Practical advice — speech perception

Practical advice — speech perception module

Advice on working with your own material: which files to prepare, how an exercise goes and what to watch out for. General advice common to both modules is here.

1. How the module works. You upload your own material — an audio recording or a text. The application splits it into short phrases, plays them one by one, and you repeat the words you have heard in your own voice. Your utterance is recognised and compared with the original text, so you can see how much you managed to understand.

2. Preferred audio format: MP3. We support a wide range of formats, but *.mp3 works best: the files are small, upload quickly and are processed most smoothly. If your recording is in another format, you can convert it with one of the widely available free audio converters.

3. Preferred text format: TXT. For e-books and documents a plain text file *.txt is best. It is free of the artefacts that clutter the recognised text with unnecessary words and phrases: footnotes, margin comments, running heads and footers, page numbers. We also accept *.docx, *.epub and *.pdf, but after conversion it is worth looking through the text and removing such fragments.

4. Scanned PDFs will not work. A PDF made of photographs or scans of pages contains no text layer — it is an image, not text. Such a file is rejected with a message, because the application would have nothing to read. Look for a text version of the material, or use a text recognition program (OCR) and save the result as *.txt.

5. Length of the material. Longer recordings are split automatically into segments of about 10 minutes and processed one after another — you can start working as soon as the first segment is ready. For daily practice shorter material works better: it is easier to come back to it the next day.

6. Surroundings and equipment. Work in a quiet place — background noise makes both listening and the recognition of your utterance harder. Speak calmly, towards the microphone, using the whole phrase rather than single words. If your hearing differs between sides, feed the signal through an induction loop to the chosen implant.

7. Repetitions and pace. You can listen to the phrase again with the Replay recording button. Do not rush: it is better to work carefully through a few dozen phrases than to race through a whole chapter. In beginner mode you see the text of the phrase after answering; in advanced mode there is no text and you rely on hearing alone. Start with beginner mode.

8. Your material. Work with material you may legally use. The choice of material stays with you. The file goes to your account only, nobody else sees it, and the audio is deleted from the server once the segment has been processed — what stays in the database are the results of your work, not the file itself.

9. Reporting problems. If a file will not upload, processing does not finish or recording does not work, write to earplanted@uwb.edu.pl. Attach a screenshot of the message and the name of the file that caused the problem.