Validate all CoNLL-U with official loader

README.md

@@ -90,7 +90,7 @@ configs:
 # Sphragis
 
 Sphragis is a benchmark for Ancient Greek authorship
-attribution. Beta 0.4.0 combines licensed, human-annotated dependency treebanks[1]
+attribution. Beta 0.4.1 combines licensed, human-annotated dependency treebanks[1]
 and the metrically annotated [Hypotactic](https://hypotactic.com/)(Chamberlain 2023)
 corpus. It supplies fixed train, validation and test splits.
 
@@ -222,6 +222,12 @@ incompatible source-specific tags are reconstructed from UPOS and FEATS, with
 unavailable distinctions represented by `-`. UPOS and FEATS remain the
 canonical UD fields.
 
+Every distinct published `conllu` document is loaded during testing with the
+[official CoNLL 2018 shared-task `load_conllu` parser](https://github.com/CoNLL-UD-2018/UDPipe-Future/blob/master/conll18_ud_eval.py).
+This checks sentence
+termination, ten-column structure, token IDs, heads, roots and dependency
+cycles across all nine configurations.
+
 ## Deduplication
 
 Raw prose input comprised 72,336 sentences. Authorship curation retained

data/prose_1/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:713c4d5690cdcaacd67100cbece5bbdad3daa05c73d1e25c6f6efda502493c92
-size 1979122
+oid sha256:2f3554227d439dcecff393a3cb48e221b6493bba0b2736cdb283403721aee367
+size 1980714

data/prose_1/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f69ebd63f93e16073f17271c15cec2d05e5d507b29c2b9586c249d24c94c53ba
-size 15617703
+oid sha256:698fff1cff621b2c0de6277bc0ac31db9b8997f00e41ab76531e8691fba551c5
+size 15629296

data/prose_1/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c36a957fc05a83104c9d69ec15b0a395017fa675dd6cf9a07218dce0f57083d8
-size 1987427
+oid sha256:0874552ac52826109722381fa457d9d5a999f9801774c6f74993f8d81d16ff18
+size 1988996

data/prose_10/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4088a5830cca67487b15f84c6f8cb90784bf88a4d9284e8de27fbd5fd9b65c53
-size 1624136
+oid sha256:88cb1a7e347f6cb1136675a5a6279bad3d7583d8dd082ccfcab9cdaf71e0f537
+size 1624165

data/prose_10/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:000acdd4129d34c1ab470dc906c60fed053c014ef3f78a57510042a6b204c8b0
-size 16734742
+oid sha256:26b317b51ff8c1f82cf4f1c40be179a830e2be1801573ecdfc6d176e88088eb1
+size 16747242

data/prose_10/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5cc6a0c096e8f6834583a6cc81986f1d3df4a6cdb7abc9c8b2b151dd5abf27f8
-size 1636813
+oid sha256:b51210cad429999ffef983930153825fc3b5f68ea9ea6be4334b5e654c783896
+size 1637038

data/prose_100/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8f4b6345d684a84262a9f65dd50428215b763509b1b017582694160687bda1a7
-size 1109747
+oid sha256:32c1eac79ca6762733b1da74a349a6bd4970859542558fb973ea2c9c62ae44c2
+size 1109778

data/prose_100/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:beacea9590dbcdacc1fd8a55ba640766158341638fd37339fb8d046a6dce325a
-size 12940692
+oid sha256:1f8ea005c0094d41ba88ffd835481ab4a1e1ae205b0161d44605245c23fb4329
+size 12950032

data/prose_100/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32f02dc9899f1d4bb816529c7f090b594d965c482a3ded45e619d78dc007fc9b
-size 1110513
+oid sha256:733c0d19d165141d5a007d5d08c81ad8be036d2d5d5cf6fc983da2f18d169e9e
+size 1110554

data/verse_metre_1/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2f24d8b2aaccb611f4153dea07f2b2c8f30f41ebaebe11fc77b99734f1bd1753
-size 1392687
+oid sha256:b94a6506fe839d15401e96622e7750b8ccdf6078d818a3aab6f08f4dbe311894
+size 1392577

data/verse_metre_1/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d20ea516f44169db2849f50111c99d8640e2e942678fb602b2b0e41a3b4e1f8b
-size 6979666
+oid sha256:15d808c053df82e8a1a0e75614a4e38a6b338eeecabd5afd55b22ccc7ee39651
+size 6978445

