A profiling tool for tabular data leveraging the powerful capabilities of LLMs, appropriate for data scientists.
- Table-related features: Table-related features of TableSage.
- Table-related features: Attribute-related features of TableSage. Dashed lines indicate optional arguments, but at least one should be provided.
| Feature | Sub-Feature | Input | Output |
|---|---|---|---|
| Table | Table Summarization | Table | Property |
| Table Type Annotation | Table | Property | |
| Column | Column Summarization | Table | Property |
| Column Type Annotation | Table | Property | |
| Extraction | Insights Extraction | List of Official Properties (Optional) List of Generated Properties (Optional) (At least one) |
List of Insights |
| Spatial Information Extraction | Table Description | Property | |
| Temporal Information Extraction | Table Description | Property | |
| Comparison | Properties Comparison | List of Official Properties List of Generated Properties |
List of Comparisons |
| Fusion | Alternative Properties Fusion | List of Official Properties (Optional) List of Generated Properties (Optional) List of Insights (Optional) (At least one) |
Property |
| Complementary Properties Merging | List of Official Properties (Optional) List of Generated Properties (Optional) (At least one) |
Property |
pip install tablesage
You can easily use TableSage by:
from tablesage import TableSage
p = TableSage()
p.load_dataset(path=<path>, separator=<separator>)
profile = p.profile_dataset(<model>, endpoint=<openai-endpoint>, token=<token>)
You can find more information about OpenAI endpoints here.
This work was partially funded by the EU Horizon Europe projects STELAR (GA. 101070122)