Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,59 @@
+---
+license: apache-2.0
+task_categories:
+  - text-retrieval
+tags:
+  - information-retrieval
+  - lotte
+---
+
+# LoTTE (small extracts)
+
+Small, teaching-sized extracts of [LoTTE](https://github.com/stanford-futuredata/ColBERT/blob/main/LoTTE.md)
+(*Long-Tail Topic-stratified Evaluation*), built from the public
+[`colbertv2/lotte_passages`](https://huggingface.co/datasets/colbertv2/lotte_passages)
+dataset (Apache 2.0). Prepared for the practicals of *Deep Learning pour le
+texte* (Formation Continue, Sorbonne Université) — see
+`practicals/scripts/prepare-lotte-corpus.py` in the course repo for the build
+script.
+
+Each variant is a self-contained folder:
+
+```
+<variant>/
+    documents/     impact-index DocumentStore (key `docno` -> text)
+    queries.tsv    qid, query
+    qrels.tsv      qid, docno, label
+```
+
+## Variants
+
+- **`technology-3000`** — technology domain, dev split. 150 search queries
+  (with qrels), 3 365 documents (365 required by the qrels + a uniform random
+  sample of ~3000 others).
+
+## Loading
+
+```python
+from huggingface_hub import snapshot_download
+import impact_index, pandas as pd
+
+local_dir = snapshot_download(
+    "bpiwowar/lotte", repo_type="dataset", allow_patterns="technology-3000/*"
+)
+docstore = impact_index.DocumentStore.load(f"{local_dir}/technology-3000/documents")
+queries_df = pd.read_csv(f"{local_dir}/technology-3000/queries.tsv", sep="\t", dtype=str)
+qrels_df = pd.read_csv(f"{local_dir}/technology-3000/qrels.tsv", sep="\t", dtype=str)
+```
+
+## Citation
+
+```bibtex
+@inproceedings{santhanam-etal-2022-colbertv2,
+    title = "{C}ol{BERT}v2: Effective and Efficient Retrieval via Lightweight Late Interaction",
+    author = "Santhanam, Keshav and Khattab, Omar and Saad-Falcon, Jon and Potts, Christopher and Zaharia, Matei",
+    booktitle = "Proceedings of NAACL-HLT 2022",
+    year = "2022",
+    url = "https://aclanthology.org/2022.naacl-main.272",
+}
+```