data/verse_metre_1/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7fb08bfdde1604c8bd7f9525837d2d05bb3877a7783274e26028042ebd0ba6c9
-size 1404667
+oid sha256:f5022ffcac0817824f3462af8cac71a5dd63da8650775e2ad546e92f8933d5fc
+size 1404481

data/verse_metre_10/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0122d069ab69c732c8b45533476bc269625fd815104e5a9e6504ee3ac54d0535
-size 1243839
+oid sha256:89e3bd097f6de102f2c0c804b0024d84b5f5aa604d7eae17ddae98b019d16eb2
+size 1243932

data/verse_metre_10/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:369d05e6a1b5e2b952960e78411d29e09614022da7fd1418845e8fb868fcea2f
-size 7738064
+oid sha256:6ade24c09c35212b177a580ced44b83ef5bec5d2842f0ca5c0d89a638380e3e5
+size 7736840

data/verse_metre_10/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:67c243cc686ffb1b0794e308226b288e2e74cac56be956f3008ada1f0d5b6094
-size 1244941
+oid sha256:1a9a53657bac81f47a1e2c12590e469611c94235db2eaa67fc1f82b3fc973b0c
+size 1244984

data/verse_metre_100/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00405b314ee35c36dc9cd0e89a0a3f55146af0fe0601d980506ec0d2f179240e
-size 1105089
+oid sha256:7d13e026a6880144ac54ff6667d6770af08057ae22eaad23eadaa55ad2991dfc
+size 1105208

data/verse_metre_100/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b0bf6e759d15b0e3e224d65fa5a6dc8f548c4ee4f9c7a4b2b52937b3a7025efd
-size 7397356
+oid sha256:8e1f8dacfca0dd540f898f06f1aef6b893937baf75562257cd7a7e7df35047d1
+size 7396134

data/verse_metre_100/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b27e0fc1f718fe14f59f9b00a2944992b0366daa42020dbecf8f22129489fe4
-size 1111111
+oid sha256:89f568e83a3ccc0fe8e5aece955ea0ea311a51667c4d4482f0c1c47bedd4f2df
+size 1111127

data/verse_sentence_1/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:05bad0ee73fe63da8a8caa476076817ff49b9e25cea2d2f568ffcfe10754fea5
-size 983119
+oid sha256:6d85a321fd454bf40f80042b6ed740f2d563e3f8f491097f31f16a1f4bea247c
+size 983173

data/verse_sentence_1/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f018aa2b7fe9bff0bfead1d6156acc17e02cab838d95e9a56e9fa8eb1cf8c2eb
-size 7021105
+oid sha256:14fb7c585b694c767a45e137a86dff43a2f10f25734d3259924ffd74972de868
+size 7021255

data/verse_sentence_1/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4287b3e483eaa960db4c50d4aa16fa31a35705ae49707c627fc9ae94144e0654
-size 991990
+oid sha256:eba2a593c9be9cfa16275f78ef6681402488e0b1613d49a31a7f6f0204f014cc
+size 992115

data/verse_sentence_10/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16e8011ad7722bff44aa13e27fba66834ddcb00ae2c973d2d95e2c023c57c8a7
-size 888242
+oid sha256:ab1c60bb7134f82080e454944ab70ddb134d6b56b4c2ab0aa9f0d71eed5b9702
+size 888272

data/verse_sentence_10/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:57a2f4b45a1c3e4aec4c4f21b178b78e9b3a28c33e25d6bade7375f6e9b4b71c
-size 7418187
+oid sha256:892c417d5138b16fad888cca918affd2f681546139d825bfb1d447777368e45b
+size 7418314

data/verse_sentence_10/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b099d56c62986de0c3cf0565390b03adb9090655a6d81b9661a63e91684454a
-size 882081
+oid sha256:269355558aeababb8b92c50e4f7d3502321727d88d4cf0d939f192fa78d5b709
+size 881989

data/verse_sentence_100/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9cfc1c393b35a97611550e36cd60b8eca54376037c03141fd5c129e089e63d5
-size 785305
+oid sha256:4acc7d892efefe91e247c9d777aecf6786831bfc3a45e8ee367875c6daff805e
+size 785299

data/verse_sentence_100/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:537f7f9598113746a9bb3da7377d8372c2343d1dabdb53a97fd404c9ace77598
-size 7083932
+oid sha256:618069a03f4d1ab7ea306ff0edb0c593b47ff733da786ddf81b4e127c5dba8e8
+size 7084097

