BUILDING A VIRTUAL EMAIL VOICE ASSISTANT FOR VISUALLY IMPAIRED PERSONS

Visually impaired persons encounter numerous complications in their mundane activities for impaired vision. Though there are different technologies, systems interaction procedure or expenditure as the key factors here, to provide a solution to this problem, a virtual email assistant was developed. The speech of recognition model was built using Way2Vec framework, and the trained model was assessed in terms of word error rate and character error rate. Finally, a responsive web-based user interface was designed and developed using Bootstrap. The result of this study demonstrates that fine-tuning a small amount of labeled speech data can reach results comparable to those of state-of-the-art ASR systems. The researchers fine-tuned Wav2Vec2 without using a language model and only used 188,422 samples of the Common Voice English dataset. It has been demonstrated that a standalone Wav2Vec2 acoustic model produced impressive results of 15.4% WER and 6% CER on the Common Voice English validation dataset. In addition, the developed virtual voice email assistant application provides visually impaired people to send and receive emails.

top viewed thesis

1. Records Management System of Municipal Agriculture, Sto. Tomas, La Union 2022-11

2. Record Management System for the Instruction Office of DMMMSU-SLUC 2018-04

3. Online Ordering System for Ryan and Ced Pizza 2022-05

4. Development of Online Examination System for Joyland School, Inc. 2022-05

5. Network Infrastructure Plan for Don Eufemio F. Eriguel Memorial National High School 2017-04