Exacom is pleased to announce audio transcription services are now available within the latest version of HindSight 4.
Transcribed audio allows agencies to respond faster to FOIA requests for court proceedings; instead of a staffer manually transcribing minutes or hours of audio, that work is done automatically as needed. Transcriptions also make it easier for agencies to perform Quality Assurance (QA) on specific recordings. Evaluators and supervisors can find keywords or topics in a recording quickly and highlight important recorded incidents.
Exacom’s transcription services within HindSight 4 are available on an on-demand, pay-as-you-go basis. This structure ensures that communication centers operate within their expected budgets and keeps costs in check, while still allowing users to automate this once-manual process.
Users can select one or multiple recordings, request the transcription(s), and in moments, the automatically generated transcription text is available in the HindSight 4 client. Once generated, transcriptions are fully searchable. When users download recordings, they have the option to include the generated transcriptions as a .txt file along with the audio and metadata files. Keeping in line with the permissions-based structure of HindSight 4, agencies can control which users can just see transcriptions and which can request transcriptions.
Aiming for Quality and Accuracy
“Exacom is not the first logging recorder company to bring this product to the public safety, corporate and campus security, utility, transportation, and government markets,” said Exacom Product Manager Mark Woody. “However, we’ve made sure that what we bring customers is going to be as accurate as technologically possible, and therefore useful and beneficial to them.”
Most AI transcription services are fairly accurate when it comes to transcribing phone audio in a call center setting, because phone audio is typically quite clear and the AI models have been trained for a specific set of scenarios. The words spoken over radio transmissions, however, are often less intelligible to most AI platforms (and humans, too). This makes transcriptions from most AI transcription services for radio audio quite inaccurate, rendering the generated transcriptions much less meaningful. With 911 phone calls, while the quality of the calls is usually pretty good, the content can be anything and there can be a lots of background noise, emotions, multiple speakers in the background, etc. All of this makes it very challenging for the AI models that do the transcription, resulting in lower quality output.
“When Exacom set out to design our transcription service, we felt it was important that our customers have a high degree of quality and accuracy for ALL of their communications,” said Woody. “We found a transcription partner for phone audio that understood the special needs of mission-critical communications, especially public safety. Then, we sought out a partner who specialized in transcribing often-difficult-to-understand radio audio. Our search led us to one who had specially trained voice models for radio and, with that, offered astounding accuracy for transcribing radio communications.”
With this two-pronged approach to phone and radio transcriptions, Exacom’s logging recorder software can record multimedia from various communications systems on a single system and, now, render clear, accurate transcriptions for the majority of recordings.