2,474,584 rows of molecular data pairing canonical SMILES strings with natural language descriptions. The dataset is structured into a single training split containing unique identifiers, chemical structures, and textual explanations for each compound.
Use Cases
- Train a translation model to convert canonical_smiles into natural language description strings.
- Develop a text-to-molecule generation system that outputs canonical_smiles based on queries in the description field.
- Perform semantic search for chemical compounds by indexing the description column.
Strengths
- Contains 2,474,584 rows in a single training split.
- Includes canonical_smiles strings representing the chemical structure of each entry.
- Provides a description column containing natural language text associated with each molecule.