The Operation of Computer Technology in Corpus-based Spoken Language

Jie ChenThe Open Automation and Control Systems Journal 2013 · DOI

Abstract

Corpus refers to the database of language materials. Cool Edit Pro is a media edit software. This paper explores how to construct spoken language corpus, how to use cool edit pro 2 to make waveforms for L2 learners at the different stages, make contrast among them and try to provide for the experimenters an intuitive observation from their own speech waveforms. The key is to offer the obvious waveforms contrast among the sampling waveform of the native speaker, the L2 learners' more times waveforms before and after necessary modifications and instructions, resulting in making the oral autonomous learning more effective and efficient. This paper follows the modern education technology methods and quantitative and qualitative analysis to make experiments on pronunciations of speech, aiming to get a satisfactory result and further promote this method to the other fields of language teaching and learning.

Introduction

In traditional English Language Teaching (ELT), feedback on spoken language is often subjective, relying on a teacher’s "one-time impression." For the learner, these critiques can feel abstract and difficult to act upon. While written corpora (databases of language) are well-established, spoken corpora are notoriously difficult to handle due to the complexities of recording, processing, and analyzing live speech.

This research addresses the gap between abstract linguistic theory and practical classroom application. By utilizing digital audio visualization—specifically waveform analysis—the study demonstrates how to revolutionize second language (L2) acquisition. By using software to compare the "visual signatures" of native speakers against L2 learners, the researchers provide a quantitative method for students to intuitively correct their pronunciation, rhythm, and intonation, moving beyond simple "listen and repeat" methods.

Methodology: Corpus Construction and Audio Processing

The researchers utilized Cool Edit Pro 2.1 to create a feedback loop for L2 learners. The technical workflow involved several critical stages of digital signal processing to ensure the data was suitable for linguistic analysis. This included noise reduction and filtering using Fast Fourier Transform (FFT) filters to isolate speech and remove background distortion without altering core vocal characteristics.

Systematic workflow for the development of a raw spoken language corpus.

The software was used to align three distinct waveforms for the same sentence: the L2 learner’s initial attempt, the native speaker’s "standard" sample, and the L2 learner’s second attempt after imitation and instruction. Beyond visual inspection, the authors used statistical measures including RMS (Root Mean Square) excitation, amplitude (dB), and frequency (Hz) to calculate how closely the learner’s mimicry approached the native speaker’s profile.

Pedagogical Framework and Corpus Applications

The paper argues that learner corpora should follow a task-based approach rather than a text-type approach, focusing on the process of speech production. This fits into a broader framework of pedagogical corpus applications, which can be divided into direct and indirect uses.

Framework for direct and indirect pedagogical applications of corpora.

By providing an "intuitive observation," the technology shifts the burden of correction from the teacher to the student. In "direct applications," learners can repeatedly record and adjust their speech until their waveform matches the shape of the native speaker's. This encourages autonomous learning where waveforms serve as linguistic maps; for example, a "narrow" waveform indicates excessive speed, while a "small" waveform relative to the sample suggests issues with accent and stress.

Datasets and Experimental Scope

The paper references several internationally recognized corpora, such as the London-Lund Corpus (LLC) and the British National Corpus (BNC), while highlighting the landscape of English corpora in China, including COLSEC (Shanghai Jiao Tong University), MSEE (South China Normal University), and SWECCL (Nanjing University).

For the experimental portion, the authors created a bespoke "learner’s video database" where students chose famous speeches or interviews to imitate, creating a personalized dataset for comparative analysis.

Key Results

The study provided quantitative evidence that imitation guided by visual feedback works. In their statistical analysis, the Total RMS Excitation of the native speaker was measured at -20.79 dB. The L2 learner’s initial attempt was -18.31 dB. After imitation and waveform comparison, the learner’s "Mimic Waveform" reached -20.5 dB, showing a significant convergence toward the native speaker’s vocal energy and intensity profile.

Qualitatively, the authors observed that the second-time recordings were "nearly tend[ing] to the native speaker’s waveform," resulting in speech that was described as more "nice and pleasant."

Conclusion and Future Work

While the results are promising, the authors acknowledge that this method requires technical literacy; teachers must be proficient in audio analysis software to act as "corpus linguists." Future work should expand these digital methods into other areas, such as using E-learning platforms to collect and analyze "inter-language" roles in written English. However, the current study is limited by a relatively small experimental scope and does not yet detail the long-term retention of these pronunciation improvements once the visual aids are removed.

Summary updated

Abstract

Corpus refers to the database of language materials. Cool Edit Pro is a media edit software. This paper explores how to construct spoken language corpus, how to use cool edit pro 2 to make waveforms for L2 learners at the different stages, make contrast among them and try to provide for the experimenters an intuitive observation from their own speech waveforms. The key is to offer the obvious waveforms contrast among the sampling waveform of the native speaker, the L2 learners' more times waveforms before and after necessary modifications and instructions, resulting in making the oral autonomous learning more effective and efficient. This paper follows the modern education technology methods and quantitative and qualitative analysis to make experiments on pronunciations of speech, aiming to get a satisfactory result and further promote this method to the other fields of language teaching and learning.
BibTeX
@article{chen2013operation,
  author = {Jie Chen},
  title = {{The Operation of Computer Technology in Corpus-based Spoken Language}},
  journal = {The Open Automation and Control Systems Journal},
  year = {2013},
  doi = {10.2174/1874444301305010045},
}
Figure