add files
Browse files- README.md +11 -0
- dev.jsonl.gz +3 -0
- train.jsonl.gz +3 -0
README.md
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#NQ-retrieval
|
| 2 |
+
|
| 3 |
+
This is a nicely formatted version of the [Natural Questions](https://ai.google.com/research/NaturalQuestions/) dataset, formatted to train and evaluate retrieval systems.
|
| 4 |
+
|
| 5 |
+
Each row contains the following entries:
|
| 6 |
+
- **question**: Original question send for Google Search Engine
|
| 7 |
+
- **title**: Title of Wikipedia article
|
| 8 |
+
- **candidates**: A list with the passages from the original Wikipedia HTML document
|
| 9 |
+
- **passage_types**: Types (text, table, list) of the candidate passages
|
| 10 |
+
- **long_answers**: IDs which candidate passages where selected as relevant from annotators. Might be empty if no relevant passage has been identified
|
| 11 |
+
- **document_url**
|
dev.jsonl.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad23b7e4f50b0f02c9395a4f8fe39946e3e1242edea7c826aaf7ac378f3e8779
|
| 3 |
+
size 72661266
|
train.jsonl.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9eb0988b337005ebcc871dd56df006ef042210e77d59cd6e0645a3993c5d3f1e
|
| 3 |
+
size 2968118739
|