New Captioning Webcasts Archived

Submitted by Barbara Keefe, PEPNet-Northeast

Two archived webcasts were streamed live last fall in Portland, Maine at an accessible distance technology conference. PEPNet collaborated with the National Assistive Technology Technical Assistance Partnership and Maine CITE.

Current Distance Learning Solutions:
Captioning Educational Materials

Kevin Erler of Automated Sync Technologies (AST) presented an informative review of issues connected to captioning educational materials. The webcast outlines the results of AST's investigation of approaches to captioning and highlights the pros and cons of each. Erler focuses on emerging technologies and the challenge of captioning new distribution models on the web, e.g., iTunes, YouTube, etc. Captioning quality is examined in an interesting and forthright manner and he points out that the mission of most universities includes "excellence" and setting high standards; he suggests those should include captioning standards.

Erler compares the error rates of a trained stenographer, a student transcriber, and trained and untrained speech recognition providers. He raises the issues of "fixing" the speech recognition output pre-processing and the human challenges involved in motivating faculty to "train" on a specific device. He also discusses post-processing issues and says the cost of repairing errors is higher than the cost of starting from scratch.

Erler reminds us that states are increasingly requiring more than basic accessibility, adopting laws that call for effective programmatic access. Captioning can help provide improved access to educational programs and products.

Finally, he advises decision makers that accuracy is key to captioning; don't waste resources on inadequate solutions.

Current Distance Learning Solutions:
Liberating Speech through Captioning

Jacob Butcher's webcast outlined the mission and work of Docsoft and its automatic speech recognition product. Docsoft's captioning solution can capture lectures, create searchable content of spoken words, and leverage the power of speech recognition technology.

This appliance has a dedicated hardware and software package designed to service multiple users. It has 80% average word recognition with quality input and a higher percentage with speaker training. The product automatically exports caption files and creates caption files from existing text transcripts.

Butcher emphasizes that the quality of speech recognition depends on audio quality. Some factors affecting audio quality include microphones, file compression, pronunciation and accents. He also describes the effects of environmental factors such as microphone placement, building noise, number of speakers, acoustics and speaker format on captioning accuracy.

Key to the success of automatic speech recognition is training the speaker. A 20% increase in accuracy occurs in building word lists in highly relevant documents. Butcher also says that editing tools designed for speech recognition output can expedite and improve the review and editing processes.

Success stories of a number of universities providing access using speech recognition are shared along with the variations and accommodations made by each university. Butcher concludes his presentation with an account of the benefits of using an automatic speech recognition product, including cost savings, elimination of outsourcing, decreasing the time it takes to receive captions and transcripts, and the increased value of digital assets.

To view either archived presentation, go to:
www.pepnet.org/sermat.asp