Does not work on Medical Transcription

There is a whole lot of difference when I feed the wav file of a physician in deepspeech model. There are a whole lot of mistakes other than medical termiologies. Can anyone guide me?

STT with DeepSpeech based on wav file:
coracomorphae mental come his dry of this life sebaceous ion come hyperpituitary of corney ardices with pride conistra so she come multiple falls reportedly book up this morning and walk to kitchen period he was doing a kayite morning but later in the afternoon he told nacionales period he will also warieties before coming to hospital period bearpark bedecorated that behind to be diacritic was as sweet key with dried by the branch block period patient is confused period family fostered but they would only one comfort measures period during the session with the ear physician the latter changed their mind and wanted patient to be treated with antibiotics but at the same time they wanted comfort measures well continued the home medication period they do not want any resuscitation but at the same time as for floods to be given to patient peritoneal perform in his revered all these to us being taken from patients family perpetrated that pechorin hospice
Inference took 245.054s for 71.038s audio file.

Original Document:
Caucasian male with vascular dementia, history of TIA/CVA, hypertension, hyperlipidemia, history of coronary artery disease with prior coronary artery bypass graft surgery, multiple falls reportedly woke up this morning and walked to the kitchen. He was doing okay in the morning but later in the afternoon he developed nausea and malaise. He also vomited couple times before coming to hospital. Patient found to be tachycardic which was SVT with right bundle branch block. Patient is confused. Family first reported that they would only want comfort measures. During discussion with the ER physician they later changed their mind and wanted patient to be treated with antibiotics but at the same time they wanted comfort measures as well continue the home medications. They do not want any resuscitation but at the same time asked for fluids to be given to patient. Patient is unable to perform any history. All the history was being taken from patient’s family. Patient daughter reported that patient was in hospice while he was at the facility few days ago but later hospice was denied. She does not want any invasive procedures. She wanted to see if antibiotics will help patient otherwise they would like to take a decision in the next day or so.
Patient is febrile at the time of presentation temperature was 102.3 F.
Patient was given antibiotics in the emergency room mainly cefepime doxycycline as well vancomycin. He is not complaining any pain. He has some cough. He is somewhat confused but recognizes his family.

Assessment and plan:
Sepsis: He appears to be in sepsis. His labs showed lactic acid was elevated to 3.4. He has been started on antibiotics. I will continue the fluids as patient family wants. We will also continue the home medications. Patient cultures have been sent. Antibiotics can be discussed again in the morning with the patient’s family. He has been given cefepime, doxycycline, vancomycin in the emergency room.

Comfort measures: Patient family also want all comfort measures and keep patient DNR/DNI. Will start patient on fentanyl patch. We will also give every 4 hours as needed morphine. Continue Zofran.
We will do palliative care consult in the morning.

I really don’t thank you for not making the basic effort of reading the guidelines that I took time to link to you on the Github issue you mis-filled because you did not bother reading the issue template that explicitely states, in the first words, to reach for support on Discourse.

Now what? I have read the guidelines. Your guidelines tells me topics which doesnt fit into and it is rather you need to make your support better

We are not here to provide you business paid support.

You don’t share any basic informations on how you build / run your model. All you share is “here is my data, here is what was transcribed”. You don’t even bother telling what version of the lib and model you are using.