Multi-language Speech Database: Creation and Phonetic Labeling Agreement
نویسندگان
چکیده
The focus of the paper is the evaluation of inter-labeler reliability on broad phonetic transcriptions when la-belers do not necessarily know the language they are labeling. We provide an analysis of label disagreements, presenting results from six languages, Spanish, and Vietnamese with a total of 2 minutes of continuous labeled speech. Labeler agreement across languages ranges from 41 percent with detailed label to label comparisons to 91 percent when less ne comparisons were made.
منابع مشابه
Automatic Labeling of Corpora for Speech
One of the bottlenecks in the development of text-to-speech synthesizers based on segment concatenation is the need for large, segmented and labeled corpora. Consequently, as manual segmentation and labeling is a tedious and time consuming task, there is a strong demand for automatic labeling systems which can label speech from many languages. Several systems have been proposed already, but the...
متن کاملACTOR: A multilingual unit-selection speech synthesis system
The ACTOR® Text-To-Speech (TTS) synthesis system, developed at Loquendo S.p.A., is here described. The system employs a unit -selection concatenative synthesis technique, relying on labeled acoustic databases providing phonetic and prosodic coverage of the intended language/domain and on an original algorithm for run-time selection of the acoustic units to be concatenated. This technique yields...
متن کاملA Prosodic Labeling System for Mandarin Speech Database
A working database needs tools to transcribe and label at both phonetic and prosodic levels. While the proposed phonetic transcription system is a simplified from of the International Phonetic Alphabet (IPA) following the SAMPA guidelines; the prosodic labeling system is an elaborated form of the ToBI (Tone and Break Indices) framework adopted for Mandarin. In particular, the proposed prosodic ...
متن کاملA Computational Platform for Development of Morphologic and Phonetic Lexica
Statistic approaches in speech technology, either based on statistical language models, trees, hidden Markov models or neural networks, represent the driving forces for the creation of language resources (LR), e.g. text corpora, pronunciation lexica and speech databases. This paper presents the system architecture for rapid construction of morphologic and phonetic lexica for Slovenian language....
متن کاملAn analysis of transcription consistency in spontaneous speech from the buckeye corpus
We present a preliminary analysis of transcriber consistency in labeling and segmentation of words and phones in the Buckeye corpus of spontaneous, informal speech. We find that pairwise inter-transcriber agreement on exact phone label match was 76%, and segmentation agreement within 20% of phone pair length was 75%, though longer phones are more consistently segmented than shorter phones. Patt...
متن کامل