README: drop the project name, exact typed F1 for the first phase
Browse files
README.md
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
language: [en]
|
| 4 |
tags: [ner, crossner, evaluation]
|
| 5 |
-
pretty_name:
|
| 6 |
dataset_info:
|
| 7 |
- config_name: crossner_ai
|
| 8 |
features:
|
|
@@ -127,12 +127,12 @@ configs:
|
|
| 127 |
path: crossner_science/test-*
|
| 128 |
---
|
| 129 |
|
| 130 |
-
#
|
| 131 |
|
| 132 |
-
Part of
|
| 133 |
|
| 134 |
The five CrossNER target domains ([DFKI-SLT/cross_ner](https://huggingface.co/datasets/DFKI-SLT/cross_ner),
|
| 135 |
-
Liu et al., 2021) in exactly the format of `
|
| 136 |
evaluation code scores them: same system and user prompt, same construction of the requested types
|
| 137 |
(up to 6: a random number of the row's present types plus absent distractors drawn with inverse
|
| 138 |
frequency from the domain's train split), plus `raw_text` and `all_gold`.
|
|
@@ -152,4 +152,4 @@ frequency from the domain's train split), plus `raw_text` and `all_gold`.
|
|
| 152 |
- The system prompt is the benchmarks' one verbatim (it mentions biomedical text), so the dataset is the
|
| 153 |
only thing that changes with respect to the biomedical evaluation.
|
| 154 |
|
| 155 |
-
Built with `04_evaluation/build_crossner_eval.py`. The second phase of the project uses a different query protocol on the same sentences (all gold types plus 1-3 distractors): see `
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
language: [en]
|
| 4 |
tags: [ner, crossner, evaluation]
|
| 5 |
+
pretty_name: CrossNER Evaluation Subsets (First-Phase Protocol)
|
| 6 |
dataset_info:
|
| 7 |
- config_name: crossner_ai
|
| 8 |
features:
|
|
|
|
| 127 |
path: crossner_science/test-*
|
| 128 |
---
|
| 129 |
|
| 130 |
+
# CrossNER Evaluation Subsets (First-Phase Protocol)
|
| 131 |
|
| 132 |
+
Part of the code, data and models of the bachelor thesis *Structuring Universal Chain-of-thought for Named Entity Recognition* (Department of Computer Science and Engineering, University of Bologna, 2026). All the repositories cited by name belong to the same organization.
|
| 133 |
|
| 134 |
The five CrossNER target domains ([DFKI-SLT/cross_ner](https://huggingface.co/datasets/DFKI-SLT/cross_ner),
|
| 135 |
+
Liu et al., 2021) in exactly the format of `bioner-eval-subset`, so the same
|
| 136 |
evaluation code scores them: same system and user prompt, same construction of the requested types
|
| 137 |
(up to 6: a random number of the row's present types plus absent distractors drawn with inverse
|
| 138 |
frequency from the domain's train split), plus `raw_text` and `all_gold`.
|
|
|
|
| 152 |
- The system prompt is the benchmarks' one verbatim (it mentions biomedical text), so the dataset is the
|
| 153 |
only thing that changes with respect to the biomedical evaluation.
|
| 154 |
|
| 155 |
+
Built with `04_evaluation/build_crossner_eval.py`. The second phase of the project uses a different query protocol on the same sentences (all gold types plus 1-3 distractors): see `free-vs-structured-cot-eval`.
|