mangoesai/StructuredHarmonyFrequency20240918

Medical Conversation Dataset

Overview

This dataset contains transcribed medical conversations between doctors and patients. It provides valuable data for research in medical dialogue analysis, natural language processing in healthcare, and related fields.

Dataset Structure

The dataset is organized as follows:

Doctor_1/
├── Patient_1/
│   ├── GMT20230103_DCC_1010.csv
│   ├── GMT20230104_DCC_1011.csv
│   └── README.md
├── Patient_2/
│   ├── GMT20230105_DCC_1012.csv
│   └── README.md
└── ...

• Each doctor has a folder.
• Within each doctor's folder, there are folders for each patient.
• Each patient folder contains:
  • CSV files for individual meetings (named with the format GMT[DATE]_DCC_[NUMBER].csv)
  • A README.md file with information about the meetings between that doctor and patient.

Data Fields

Each CSV file contains the following fields:

• doctor_name: Name of the doctor (anonymized if necessary)
• patient_name: Name of the patient (anonymized if necessary)
• meeting_number: Unique identifier for the meeting
• transcription: The transcribed text of the conversation
• chunk_speaker: Identifier of who is speaking in this chunk of conversation
• time_interval: The time interval for this chunk of conversation
• 'date': The time of the conversation

Usage

To load the entire dataset:

from datasets import load_dataset

dataset = load_dataset("your-username/your-dataset-name")

This will give you a DatasetDict with a 'train' split containing all the data.

To access specific parts of the dataset:

# Get all data
all_data = dataset['train']

# Get a specific row
first_row = all_data[0]

# Filter data
doctor_data = all_data.filter(lambda example: example['doctor_name'] == 'Dr. Smith')