---
title: "Parsing Long English Sentences with Pattern Rules — Li 1990"
description: "This paper addresses a practical problem in rule-based machine translation. 4 references, cited by 4 papers here."
authors: ["Wei-Chuan Li", "Tzusheng Pei", "Bing-Huang Lee", "Chuei-Feng Chiou"]
year: 1990
venue: "COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics"
source: "https://aclanthology.org/C90-3088/"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/li-1990-parsing-long-english-sentences-pattern/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/li-1990-parsing-long-english-sentences-pattern/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1990](https://md.research.elamur.ai/free/en/papers/1990/) / Parsing Long English Sentences with Pattern Rules

# Parsing Long English Sentences with Pattern Rules

**Authors:** Wei-Chuan Li, Tzusheng Pei, Bing-Huang Lee and Chuei-Feng Chiou · **Year:** 1990 · **Venue:** [COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 1990](https://md.research.elamur.ai/free/en/venues/coling/1990/)) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/C90-3088/) · [PDF](https://aclanthology.org/C90-3088.pdf) · [OpenAlex](https://openalex.org/W2059230065)

**Citations:** 13 citations (OpenAlex, 2026-10-14) · 4 in our library, of the 149,439 papers whose bibliographies we read

## Summary
This paper addresses a practical problem in rule-based machine translation. Long or structurally complex English sentences make the number of candidate parses explode, so parsing time and error rates rise sharply. The authors work inside ERSO-ECMT, an English-Chinese system that uses an augmented context free grammar. They add a separate layer of "long sentence pattern" rules. Each rule is a regular expression over words, categories and variables, augmented with LISP test functions. Sentences longer than a threshold of 40 words are matched against these rules and split into large segments such as SDEC, NP, INF and VP. Each segment is parsed by the existing grammar, and the results are merged or reordered into the Chinese output. If no pattern fits, or a pattern does not yield a full tree, the system falls back to the ordinary grammar. On an environmental-protection corpus, 36 of 42 long sentences were partitioned correctly, and 28 of those were also translated correctly. The approach matters because it limits search depth without trying to solve deep semantic attachment problems.

## Key points
- ERSO-ECMT has been developed at ERSO/ITRI in Taiwan since July 1986. It follows an analysis/transfer/synthesis design, translates one sentence at a time without context, and targets computer science and environmental protection texts.
- On the computer science corpus (first chapter of the UNIX manual), running on a LAMBDA LISP machine:
  - sentences under 34 words translate in about a minute on average;
  - from around 40 words upward, translation time grows exponentially.
- As motivation, the authors cite ATLAS II. Its usable translations averaged 33.5 words, while translations needing post-editing or unusable ones averaged 45.7 and 46.8 words.
- The environmental test set has 871 sentences, 19,539 words and a mean of 18.3 words per sentence. Of these, 42 sentences (4.8%) exceed 40 words.
  - Results on those 42: 36 correct partitions, 28 correct translations, about 2 minutes per sentence.
  - The paper gives 66.7% as the ratio of the last two counts, although 28/36 works out to about 78%.
- Example patterns include:
  - (In order INF, SDEC);
  - (SDEC, because SDEC);
  - a repeated gerund-list pattern using closure;
  - (Not that SDEC but that SDEC), which maps to a fixed Chinese construction.

## Interesting ideas
- Before full parsing, coarse cue-word patterns (e.g. "because", "in order", comma plus gerund lists) cut a long sentence into big clause-level chunks, and the chunks are then parsed independently, so a local error stays local.
- Pattern rules act as an optional front layer with a strict fallback: if no pattern fits, or a pattern fails to produce a complete tree, the system simply uses the normal grammar, so adding rules never reduces coverage.
- Partitioning is applied recursively until every segment falls below a length threshold or no pattern applies.
- To avoid costly matching against every rule, the authors propose attaching each pattern to its leading word in the lexicon, or to the relevant grammar nonterminal, so that only plausible patterns are tried.
- Some segment patterns carry their own target-language template, so a fixed English construction maps directly to a fixed output form rather than being translated piece by piece.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (tree): A pre-segmentation pass for very long inputs could let the perceptron parser or gparse handle bounded spans, which would then be reattached at the cue word with an explainable record of the split.
- **I2** (learn): Self-induced repair rules could be gated the same way, kept only when they yield a valid tree and otherwise bypassed, and checked with negative controls to show they never hurt baseline output.
- **I3** (tree): A configurable length threshold with recursive splitting would give predictable parse cost on long school texts or tales without changing the parser itself.
- **I4** (tokens): The crate en lexicon could store construction triggers on lemmas such as "not that … but that", letting later stages fire construction-specific handling cheaply.
- **I5** (back): Generation and translation through the meta-world could keep a small table of construction-to-template mappings for idiomatic contrastive or purpose structures.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🧮 Computational Theory and Mathematics](https://md.research.elamur.ai/free/en/shelves/computer-science/computational-theory-and-mathematics/) · 321 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works

