A Word Selection Model Based on Lexical Semantic Knowledge in English Generation

What it is about

The paper proposes a word-selection model for an interlingua-based Chinese-to-English system. Word selection chooses one English word among several translations of a concept. The authors compare two existing methods: semantic patterns (hand-made slot constraints, which are often tied, or fail for unseen words) and examples (similar sentences from a corpus, which need many examples and can explode in combinations). Their hybrid first uses patterns to drop poor candidates, then uses examples to choose the best one among the rest. Both knowledge bases are built automatically. The example base comes from SemCor, mapped from WordNet senses to HowNet concepts and turned from linear collocations into frame-like entries, giving 4,362 examples. The pattern base is trained from those examples, so that each slot has a probability instead of a yes or no. The paper gives no accuracy figures for word selection; it reports the sizes of the bases, the algorithms and sample entries. The claim is that the hybrid method is more flexible than either method alone.

Ideas

• I1. en/src/dict.rs and world/src/dictionary.rs: keep the sense-to-word choice as two stages, a fixed slot filter first and a count-based example score second, each stage writing its reasons into the output. Ties are broken by lemma order in a BTreeMap.
• I2. Probability in slot constraints: replace the yes or no of a semantic slot with a count ratio (how often the word takes that slot value in the school texts), with a smoothing constant. Use it to rank the candidate words for a concept.
• I3. Measure word selection on the school lessons: take each sentence, remove one content word, and check that the selected word is the one in the text, with the rate reported per concept. Report the rate with the slot filter only, the examples only, and both, as the paper's three-way comparison suggests.
• I4. The sense map: write a checked table from the world's concept names to the WordNet-style senses of the dictionary, and keep every mapping that has more than one candidate in a list for hand review, since the paper's mapping step resolves ties by frequency.

For Mova

The slot filter and example scores belong to world/src/dictionary.rs (word senses for the school) and en/src/dict.rs. Choosing a word for a concept belongs with world/src/knowledge.rs. The collocational slots relate to 19. The lexicon: vocabulary, multiword expressions and register. Measures: word-selection accuracy per concept, with the two stages separately, and the count of mappings left for review. Run twice to confirm the output is identical.

Cite

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

@inproceedings{chen2003word,
  title = {{A Word Selection Model Based on Lexical Semantic Knowledge in English Generation}},
  author = {Chen, Yi-Dong and Li, Tang-Qiu and Zheng, Xu-Ling},
  booktitle = {Proceedings of the 17th Pacific Asia Conference on Language, Information and Computation},
  year = {2003},
  url = {https://aclanthology.org/Y03-1043/}
}

References

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

1. Chen Qunxiu. 1998. An On-line Thesaurus: WordNet. Application of Language and Literary, (2):93-99 find
2. Chen Yidong, Li Tangqiu, Hong Qingyang and Zheng Xuling. 2001. Designing of a Model of Word selection in English Generation. Journal of Chinese Information Processing, 15(6): 19 - 26. find
3. Chen Yidong, Li Tangqiu and Zheng Xuling. 2002. Fuzzy Semantic Pattern and Its Application in the Word selection for English Generation. Journal of Chinese Information Processing, 16(5): 15-22. find
4. Dong Zhendong and Dong Qiang. HowNet. http://www.keenage.com W. N., and Kucera H. 1982. Frequency Analysis of English Usage: Lexicon and Grammar. Houghton Mifflin Company, Boston. ↗
5. Liu Xiaohu and Li Sheng. 1998. Target Word Selection in Machine Translation Based on Machine Learning. Computer Research & Development, 35(10):946 - 950 find
6. Miller G.A., Leacock C., Tengi R. and Bunker R.T. 1993. A Semantic Concordance. In: Proceedings of the ARPA Workshop on Human Language Technology. find
7. Nirenburg, Sergei and Nirenburg, Irene. 1998. A framework for word selection in NLG. In: Proceedings of the 12th International Conference on Computational Linguistics. find
8. Yang Xiaofeng, Li Tangqiu and Hong Qingyang. 2001. Word Sense Disambiguation Method in Chinese - English Machine Translation System. Journal of Chinese Information Processing, 15(3):22 - 28. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/chen-2003-word-selection-model-based-lexical/