The dataset viewer is not available for this split.
Parquet error: Scan size limit exceeded: attempted to read 1538110091 bytes, limit is 300000000 bytes
Make sure that
1. the Parquet files contain a page index to enable random access without loading entire row groups2. otherwise use smaller row-group sizes when serializing the Parquet files
Error code: TooBigContentError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
CaBot Clinical Literature Embedding Index
Exact embedding-search index over 3,474,244 works from 204 high-impact clinical
journals (2023 JIF >= 10), built from an OpenAlex snapshot (~June 2025) and
embedded with OpenAI text-embedding-3-small at 1536 dimensions (float32).
This is used for CaBot. The build and search code is in the CaBot-Search/
folder of the source repository.
Columns
| column | type | notes |
|---|---|---|
| id | string | OpenAlex work id |
| doi | string | |
| title | string | |
| abstract | string | reconstructed from OpenAlex inverted index ("" if none) |
| journal | string | |
| year | int32 | |
| publication_date | string | YYYY-MM-DD |
| cited_by_count | int32 | |
| authors | list[string] | raw author names |
| is_pubmed_indexed | bool | |
| is_open_access | bool | |
| article_type | string | OpenAlex type |
| has_abstract | bool | |
| embedding | list[float32] x 1536 | text-embedding-3-small, document text = lowercased title (+ abstract) |
Reproducing search
from datasets import load_dataset
ds = load_dataset("tbuckley/cabot-search", split="train") # streams the parquet shards
Embed queries with a "query: " prefix and use cosine similarity (L2-normalize
- inner product) for exact search. See
04_search.py.
- Downloads last month
- 126