data/verse_sentence_100/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd2f56d635093b396235acbfd4b33dd99b1f7795d26f1c947e7521a5e24cb83b
-size 779917
+oid sha256:4ee94cba3f6996fa6f407d787ebe86f02d1d5259a0bd62e0af3a07e9634d926d
+size 779883

metadata/build_report.json

@@ -940,7 +940,9 @@
   "validation": {
     "checks": {
       "fixed_exact_evaluation_chunks": true,
+      "official_conll18_loader": true,
       "prose_exact_text_unique": true,
+      "unique_conllu_documents_checked": 62713,
       "unique_ids": true,
       "verse_component_exact_coverage": true
     },

scripts/build_dataset.py

@@ -11,6 +11,7 @@ from __future__ import annotations
 import argparse
 import hashlib
 import html
+import io
 import json
 import random
 import re
@@ -29,8 +30,10 @@ import pyarrow.parquet as pq
 
 try:
     from scripts.dataset_variants import make_dataset_variants, variant_schema
+    from scripts.vendor.conll18_ud_eval import UDError, load_conllu
 except ModuleNotFoundError:  # Direct execution from the scripts directory.
     from dataset_variants import make_dataset_variants, variant_schema
+    from vendor.conll18_ud_eval import UDError, load_conllu
 
 
 RANDOM_SEED = 776
@@ -338,7 +341,7 @@ def canonicalize_conllu_xpos(conllu: str) -> str:
             columns[4] = canonical_xpos(columns[3], columns[4], columns[5])
             line = "\t".join(columns)
         lines.append(line)
-    return "\n".join(lines) + "\n"
+    return "\n".join(lines).rstrip("\n") + "\n\n"
 
 
 def canonical_native_deprel(relation: str, child_upos: str, head: int) -> str:
@@ -679,7 +682,7 @@ def xml_sentence_to_conllu(sentence: ET.Element, metadata: dict[str, str]) -> tu
             }),
         ]
         lines.append("\t".join(row))
-    return text, "\n".join(lines) + "\n", cites
+    return text, "\n".join(lines) + "\n\n", cites
 
 
 def parse_agdt(root: Path, revisions: dict[str, str]) -> tuple[list[Sentence], list[Sentence]]:
@@ -1212,7 +1215,7 @@ def align_verse_blocks(
                     "id": line_ids[line_index], "parent_sentence_ids": parent_ids,
                     "author": parents[0].author, "work": parents[0].work, "work_id": parents[0].work_id,
                     "genre": "verse_metre", "text": line["text"],
-                    "conllu": "\n\n".join(parent.conllu.strip() for parent in parents) + "\n",
+                    "conllu": "\n\n".join(parent.conllu.strip() for parent in parents) + "\n\n",
                     "cts_urn": parents[0].cts_urn,
                     "passage": " | ".join(dict.fromkeys(parent.passage for parent in parents if parent.passage)),
                     "alignment_component_id": component_id,
@@ -1387,6 +1390,7 @@ def write_parquet(
 
 def validate(rows_by_config: dict[str, list[dict]]) -> dict:
     report = {}
+    checked_conllu = set()
     for config, rows in rows_by_config.items():
         ids = [row["id"] for row in rows]
         dedup = [row["dedup_key"] for row in rows]
@@ -1406,6 +1410,17 @@ def validate(rows_by_config: dict[str, list[dict]]) -> dict:
             assert all(len(row["constituent_ids"]) == row["chunk_size"] for row in rows)
         if base_config == "prose":
             assert len(dedup) == len(set(dedup)), "prose exact-text deduplication failed"
+        for row in rows:
+            digest = hashlib.sha256(row["conllu"].encode("utf-8")).digest()
+            if digest in checked_conllu:
+                continue
+            try:
+                load_conllu(io.StringIO(row["conllu"]))
+            except UDError as error:
+                raise AssertionError(
+                    f"Malformed CoNLL-U in {config}, row {row['id']}: {error}"
+                ) from error
+            checked_conllu.add(digest)
         report[config] = {
             "rows": len(rows),
             "authors": len({row["author"] for row in rows}),
@@ -1429,6 +1444,8 @@ def validate(rows_by_config: dict[str, list[dict]]) -> dict:
         "unique_ids": True, "prose_exact_text_unique": True,
         "verse_component_exact_coverage": True,
         "fixed_exact_evaluation_chunks": True,
+        "official_conll18_loader": True,
+        "unique_conllu_documents_checked": len(checked_conllu),
     }
     return report
 

