With any new technology, there is often confusion about what we do and just as importantly, what we don’t do. Our product, SpeechMED, is a multilingual voice platform that empowers you to hear medication notifications, care instructions, and vital information in the language that you understand best. With our proprietary machine learning engine, we are the only software that provides medication reminders in different languages with 100% accuracy. We make it easy for users, with our mobile application, to have their information anywhere and at any time.
We are addressing a health literacy crisis that marginalizes the elderly, minorities, and those with limited English proficiencies. We are doing this by using cutting-edge technology that puts the power of voice and language in the user’s hands. Some of SpeechMED’s many multilingual audio capabilities include:
- Reminders to help avoid missed appointments and medicine doses
- Medication notifications with a picture of a specific pill, dropper, or applicator used for each dose
- Multilingual medication and health care educational materials in multiple formats, including audio text, documents, and video animations
- Immediate access to personal medical information that users can hear and understand in the language and format users prefer
- Assurance that caregivers, regardless of their native language, can understand instructions related to patient care and the medications being administered
- Management of allergies, medications, and other vital health information that a user needs for emergencies or their next medical appointment
- The ability to record audio notes from providers or health professionals that can be heard on repeat by the user or caregiver
- Secure document storage so that the user can access their COVID vaccination card, discharge instructions, blood results, will, or medical record on the fly
- Compliance reminders to emergency contacts and providers if the user has missed their medication, dosage, or follow-up appointment
- Privacy. Privacy. Privacy.
What we don’t do is replace human translators with a direct translation.
We also don’t provide on-the-fly translation of medical documents.
Why? Because no machine translator is skilled enough to give medical information with 100% accuracy the way a human translator can. And while many companies and academics continue to spend millions of dollars on machine translation research, producing accurate translations in a clinical setting is limited by Polysemy, the coexistence of many possible meanings for a word or phrase. Often, a direct translation may not give an accurate equivalent word.
Take the word “tears” as an example. Are we talking about tears that are salty, wet, and painful, or are we talking about water-like fluid secreted by the tear glands? Or are we talking about something that rips? And that’s only a few meanings of the word “tears.” There are literally hundreds of similar examples of words and expressions in English alone. Even with these difficulties in mind, neural networks trained on hundreds of millions of words have been shown to produce some level of accuracy of up to 70% for some languages like English, French, and Spanish.
Several other problems that prevent accurate translation are due to cultural differences, and languages that are structured differently from the language of origin. There may also not be an equivalent word or a way to express something in another language. For example, Cha’paala, spoken in South America, has no word for “thank you.” For some languages like Siwu (spoken in Western Africa) and Lao (spoken in Southeast Asia) saying “thank you” can be even perceived as out of place and bizarre.
An administrator at a children’s hospital told me that he used Google Translate to write a quick email about the fact that he was going to “expedite the discharge of their son.” Because there was no word match for discharge, he basically sent them an email that said he was going to “hurry along to shoot their son.”
For accurate translations, the context of the word, sentence, and phrase must be understood. In spite of its ability to process data, AI is not capable of fully understanding complex sentences or paragraphs in medical text without the help of human experts.
And that’s where we come in. SpeechMED’s proprietary platform fills in the gaps that human translation cannot fill. By combining the power of machine translation with the skills of human translators and medical transcriptionists with expertise in specific fields, we have broken down complex medical processes and reminders into a digestible format that makes it easy for patients to understand and follow in any language.
