---
title: "Semantic Information and Derivation Rules for Robust Dialogue"
description: "The paper addresses how a spoken dialogue system can recognise the user's… 84.3% accuracy. 12 references, cited by 1 paper here."
authors: ["Wei-Bin Liang", "Chung-Hsien Wu", "Chia-Ping Chen"]
year: 2011
venue: "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies"
source: "https://aclanthology.org/P11-2106/"
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/liang-2011-semantic-information-derivation-rules-robust/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/liang-2011-semantic-information-derivation-rules-robust/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2011](https://md.research.elamur.ai/free/en/papers/2011/page/2/) / Semantic Information and Derivation Rules for Robust Dialogue Act Detection in a Spoken Dialogue System

# Semantic Information and Derivation Rules for Robust Dialogue Act Detection in a Spoken Dialogue System

**Authors:** Wei-Bin Liang, Chung-Hsien Wu and Chia-Ping Chen · **Year:** 2011 · **Venue:** [Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2011](https://md.research.elamur.ai/free/en/venues/acl/2011/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **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/P11-2106/) · [PDF](https://aclanthology.org/P11-2106.pdf) · [OpenAlex](https://openalex.org/W2098068119)

**Citations:** 2 citations (OpenAlex, 2026-10-14) · [1](https://md.research.elamur.ai/free/en/papers/liang-2011-semantic-information-derivation-rules-robust/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper addresses how a spoken dialogue system can recognise the user's dialogue act (DA) when speech recognition output contains errors. The authors decompose DA detection into three multiplied scores: an ASR recognition score, a dialogue-history score from a back-off bigram over DA sequences, and a lexical score. The lexical score is the main contribution. Each recognised sentence becomes a partial sentence tree (PST), which lists every variant that keeps all keywords and some subset of non-keywords. Low-confidence words are replaced by a Filler token, and keywords are mapped to named entity classes. Sentences are then described by derivation rules (non-terminal → preterminal → word) taken from Stanford parser output. The DA inventory is not hand-defined: spectral clustering groups the sentences into DAs. An entropy-weighted DR-DA matrix links rules to acts. On a Mandarin tour-guiding system for Tainan, the full model reaches 84.3% accuracy, compared with 49.6% for a keyword-slot baseline. This shows that structure-aware, error-tolerant features can substantially reduce the damage caused by noisy ASR.

## Key points
- Data came from two sources. A-data contains 1,603 sentences and 317 word types, gathered from web sources such as Wikipedia and Google Map. Q-data is a Wizard-of-Oz corpus of 144 dialogues and 1,586 utterances. Using these corpora, the authors derived a total of 796 derivation rules along with 28 classes of named entities.
- The ASR system was built with HTK and used a bigram language model from SRILM. It achieved 86.1% word accuracy on a 297-word lexicon.
- The number of DA clusters was swept: 37 clusters gave 82.7%, 38 gave 84.3%, and 39 gave 77.2%. The system therefore uses q = 38.
- Components were added one at a time and tested on ASR transcripts, giving these accuracies:
  - baseline: 49.6%
  - adding NEC: 56.8%
  - adding PST: 76.2%
  - adding DR: 81.6%
  - adding DR-DA: 82.9%
- On reference transcripts, accuracy rose from 60.9% to 93.3%. However, DR-DA performed worse than PST when only 40% or 60% of the words were kept.
- The fusion weights were tuned, and setting the history weight λ_L to 0.7 raised accuracy to 85.1%.

## Interesting ideas
- **Partial sentence trees:** The system enumerates subsequences that keep every keyword but drop some non-keywords, then scores the best-matching variant. This lets unreliable function words fail without affecting the classification.
- **Derivation rules as features:** Instead of bare words, each word is represented by its parse path, such as phrase → POS → word. This keeps part-of-speech information and helps separate noun/verb homonyms.
- **Entropy-based feature weighting:** Each feature's weight is scaled by one minus the normalised entropy of its distribution over classes. Features spread evenly across all acts are suppressed, and features concentrated in a few acts are emphasised.
- **Inducing the label set by clustering:** The act inventory is discovered by spectral clustering, and the number of clusters is chosen by downstream accuracy instead of being fixed in advance.
- **Controlled degradation tests:** Inputs are evaluated with only 40% or 60% of words kept, alongside perfect and real ASR transcripts. This revealed that a component that helps on clean text can hurt on damaged text.