## Cite

Cite as: Wei-Chuan Li, Tzusheng Pei, Bing-Huang Lee and Chuei-Feng Chiou. 1990. Parsing Long English Sentences with Pattern Rules. COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics.

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

```bibtex
@inproceedings{li1990parsing,
  title = {{Parsing Long English Sentences with Pattern Rules}},
  author = {Wei-Chuan Li, Tzusheng Pei, Bing-Huang Lee and Chuei-Feng Chiou},
  booktitle = {COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics},
  year = {1990},
  doi = {10.3115/991146.991234},
  url = {https://doi.org/10.3115/991146.991234}
}
```

- [← previous Reversible Unification Based Machine Translation](https://md.research.elamur.ai/free/en/papers/noord-1990-reversible-unification-based-machine-translation/)
- [next → A Formal Description of Arabic Syntax in Definite Clause Grammar](https://md.research.elamur.ai/free/en/papers/shishiny-1990-formal-description-arabic-syntax-definite/)
- [↑ back to papers of 1990](https://md.research.elamur.ai/free/en/papers/1990/)

## References

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

1. [i] Jane J. Robinson, "DIAGRAM: A Gra~unar for Dialogues", CoMmunications of ACM, January 19S2, pp. 27-47. [2] W.S. Bennett and J. Slocum, "The LRC Machine T r a n s l a t i o n System", Computational Linguistics, Vol. ii, No. 2-3, April-September, 1985, pp. 112-121. [find](https://scholar.google.com/scholar?q=%5Bi%5D+Jane+J%2E+Robinson%2C+%22DIAGRAM%3A+A+Gra%7Eunar+for+Dialogues%22%2C+CoMmunications+of+ACM%2C+January+19S2%2C+pp%2E+27%2D47%2E+%5B2%5D+W%2ES%2E+Bennett+and+J%2E+Slocum%2C+%22The+LRC+Machine+T+r+a+n+s+l+a+t+i+o+n+System%22%2C+Computational)
2. [3] Keh-Yih Su, et al., "A New Parsing Strategy in Natural Language Processing Based on the Truncation Algorithm"~ NCS 1987, Vol 2, pp. 580-586. [find](https://scholar.google.com/scholar?q=%5B3%5D+Keh%2DYih+Su%2C+et+al%2E%2C+%22A+New+Parsing+Strategy+in+Natural+Language+Processing+Based+on+the+Truncation+Algorithm%22%7E+NCS+1987%2C+Vol+2%2C+pp%2E+580%2D586%2E)
3. [4] P.H. Winston, "Artificial Intelligence", Addison-Wesley, Reading, MA, USA, 1984. [find](https://scholar.google.com/scholar?q=%5B4%5D+P%2EH%2E+Winston%2C+%22Artificial+Intelligence%22%2C+Addison%2DWesley%2C+Reading%2C+MA%2C+USA%2C+1984%2E)
4. [5] Makoto Shiotsu, "Japanese Polish Support System for the J a p a n e s e - E n g l i s h Translation", The 35th National Conference of Information Processing Society of Japan, September 1987(in Japanese), pp. 1249-1250. [find](https://scholar.google.com/scholar?q=%5B5%5D+Makoto+Shiotsu%2C+%22Japanese+Polish+Support+System+for+the+J+a+p+a+n+e+s+e+%2D+E+n+g+l+i+s+h+Translation%22%2C+The+35th+National+Conference+of+Information+Processing+Society+of+Japan%2C+September+1987%28in+Jap)

*Updated 30 October 2026*