AI-powered speech training model for business-oriented english learners

With the increasing level of internationalization, traditional foreign language training methods cannot satisfy the current requirement for compound talents’ foreign language abilities in the business environment. Based on this, this study combines artificial intelligence technology to propose an in...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Discover Artificial Intelligence Ročník 5; číslo 1; s. 361 - 17
Hlavní autor: Wu, Jing
Médium: Journal Article
Jazyk:angličtina
Vydáno: Cham Springer International Publishing 01.12.2025
Springer Nature B.V
Springer
Témata:
Accuracy
AI technology
Algorithms
Artificial Intelligence
Business environment
Computer Science
Deep learning
Digital economy
Distance learning
Education
Engineering
English study
Feature parameters
Feedback
Foreign language learning
Fourier transforms
Globalization
Grammar
Language proficiency
Methods
SMFCC
Speech
Students
Voice recognition
Wavelet transforms
China
Feature parameters
SMFCC
English study
Business environment
AI technology
ISSN:2731-0809
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:With the increasing level of internationalization, traditional foreign language training methods cannot satisfy the current requirement for compound talents’ foreign language abilities in the business environment. Based on this, this study combines artificial intelligence technology to propose an intelligent English training conversation model based on speech recognition and multi-feature parameters. The experimental results demonstrated that: first, in terms of model performance, the prediction accuracy of the training set and verification set reached 0.968 and 0.975, respectively. The word error rate in the cross dataset test was 12%-18% lower than that of the baseline method, and the processing time (4.5s/5.0s/5.7s) of single/double/multi-syllable processing of student groups was increased by more than 20%. Second, regarding speech recognition performance, the single syllable recognition rate was 96.6%, the multi-syllable recognition rate was 94.8%, and the feedback correction efficiency was as high as 98%. The difference from manual scoring was controlled within 0.5 (single syllable) and 0.25 (double/multi-syllable). Third, in the platform testing of application verification, learners’ English application ability in business scenarios improved by 28%, while the multi-syllable recognition rate of social groups (94.6%) and student groups differed by less than 2%. The research conclusion showed that the model achieved breakthroughs in speech recognition accuracy (> 94.8%), real-time response (< 5.8s), and teaching adaptability through a multi-feature dynamic feedback mechanism. The research model could significantly improve the scenario-based ability of business negotiation terminology application (such as increasing the accuracy of contract terms expression by 19%), providing quantifiable and practical intelligent training solutions for business foreign language teaching.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2731-0809
DOI:10.1007/s44163-025-00639-5