Frequently Asked Questions about MOR-ASR

In the context of MOR-ASR, it stands for AI Assisted Summary for Medical Records — a technology that converts medical conversations into text and summarizes the key points into a format that can be used to support medical records.

MOR-ASR uses AI Assisted Summary for Medical Records technology developed specifically for medical work. The system receives the doctor's spoken input and transcribes and summarizes it in both Thai and English, supporting medical terminology through AI and a Large Language Model tuned for real-world clinical use.

Users can access MOR-ASR through Cariva Space by following these steps:

1. Go to http://space.cariva.co.th
2. Enter the E-mail and Password registered with Cariva
3. Click the Login button to sign in and start using the service

MOR-ASR can be used as a Web Application on any device with a browser, internet connection, and microphone, such as a computer, tablet, or other web-capable device.

For organizations that need to connect with a HIS, integration via SDK can be considered.

MOR-ASR outputs support both Thai and English, and can retain English medical terms where appropriate, such as disease names, medications, procedures, or abbreviations doctors use when documenting.

There are currently two main modes for OPD and IPD work:

1. OPD-Clinical Record ASR for outpatient documentation
2. IPD-SOAP ASR for inpatient documentation in SOAP format

An internet connection is required throughout use, as the system needs to send audio for processing and return the results to the user.

General Voice to Text typically focuses on transcribing speech word-for-word without structuring or summarizing it in a medical context.

MOR-ASR is designed to summarize key medical information from conversations and organize it into a Medical Record with clearly defined sections, making it easier for doctors to review and use in support of medical records.

Yes. The system is designed and tuned to understand medical terminology, abbreviations, and clinical documentation context better than general transcription systems.

MOR-ASR reduces the burden of documentation and typing medical records, giving doctors more time for patient care. Key benefits include:

1. Less time spent typing and organizing medical records
2. Reduced risk of typos or omissions
3. Fast results after the recording ends
4. Lets doctors focus more on the conversation and patient care

There are a few considerations for the best results:

1. Use it in a low-noise environment
2. Unclear pronunciation may affect accuracy
3. Certain accents or speaking styles may be harder for the system to process
4. Doctors should always review the results before saving them to the medical record

Generally, MOR-ASR supports continuous speech of about 5–10 minutes per session. The ideal duration may depend on the use case, case length, and internet quality.

MOR-ASR suits medical documentation across many specialties, such as general practice, internal medicine, surgery, pediatrics, obstetrics & gynecology, emergency medicine, and other specialties.

Examples of supported specialties include Cardiology, Pulmonology, Gastroenterology, Neurology, Nephrology, Oncology, Endocrinology, Dermatology, Allergy and Immunology, Orthopedic, Ophthalmology, and Otolaryngology (ENT).

MOR-ASR accuracy is evaluated using the Concept Error Rate concept, prioritizing capturing all key medical information over word-for-word transcription.

Results are grouped into 4 categories:

1. Green: the system captured the concept correctly and no editing is needed
2. Yellow: the meaning is correct, but wording, phrasing, or minor details should be adjusted
3. Red: the system clearly misinterpreted and should be corrected before use
4. Gray: a concept that should have been summarized, but the system did not transcribe or summarize it

Accuracy may vary by subspecialty. For example, General Surgery has 91.11% green, 4.44% yellow, 2.22% red, and 2.22% gray.

In OPD ASR mode, the user presses the record button, speaks the information to document, and presses Submit for the system to process. Minor slips, pauses, or thinking while recording don't require stopping and re-recording.

If the workflow needs a break — such as performing a physical exam first and then continuing to speak — the user can pause and resume recording as appropriate.

SOAP ASR mode works like OPD ASR: press record, speak the information, and press Submit to process. The difference is that the output is organized in SOAP format — S, O, A, and P — which suits inpatient documentation or cases that require a SOAP structure.

MOR-ASR is designed to process audio to produce results, and does not store users' audio files after processing is complete.

Sending more processing requests than the system or usage agreement supports may affect service stability and the limits on the hospital's side.

In real use, concurrent volume and rate limits should be set appropriately together with the Cariva team and the hospital's IT team.

In principle, Cariva does not store patient-identifying data alongside the audio used for processing. During processing, the system does not know which patient the audio belongs to.

Once processing is complete, the system returns the result to the hospital, and the hospital matches the result to the medical record or patient data in its own system. For details on connection security measures, this should be evaluated with the Cariva team based on the actual integration model.

MOR-ASR follows a Zero Retention principle for data used in processing — that is, it does not retain customer data after the service is complete. Audio files are deleted after processing according to the service terms agreed with the customer.

The system uses a Large Language Model further fine-tuned by Cariva for the medical context, starting from an open-source model and trained on medical knowledge.

The process relies not only on the core model but also on pre-processing and post-processing steps to make the results more suitable for use in medical records.

Processing time depends on several factors, especially internet speed and audio length, since the system needs to upload the audio for processing and return the result to the user.

The system is designed to scale resources automatically, which helps reduce impact when many doctors use it at the same time. However, actual performance may depend on usage volume, package terms, and the hospital's system environment.

Very loud background noise may affect processing quality and result accuracy, so it's best to use the system in the clearest-sounding environment possible.

If there is general speech unrelated to the medical content, the system is designed to focus primarily on medical terms and treatment-related content.

The recommended bandwidth is about 100 Mbps, referencing support for roughly 10 cases per second. This should be evaluated alongside actual usage volume and the hospital's system architecture.