## For Mova

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

- **I1** (frames): Speech-act detection on noisy or ungrammatical input could score several pruned variants of a clause and keep the best match, while recording which words were dropped so the decision stays explainable.
- **I2** (frames): Speech-act and frame cues could be keyed on (dependency relation, UPOS, lemma) triples taken from the UD tree rather than on lemmas alone, giving structural features that are still easy to read.
- **I3** (frames): A cheap, transparent way to rank which cues actually indicate a given speech act or implicature, with the weight itself serving as the explanation.
- **I4** (learn): Candidate frame or speech-act categories could be proposed from silver data and kept only if they improve task scores, rather than relying on a hand-built inventory.
- **I5** (learn): The evaluation suite could include word-dropout variants of test texts as stress controls, so that a rule or model is accepted only if it does not degrade under corrupted input.

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

## Institutions

- [National Cheng Kung University](https://md.research.elamur.ai/free/en/universities/national-cheng-kung-university/) — University, Tainan, Taiwan · 3 works · 613 citations

Also on the byline: [National Sun Yat-sen University](https://ror.org/00mjawt10).

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (3): [🗨️ Dialogue systems](https://md.research.elamur.ai/free/en/topics/dialogue-systems/) · 401 works · [🎙️ Speech & spoken language](https://md.research.elamur.ai/free/en/topics/speech/) · 233 works · [💬 Speech act theory](https://md.research.elamur.ai/free/en/topics/speech-act-theory/) · 160 works

## Cite

Cite as: Wei-Bin Liang, Chung-Hsien Wu and Chia-Ping Chen. 2011. Semantic Information and Derivation Rules for Robust Dialogue Act Detection in a Spoken Dialogue System. Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies.

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

```bibtex
@inproceedings{liang2011semantic,
  title = {{Semantic Information and Derivation Rules for Robust Dialogue Act Detection in a Spoken Dialogue System}},
  author = {Wei-Bin Liang, Chung-Hsien Wu and Chia-Ping Chen},
  booktitle = {Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies},
  year = {2011},
  url = {https://aclanthology.org/P11-2106/}
}
```

- [← previous Derivation in a word-based morphology: on the origin of Old English verbs of perception, cognition and emotion](https://md.research.elamur.ai/free/en/papers/vera-2011-derivation-word-based-morphology-origin/)
- [next → Domain-Dependent Detection of Light Verb Constructions](https://md.research.elamur.ai/free/en/papers/nagy-2011-domain-dependent-detection-light-verb/)
- [↑ back to papers of 2011](https://md.research.elamur.ai/free/en/papers/2011/page/2/)

