语言学论文哪里有？本研究旨在提出用词首塞音的VOT值进行语音比对的可行性 ,为未来跨语言司法语音比对研究提供了新思路。
CHAPTER 1 INTRODUCTION
1.2 Rationale
The primary motivation behind this study stems from the imperative to combat TNFCs within the realm of forensic linguistics. Presently, there exists a multitude of avenues through which perpetrators can engage in TNFC, with a significant portion of these activities being conducted remotely. Consequently, in numerous instances, fraudulent texts or audio recordings serve as the sole evidence for apprehending these criminals, underscoring the critical need to leverage linguistic resources for convicting them. Mainstream linguistic inquiries into TNFC have predominantly focused on written discourse associated with telecom fraud, employing a diverse array of linguistic tools, from lexico-grammatical analyses to considerations of discourse structure, rhetoric, and socio-semiotic elements, to elucidate how fraudsters cultivate trust, manipulate victims through persuasion or threats, and ultimately ensnare them in a carefully orchestrated scheme. While existing research extensively explores these textual aspects, scant attention has been paid to forensic voice comparison based on acoustic-phonetic features, with Rose (2013) standing out as a notable exception. Through its detailed examination of FVC analysis in a genuine telephone fraud case, Rose's work not only addresses this research void but also serves as a pivotal catalyst for the current study.
CHAPTER 3 THEORETICAL FRAMEWORK
3.1 Acoustic-phonetic foundations
VOT, as the name suggests, refers to the time process when the articulatory organs alternate between the production of consonants and vowels, specifically the timing between the release of a stop or affricate and the onset of vocal fold vibration. Taking a consonant-vowel structure as an example, if the place of articulation of the consonant is close to that of the vowel and the vowel immediately follows the release, the VOT is very short and approaches zero; if the places of articulation are distant, causing a delay between the release and the onset of the vowel (resulting in a brief period of silence where the vocal folds have not yet started vibrating, visible as a gap on the spectrogram between the burst and the formants), the VOT is positive; in the case of a voiced stop, where the vocal folds begin vibrating before the release, the vibration continues into the vowel without a silent gap between the consonant and the vowel, resulting in a negative VOT value. These phenomena are illustrated in Figure 6.12, with the aspiration line as the zero point. The left panel represents a voiced stop, with a VOT less than 0; the middle panel represents a voiceless stop, with a VOT equal to 0 (sometimes slightly greater than 0, with a very short silent gap); the right panel represents an aspirated voiceless stop, where there is a period of airflow after the release before the vowel begins, resulting in a longer, positive VOT.
CHAPTER 5 RESULTS AND DISCUSSION
5.1 Inter-speaker comparison
Inter-speaker comparison is carried out to explore the voice-onset-time of the stops in the four syllables, specifically /kuo/<guo>, /kha/<ka>, /pan/<ban>, and /te/ <de>. During this analysis, speech samples from Speakers No. 5, 21, 23, and 24 are individually compared with speech samples from the fraudster.
When comparing different origins, an LR smaller than 1 is anticipated, suggesting a greater likelihood of supporting the hypothesis of different speakers over that of the same speaker. Consequently, the anticipated LLR values are less than 0, with a preference for lower negative LLR values. After outlining the aforementioned procedures, Table 5-1, 5-2, 5-3, and 5-4 present the LRs and the converted LLRs for formant transitions of the five syllables in various speaker comparisons. Subsequent to this, the evaluations are discussed accordingly.
5.2 Intra-speaker comparison
In forensic voice comparison, exploring intra-speaker variation is a crucial aspect when examining parameters. Following the inter-speaker comparison conducted in the previous section, this segment will concentrate on assessing the stability of VOT. Similar to the inter-speaker comparison, intra-speaker variation will be examined using paired speech samples, with the distinction that the compared samples originate from the same speakers. In this study, each of the four suspects is required to provide two recordings.
In the context of intra-speaker comparison, a log-likelihood ratio greater than 0 is expected to support the hypothesis that the recordings are from the same speaker; otherwise, a contrary conclusion will be drawn. The following section will present the calculated LR, Log LRs, and analysis for the intra-speaker variation test.
CHAPTER 6 CONCLUSION
6.1 Major findings
To enhance the range of features available for FVC within the realm of cellphone fraud in China, this thesis has formulated three key questions in Chapter 1. In Chapter 5, a thorough examination of the results leads to the presentation of the primary discoveries related to each research query.
The initial and primary phase of the study involves evaluating the ability to distinguish speakers based on selected acoustic features through comparisons. Analysis of the Log-Likelihood Ratios (LRs) obtained from comparisons between different speakers reveals that the VOT of the syllables /kuo/<guo>, /kha/<ka>, and /te/<de> exhibit discriminative potential for FVC, with discrimination rates of 100%, 75%, and 50% respectively. Conversely, the VOT of the syllable /pan/<ban> are deemed unsuitable for FVC, as they do not effectively differentiate speech samples from different speakers, showing a 0% success rate in this regard. This stage involves exploring the internal consistency of the parameters under scrutiny within the same speaker.
Upon analyzing the Log-Likelihood Ratios (LRs) obtained from within-speaker comparisons, it becomes evident that the VOT of all examined syllables exhibits strong stability within speakers. Specifically, the VOT of the syllables /kuo/<guo>, /pan/<ban>, and /te/<de> demonstrate stability rates of up to 100%, while /kha/<ka> shows a stability rate of 75%. In conclusion, considering both inter- and intra-speaker variations, the VOT of the syllables /kuo/<guo>, /kha/<ka>, and /te/<de> emerges as robust parameters suitable for FVC.
reference(omitted)
