228. find
14. P. Schone and D. Jurafsky. 2000. Knowldedge-free induction of morphology using latent semantic analysis. Proceedings of the Conference on Computational Natural Language Learning. find
15. M. Snover and M. Brent. 2001. A Bayesian model for morpheme and paradigm identification. Proceedings of ACL 39, 482-490. find
16. D. Yarowksy and R. Wicentowski. 2000. Minimally supervised morphological analysis by multimodal alignment. Proceedings of ACL 38, 207–216. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2002-baroni-unsupervised-discovery-morphologically-related-words/