2024 Asia-Pacific Conference on Image Processing, Electronics and Computers (IPEC)
Download PDF

Abstract

Phonetic conversion technology is pivotal in developing resources for Russian speech information processing. This paper aims to enhance the Russian phoneme set based on SAMPA, ensuring accurate reflection of stress position and vowel reduction in Russian words. Utilizing the improved phoneme set, this paper successfully constructed a Russian pronunciation dictionary with 20,000 words. A data-driven Russian phonetic conversion algorithm leveraging the Hidden Markov Model (HMM) for alignment, modeling, and decoding is proposed. Initially, the Expectation-Maximization algorithm aligns Russian phonetics in a “many-to-many” manner. Subsequently, alignment results are trained using a joint N-gram model and transformed into a HMM pronunciation model. The HMM decoding algorithm ultimately predicts word pronunciation accurately. Cross-validation experiments yielded satisfactory results, and they show that the proposed can be easily implemented in the computer assisted system and provide enhanced accuracy and convenience to the Russian conversion process.
Like what you’re reading?
Already a member?
Get this article FREE with a new membership!

Related Articles