Synchronizing tool for audio and transcripts

We seek a server-based utility developed to rapidly synchronize streaming media files with transcripts. Users will upload to their uniquely-named directory a full transcript file in .txt format and an associated media file in .wav, .mp3, .ra, .avi, .wmv, .mpg, or .rm format. The new utility will examine their uploads and store a) the transcript wordcount and b) the duration of the media file. The utility will then scan and quickly return time values for the location of distinctive waveforms within the media file, using a speaker-independent system. It will display a “confidence?? value based on success in mapping these distinctive waveforms with associated words in the transcript. It will insert time code tags into a transcript file copy for each of these successfully mapped anchor points. The utility will then email the user a URL for downloading the tagged transcript, with a cc: to our system administrator with session data.

## Deliverables

1) Complete and fully-functional working program(s) in executable form as well as complete source code of all work done. 2) Installation package that will install the software (in ready-to-run condition) on the platform(s) specified in this bid request. 3) Complete ownership and distribution copyrights to all work purchased. =============== Expanded Description =========== We are looking to have a server-based software utility developed within the next 60 days to automatically associate words in transcript files with corresponding words in media files. The utility will present a web page for end users to enter their email address, password, project title, and turnaround time requirements. It will then prompt the user to upload into a unique directory a (completed) transcript file in .txt format and an associated media file in .wav, .mp3, .ra, .avi, .wmv, .mpg, or .rm format. Next, the utility will examine the uploaded files to determine a) the transcript wordcount and b) the duration of the media file. The session will then propose a default number of desired "anchor" points where words in the transcript will be attempted to be matched with associated portions of the media file. Users will have the option to accept, reject, or increase/reduce this recommended accuracy level.. The utility will then generate a time-stamped session ID number and begin processing the uploaded files. It will quickly scan the most universally recognizeable common waveforms within the uploaded media file, using a speaker-independent phoneme recognition funciton or an open source word recognition system (e.g. Sphinx). Time values for these recognized waveform points will be gathered by the utility. It will then create a copy of the transcript text file, and scan it to identify correspondances of words with the media file's recognized "anchor" points. The utility should then should display an overall "confidence value" indicating in percentage terms its success in sequentially assigning anchor points in the the media file to associated words (or phonemes) in the transcript. If this confidence level exceeds a user's defined thresholds for the session, the utility will then insert time code tags into the transcript file copy (in hour:minute:second:fractional second format) for each of the successfully mapped anchor points. The utility should then generate an email message to the user indicating that the tagged transcript is available for downloading from a server-based, user-specific directory, and including in this message a hyperlink to this directory. This utility will help users of our freeware/shareware e-learning tool ([url removed, login to view]) easily synchronize transcripts with streaming video

## Platform

The utility should be capable of operating in unattended batch processing mode on Windows 2000 or Windows XP servers, and generate a password protected, web viewable log file and report for our system administrator summarizing the session data (including username, time of session, transcript file wordcount, media file type, size, and time duration).

PHP

Project ID: #2942885

About the project