For busy readers
- Transcription converts spoken words into written text. This text can then be used in various ways within a business context. We call this speech automation.
- Summaries of conversations, video conferences, or YouTube videos are the most well-known use cases. However, through AI, numerous additional application-specific reports can be created and further automation processes can be initiated.
- The prerequisite is to unambiguously identify the conversation participants in the recording and correctly assign the texts to them. This process is called diarization ("diary-keeping").
- Diarization enables speaker-specific interpretation of content and its utilization. It is the foundation for automatically generated medical letters, lawyer-client conversations, order documentation in banking and insurance, and much more.
- Additionally, follow-up processes can be automatically triggered, for example when a supervisor approves a measure during a conversation, which then initiates and completes an approval process in the ERP system.
Tip to try out
If you use ChatGPT, you should check out the new prompt guide from OpenAI. The maker of ChatGPT has published a dedicated prompt creation guide on how a good and meaningful prompt should look in ChatGPT -- and also via the API -- to achieve the highest quality results. It's worth highlighting that OpenAI generally writes very understandable documentation, so even non-IT professionals can get the most out of ChatGPT, DALL-E, and Whisper.
Actions Require Precision
If transcription is to go beyond mere speech recognition and translation of spoken words and sentences, the unambiguous assignment of what was said to individual speakers is necessary.
Video conference providers like Microsoft Teams, Zoom, Google Meet, GoToMeeting, or Cisco WebEx can already identify each speaker in their products and precisely assign their statements, since each video conference participant uses their own channel. This fundamentally works reliably, setting aside minor assignment errors during interruptions (e.g., when "talking over each other").
If you want to automatically create medical documentation based on one or more doctor-patient conversations and automatically feed it into the hospital information system or practice system for documentation purposes, then using the aforementioned video conference systems is often not practical. Although the doctor can help themselves by speaking the essential information into their smartphone during or after the appointment, from which an automatic transcription process takes place, there is an understandable desire to directly process the normal doctor-patient conversation so that full attention can be given to the patient.
Diarization
AI-based transcription platforms like the OpenAI Whisper model can convert entire audio files into text files -- making them accessible for further processing -- but they do not offer the ability to identify individual speakers, leading to misinterpretations by the AI model when, for example, complaints at the beginning of the hospital admission report are to be listed separately.
To identify speakers (e.g., doctor, patient, nurse, family member, etc.), other AI models are used. They are called diarization models and return a list of entries showing which speaker spoke from which second to which second.
With this information, the recording is then further processed through transcription models into text, so that in the subsequent, also AI-based text analysis, the information about who said what can be utilized. This is important for differentiating content. For example, the complaint comes from the patient, while the therapy suggestion comes from the doctor. Without the vocal differentiation -- as is the case with text -- no computer can unambiguously assign what was said. Misinterpretations would increasingly creep in, which we must avoid, especially in critical areas.
Use Cases
This combination of multiple AI models enables the automation of industry-specific use cases. How AI agents in business orchestrate such processes is explained in a separate article.
Medical letters and nursing reports can be automatically generated and delivered to the desired recipient. Lawyers and tax advisors can document the results of their consulting conversations and the next steps agreed upon with their clients in their digital files. Banks and insurance companies can not only track orders and customer interactions but also immediately initiate automated actions such as buy or sell orders or sending a policy.
Customer service desks and helpdesks can take bookings with specific details shared by the customer during the conversation or activate or deactivate licenses for the caller. Our free audio-to-text converter shows how easy getting started with transcription can be.
What all use cases share is that artificial intelligence can interpret the meaning of the conversation and, thanks to speaker assignment, place it in context. This enables further automation processes to be initiated in downstream systems without explicit human action. Human communication serves problem-solving; the implementation is automatically carried out thanks to AI.
Transcription with diarization opens up entirely new possibilities for businesses in any industry to automate their daily operations, increase their own productivity, expand their competitive advantage, and improve employee satisfaction by eliminating monotonous tasks.
In short: Words lead to actions.
