---
title: "Some Prosodic Characteristics of Taiwan English Accent (2014)"
description: "The paper looks at how the English of native Taiwan Mandarin speakers differs prosodically from native English, with the aim of informing…"
authors: ["Chao-yu Su", "Chiu-yu Tseng", "Jyh-Shing Roger Jang"]
year: 2014
venue: "Proceedings of the 26th Conference on Computational Linguistics and Speech Processing (ROCLING 2014)"
source: "https://aclanthology.org/O14-1008/"
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/su-2014-some-prosodic-characteristics-taiwan-english/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/su-2014-some-prosodic-characteristics-taiwan-english/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2014](https://md.research.elamur.ai/free/en/papers/2014/page/4/) / Some Prosodic Characteristics of Taiwan English Accent

# Some Prosodic Characteristics of Taiwan English Accent

**Authors:** Chao-yu Su, Chiu-yu Tseng and Jyh-Shing Roger Jang · **Year:** 2014 · **Venue:** [Proceedings of the 26th Conference on Computational Linguistics and Speech Processing (ROCLING 2014)](https://md.research.elamur.ai/free/en/venues/rocling/) ([ROCLING 2014](https://md.research.elamur.ai/free/en/venues/rocling/2014/)) · **Publisher:** [The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)](https://md.research.elamur.ai/free/en/publishers/the-association-for-computational-linguistics-and-chinese-language-processing/) · **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/O14-1008/) · [PDF](https://aclanthology.org/O14-1008.pdf) · [OpenAlex](https://openalex.org/W2279377941)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper looks at how the English of native Taiwan Mandarin speakers differs prosodically from native English, with the aim of informing computer-assisted prosody training tools. These tools target suprasegmental errors rather than segmental ones. The English data come from AESOP-ILAS, a corpus built to elicit specific phenomena: word stress in carrier sentences, phrase boundaries, broad and narrow focus, stressed versus unstressed function words, and syntactic ambiguity resolved by prosody. Mandarin data come from SINICA COSPRO. The authors combine prosodic features from their earlier work and run three analyses. First, SVM and KNNC classifiers try to tell the three speech types apart. Second, individual features are inspected for signs of transfer from Mandarin. Third, speakers are compared by cosine similarity of their prosodic templates. Taiwan L2 English separates clearly from L1 English but shares traits with Taiwan Mandarin, which points to prosodic transfer. L2 speakers show weaker tempo contrast and a narrower loudness range for stress. The results can guide CALL and CAPT refinement.

## Key points
- Data: English from 9 L1 speakers (4M, 5F) and 9 Taiwan L2 speakers (5M, 4F) from AESOP-ILAS, plus Mandarin from 2 L1 speakers (1M, 1F) from the intonation-balanced part of SINICA COSPRO.
- Earlier work by the authors found that native speakers reduce the three-level primary/secondary/tertiary stress system to a binary stress/no-stress contrast:
  - The contrast is anchored at the primary stress.
  - Syllables after the primary stress are weakened.
  - Syllables before it can rise toward primary F0 levels.
  - L2 speakers struggle with this positional merge.
- SVM and KNNC classification separated L1 English from Taiwan L2 English. L2 English and Taiwan Mandarin grouped together against L1 English.
- Distinctive L2 traits were reduced tempo contrast within sentences, which changes rhythm, and a narrower loudness range for word stress, which blurs the strong/weak distinction.
- Speaker similarity was measured as the average cosine between prosodic feature vectors. It was higher within each group (L1 or L2) than across groups. The main differences lay in sentence loudness and in the timing and pitch of words, linked to syntax-driven narrow focus and lexical stress.

## Interesting ideas
- Adding the learners' native language as a third class makes transfer testable: if L2 output clusters with the mother tongue rather than with the target, transfer is the likely cause.
- Comparing within-group and between-group cosine similarity of per-item feature templates shows whether a group behaves consistently and whether it differs systematically from another group.
- The corpus was designed with linguistic targets built in: carrier sentences, words placed at boundaries, focus conditions, and syntactically ambiguous sentences. This turns the data into a controlled diagnostic set rather than a random sample.
- A three-level contrast that native speakers collapse into a binary one anchored at a key position suggests modelling distinctions relative to an anchor rather than as fixed absolute levels.
- The paper ties narrow focus partly to syntax, so structure can signal emphasis even when no audio is available.

## For Mova

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

- **I1** (learn): when a silver treebank is converted from another UD dialect, check whether its residual errors cluster with the source dialect's conventions, which would reveal leftover conversion artifacts.
- **I2** (learn): profile each witness parser by its label or attachment vectors on fixed sentences, and flag parsers whose outputs sit outside the agreeing cluster before trusting them for silver data.
- **I3** (answer): build small, targeted test suites with minimal pairs, such as ambiguous attachments or focus constructions, so each failure points to one specific explainable step.
- **I4** (tokens): if fine-grained tags prove unreliable, consider collapsing them into a binary distinction relative to a head or anchor token where the finer split adds noise.
- **I5** (frames): mark narrow focus from syntactic cues such as clefts or contrastive constructions, so that implicature and speech-act reasoning know which element is being emphasized.

*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

- [Academia Sinica](https://md.research.elamur.ai/free/en/labs/academia-sinica/) — Research institute, Taipei, Taiwan · 13 works · 11,163 citations
- [National Taiwan University](https://md.research.elamur.ai/free/en/universities/national-taiwan-university/) — University, Taipei, Taiwan · 32 works · 2,116 citations

📚 Shelves (4): [🎭 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 · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 1,649 works · [💭 Experimental and Cognitive Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/experimental-and-cognitive-psychology/) · 930 works

🧭 Topics (2): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works · [🎙️ Speech & spoken language](https://md.research.elamur.ai/free/en/topics/speech/) · 519 works

## Cite

Cite as: Chao-yu Su, Chiu-yu Tseng and Jyh-Shing Roger Jang. 2014. Some Prosodic Characteristics of Taiwan English Accent. Proceedings of the 26th Conference on Computational Linguistics and Speech Processing (ROCLING 2014).

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

```bibtex
@inproceedings{su2014some,
  title = {{Some Prosodic Characteristics of Taiwan English Accent}},
  author = {Chao-yu Su, Chiu-yu Tseng and Jyh-Shing Roger Jang},
  booktitle = {Proceedings of the 26th Conference on Computational Linguistics and Speech Processing (ROCLING 2014)},
  year = {2014},
  url = {https://aclanthology.org/O14-1008/}
}
```

- [← previous Setting Syntactic Parameters with Implicit Negative Evidence: The Case of Zero-derived Causatives in English](https://md.research.elamur.ai/free/en/papers/gould-2014-setting-syntactic-parameters-implicit-negative/)
- [next → The ASMAT project – Arabic social media analysis tools](https://md.research.elamur.ai/free/en/papers/anon-2014-asmat-project-arabic-social-media/)
- [↑ back to papers of 2014](https://md.research.elamur.ai/free/en/papers/2014/page/4/)

*Updated 2 November 2026*