CUPE: Contextless Universal Phoneme Encoder for Language-Agnostic Speech Processing
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918128317890560 |
|---|---|
| author | Rehman, Abdul Zhang, Jian-Jun Yang, Xiaosong |
| author_facet | Rehman, Abdul Zhang, Jian-Jun Yang, Xiaosong |
| contents | Universal phoneme recognition typically requires analyzing long speech segments and language-specific patterns. Many speech processing tasks require pure phoneme representations free from contextual influence, which motivated our development of CUPE - a lightweight model that captures key phoneme features in just 120 milliseconds, about one phoneme's length. CUPE processes short, fixed-width windows independently and, despite fewer parameters than current approaches, achieves competitive cross-lingual performance by learning fundamental acoustic patterns common to all languages. Our extensive evaluation through supervised and self-supervised training on diverse languages, including zero-shot tests on the UCLA Phonetic Corpus, demonstrates strong cross-lingual generalization and reveals that effective universal speech processing is possible through modeling basic acoustic patterns within phoneme-length windows. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_15316 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CUPE: Contextless Universal Phoneme Encoder for Language-Agnostic Speech Processing Rehman, Abdul Zhang, Jian-Jun Yang, Xiaosong Computation and Language Machine Learning Audio and Speech Processing I.2.7 Universal phoneme recognition typically requires analyzing long speech segments and language-specific patterns. Many speech processing tasks require pure phoneme representations free from contextual influence, which motivated our development of CUPE - a lightweight model that captures key phoneme features in just 120 milliseconds, about one phoneme's length. CUPE processes short, fixed-width windows independently and, despite fewer parameters than current approaches, achieves competitive cross-lingual performance by learning fundamental acoustic patterns common to all languages. Our extensive evaluation through supervised and self-supervised training on diverse languages, including zero-shot tests on the UCLA Phonetic Corpus, demonstrates strong cross-lingual generalization and reveals that effective universal speech processing is possible through modeling basic acoustic patterns within phoneme-length windows. |
| title | CUPE: Contextless Universal Phoneme Encoder for Language-Agnostic Speech Processing |
| topic | Computation and Language Machine Learning Audio and Speech Processing I.2.7 |
| url | https://arxiv.org/abs/2508.15316 |