## References

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

1. J. Bellegarda. 2000. Exploiting latent semantic information in statistical language modeling. Proceedings of the IEEE, 88:1279–1296. [↗](https://doi.org/10.1109/5.880084)
2. N. Fraser and G. N. Gilbert. 1991. Simulating speech systems. Computer Speech and Language, 5(1):81– 99. [↗](https://doi.org/10.1016/0885-2308(91)90019-m)
3. C. Hori, K. Ohtake, T. Misu, H. Kashioka, and S. Nakamura. 2009a. Recent advances in wfst-based dialog system. In Proc. INTERSPEECH, pages 268–271. [find](https://scholar.google.com/scholar?q=C%2E+Hori%2C+K%2E+Ohtake%2C+T%2E+Misu%2C+H%2E+Kashioka%2C+and+S%2E+Nakamura%2E+2009a%2E+Recent+advances+in+wfst%2Dbased+dialog+system%2E+In+Proc%2E+INTERSPEECH%2C+pages+268%E2%80%93271%2E)
4. C. Hori, K. Ohtake, T. Misu, H. Kashioka, and S. Nakamura. 2009b. Statistical dialog management applied to wfst-based dialog systems. In Proc. IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), pages 4793–4796. [find](https://scholar.google.com/scholar?q=C%2E+Hori%2C+K%2E+Ohtake%2C+T%2E+Misu%2C+H%2E+Kashioka%2C+and+S%2E+Nakamura%2E+2009b%2E+Statistical+dialog+management+applied+to+wfst%2Dbased+dialog+systems%2E+In+Proc%2E+IEEE+International+Conference+on+Acoustics+Speech+and+Sig)
5. C. Hori, K. Ohtake, T. Misu, H. Kashioka, and S. Nakamura. 2009c. Weighted finite state transducer based statistical dialog management. In Proc. ASRU. [find](https://scholar.google.com/scholar?q=C%2E+Hori%2C+K%2E+Ohtake%2C+T%2E+Misu%2C+H%2E+Kashioka%2C+and+S%2E+Nakamura%2E+2009c%2E+Weighted+finite+state+transducer+based+statistical+dialog+management%2E+In+Proc%2E+ASRU%2E)
6. D. Jurafsky and J. H. Martin. 2009. Speech and Language Processing, 2nd Edition. Pearson Education. [find](https://scholar.google.com/scholar?q=D%2E+Jurafsky+and+J%2E+H%2E+Martin%2E+2009%2E+Speech+and+Language+Processing%2C+2nd+Edition%2E+Pearson+Education%2E)
7. R. J. Larsen and M. L. Marx. 2000. An Introduction to Mathematical Statistics and Its Applications, 3rd Edition. ISBN: 0139223037. [find](https://scholar.google.com/scholar?q=R%2E+J%2E+Larsen+and+M%2E+L%2E+Marx%2E+2000%2E+An+Introduction+to+Mathematical+Statistics+and+Its+Applications%2C+3rd+Edition%2E+ISBN%3A+0139223037%2E)
8. R. Levy and C. Manning. 2003. Is it harder to parse chinese, or the chinese treebank? In Proc. Annual Meeting of ACL, pages 439–446. [find](https://scholar.google.com/scholar?q=R%2E+Levy+and+C%2E+Manning%2E+2003%2E+Is+it+harder+to+parse+chinese%2C+or+the+chinese+treebank%3F+In+Proc%2E+Annual+Meeting+of+ACL%2C+pages+439%E2%80%93446%2E)
9. A. Stolcke. 2002. Srilm - an extensible language modeling toolkit. In Proc. International Conference on Spoken Language Processing, pages 901–904. [find](https://scholar.google.com/scholar?q=A%2E+Stolcke%2E+2002%2E+Srilm+%2D+an+extensible+language+modeling+toolkit%2E+In+Proc%2E+International+Conference+on+Spoken+Language+Processing%2C+pages+901%E2%80%93904%2E)
10. U. von Luxburg. 2007. A tutorial on spectral clustering. Statistics and Computing, 17(4). [↗](https://doi.org/10.1007/s11222-007-9033-z)
11. C.-H. Wu and Y.-J. Chen. 2004. Recovery from false rejection using statistical partial pattern trees for sentence verification. Speech Communication, 43(1-2):71–88. [↗](https://doi.org/10.1016/j.specom.2004.02.003)
12. Steve J. Young, D. Kershaw, J. Odell, D. Ollason, V. Valtchev, and P. Woodland. 2006. The HTK Book Version 3.4. Cambridge University Press. [find](https://scholar.google.com/scholar?q=Steve+J%2E+Young%2C+D%2E+Kershaw%2C+J%2E+Odell%2C+D%2E+Ollason%2C+V%2E+Valtchev%2C+and+P%2E+Woodland%2E+2006%2E+The+HTK+Book+Version+3%2E4%2E+Cambridge+University+Press%2E)

*Updated 24 October 2026*