| |
|
|
| """ |
| A script for creating splits for the copenhagen dependency treebank which align with DDT. |
| """ |
|
|
| import json |
| from pathlib import Path |
|
|
| from conllu import parse |
|
|
|
|
| def load_cdt( |
| assets_path=Path(__file__).parent.parent / "assets", |
| ): |
| """ |
| Load the copenhagen dependency treebank / DaCoref dataset |
| """ |
| cdt_path = assets_path / "dacoref" / "CDT_coref.conllu" |
| with cdt_path.open(encoding="utf-8") as f: |
| text = f.read() |
|
|
| sentences = parse( |
| text, |
| fields=[ |
| "id", |
| "form", |
| "lemma", |
| "upos", |
| "xpos", |
| "feats", |
| "head", |
| "deprel", |
| "deps", |
| "misc", |
| "coref_id", |
| "coref_rel", |
| "doc_id", |
| "qid", |
| ], |
| ) |
|
|
| split_ids = {"train": [], "dev": [], "test": []} |
|
|
| for split in split_ids: |
| split_path = assets_path / "dacoref" / f"CDT_{split}_ids.json" |
| with split_path.open(encoding="utf-8") as f: |
| split_ids[split] = json.load(f) |
|
|
| return sentences, split_ids |
|
|
|
|
| sents, split_ids = load_cdt() |
|
|
|
|
| doc_ids = {} |
|
|
| _doc_id = "nan" |
| for sent in sents: |
| if "newdoc id" in sent.metadata: |
| _doc_id = sent.metadata["newdoc id"] |
| sent_id = sent.metadata["sent_id"] |
| split = sent_id.split("-")[0] |
| if _doc_id not in doc_ids: |
| doc_ids[_doc_id] = [] |
| doc_ids[_doc_id].append(split) |
|
|
| test_docs = [] |
| for s in split_ids["train"]: |
| if s not in doc_ids: |
| print("not in doc_ids") |
| break |
|
|
| if "test" in doc_ids[s]: |
| test_docs.append(s) |
|
|
| new_split = {"test": test_docs} |
| dev_no_test = [s for s in split_ids["dev"] if s not in test_docs] |
| new_split["dev"] = dev_no_test |
| train_no_test = [s for s in split_ids["train"] if s not in test_docs] |
| test_no_test = [s for s in split_ids["test"] if s not in test_docs] |
| new_split["train"] = train_no_test + test_no_test |
|
|
| with open("assets/CDT_ddt_compatible_splits.json", "w") as f: |
| json.dump(new_split, f, indent=2) |
|
|
