Update README.md

README.md

@@ -33,3 +33,70 @@ configs:
   - split: train
     path: data/train-*
 ---
+
+# Dataset Card for MRREL Data
+
+## Dataset Description
+### Dataset Summary
+The MRREL data is a component of the UMLS (Unified Medical Language System) Metathesaurus, available in the Rich Release Format (RRF). This dataset contains information about relationships between medical concepts and atoms, providing a detailed resource for understanding the interconnections within the UMLS Metathesaurus. It supports developers and researchers in comprehensively representing and utilizing medical terminology and their relationships.
+
+### Purpose
+The MRREL dataset offers structured information on the relationships between medical concepts and atoms. This data is crucial for medical informatics applications, clinical decision support systems, and other health-related software solutions, aiding in the precise representation and understanding of the relationships among medical terms.
+
+## Column Descriptions
+### CUI1 (Concept Unique Identifier 1)
+- **Description**: A unique identifier assigned to the first concept in the relationship within the UMLS Metathesaurus.
+- **Values**: Alphanumeric strings, such as `C0001175`. Each value uniquely identifies a medical concept.
+
+### AUI1 (Atom Unique Identifier 1)
+- **Description**: A unique identifier for the first atom in the relationship, which is a specific occurrence of a string in a source vocabulary.
+- **Values**: Variable length field, typically 8 or 9 characters, such as `A0019180`. Each value uniquely identifies an atom within the dataset.
+
+### STYPE1 (Source Type 1)
+- **Description**: The name of the column in MRCONSO.RRF that contains the identifier used for the first element in the relationship, e.g., AUI, CODE, CUI, SCUI, SDUI.
+- **Values**: Alphanumeric strings indicating the source type, such as `AUI`, `CODE`, `CUI`, `SCUI`, `SDUI`.
+
+### REL (Relationship)
+- **Description**: The type of relationship of the second concept or atom to the first concept or atom.
+- **Values**: Alphanumeric strings indicating the relationship type, such as `RN`, `RO`, `RQ`, `SY`, `RL`, `CHD`, `PAR`, `QB`, `QB2`, `QB3`.
+
+### CUI2 (Concept Unique Identifier 2)
+- **Description**: A unique identifier assigned to the second concept in the relationship within the UMLS Metathesaurus.
+- **Values**: Alphanumeric strings, such as `C0001175`. Each value uniquely identifies a medical concept.
+
+### AUI2 (Atom Unique Identifier 2)
+- **Description**: A unique identifier for the second atom in the relationship, which is a specific occurrence of a string in a source vocabulary.
+- **Values**: Variable length field, typically 8 or 9 characters, such as `A0019180`. Each value uniquely identifies an atom within the dataset.
+
+### STYPE2 (Source Type 2)
+- **Description**: The name of the column in MRCONSO.RRF that contains the identifier used for the second element in the relationship, e.g., AUI, CODE, CUI, SCUI, SDUI.
+- **Values**: Alphanumeric strings indicating the source type, such as `AUI`, `CODE`, `CUI`, `SCUI`, `SDUI`.
+
+### RELA (Additional Relationship Label)
+- **Description**: An optional, more specific label for the relationship.
+- **Values**: Alphanumeric strings providing additional specificity for the relationship, such as `isa`, `part_of`, `has_part`, `associated_with`. There are 964 unique values.
+
+### RUI (Relationship Unique Identifier)
+- **Description**: A unique identifier for the relationship.
+- **Values**: Alphanumeric strings, such as `R123456`. Each value uniquely identifies a relationship within the dataset.
+
+### SL (Source of Relationship Labels)
+- **Description**: The source of the relationship labels.
+- **Values**: Alphanumeric strings indicating the source, such as `MSH`, `SNOMEDCT_US`, `LNC`, `CSP`.
+
+## Example Record
+```
+CUI1:       C0001175
+AUI1:       A0019180
+STYPE1:     AUI
+REL:        RN
+CUI2:       C0001176
+AUI2:       A0019181
+STYPE2:     AUI
+RELA:       isa
+RUI:        R123456
+SL:         MSH
+```
+
+## Usage
+This dataset is primarily intended for developers and researchers working on medical informatics applications. It can be used to ensure accurate and comprehensive understanding of the relationships between medical terms and their attributes, support clinical decision-making, and enhance health information systems.