scripts/dataset_variants.py

@@ -154,7 +154,7 @@ def _aggregate_chunk(base_config: str, rows: list[dict], target: int, split: str
         "text": ("\n" if base_config == "verse_metre" else "\n\n").join(
             row["text"].strip() for row in rows
         ),
-        "conllu": "\n\n".join(row["conllu"].strip() for row in rows) + "\n",
+        "conllu": "\n\n".join(row["conllu"].strip() for row in rows) + "\n\n",
         "cts_urn": rows[0]["cts_urn"] if len({row["cts_urn"] for row in rows}) == 1 else None,
         "passage": (
             rows[0]["passage"]

scripts/validate_dataset.py

@@ -4,6 +4,8 @@
 from __future__ import annotations
 
 import argparse
+import hashlib
+import io
 import json
 import re
 import unicodedata
@@ -12,6 +14,11 @@ from pathlib import Path
 
 import pyarrow.parquet as pq
 
+try:
+    from scripts.vendor.conll18_ud_eval import UDError, load_conllu
+except ModuleNotFoundError:  # Direct execution from the scripts directory.
+    from vendor.conll18_ud_eval import UDError, load_conllu
+
 try:
     from scripts.dataset_variants import (
         actual_order_key,
@@ -381,12 +388,38 @@ def validate_combined_conllu(root: Path, parent_conllu: dict[str, str]) -> int:
         parquet = pq.ParquetFile(path)
         for batch in parquet.iter_batches(columns=["parent_sentence_ids", "conllu"], batch_size=256, use_threads=False):
             for row in batch.to_pylist():
-                expected = "\n\n".join(parent_conllu[parent_id].strip() for parent_id in row["parent_sentence_ids"]) + "\n"
+                expected = "\n\n".join(
+                    parent_conllu[parent_id].strip()
+                    for parent_id in row["parent_sentence_ids"]
+                ) + "\n\n"
                 assert row["conllu"] == expected
                 checked += 1
     return checked
 
 
+def validate_conllu_with_official_loader(root: Path) -> int:
+    checked = set()
+    for config in CONFIGS:
+        for split in SPLITS:
+            path = root / config / f"{split}-00000-of-00001.parquet"
+            parquet = pq.ParquetFile(path)
+            for batch in parquet.iter_batches(
+                columns=["id", "conllu"], batch_size=256, use_threads=False,
+            ):
+                for row in batch.to_pylist():
+                    digest = hashlib.sha256(row["conllu"].encode("utf-8")).digest()
+                    if digest in checked:
+                        continue
+                    try:
+                        load_conllu(io.StringIO(row["conllu"]))
+                    except UDError as error:
+                        raise AssertionError(
+                            f"Malformed CoNLL-U in {config}/{split}, row {row['id']}: {error}"
+                        ) from error
+                    checked.add(digest)
+    return len(checked)
+
+
 def main() -> None:
     parser = argparse.ArgumentParser()
     parser.add_argument("--data", type=Path, default=Path("data"))
@@ -410,10 +443,12 @@ def main() -> None:
     checked_lines = validate_provenance_and_verse(args.data)
     checked, parent_conllu = validate_conllu(args.data)
     checked_combined = validate_combined_conllu(args.data, parent_conllu)
+    checked_official = validate_conllu_with_official_loader(args.data)
     print("validated aligned metrical lines", checked_lines)
     print("validated cross-sentence metrical lines", crossing_lines)
     print("validated CoNLL-U sentences", checked)
     print("validated combined metrical CoNLL-U documents", checked_combined)
+    print("validated unique CoNLL-U documents with official CoNLL 2018 loader", checked_official)
     print("all checks passed")
 
 

scripts/vendor/conll18_ud_eval.py

@@ -0,0 +1,203 @@
+#!/usr/bin/env python
+
+# Compatible with Python 2.7 and 3.2+, can be used either as a module
+# or a standalone executable.
+#
+# Copyright 2017, 2018 Institute of Formal and Applied Linguistics (UFAL),
+# Faculty of Mathematics and Physics, Charles University, Czech Republic.
+#
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this
+# file, You can obtain one at http://mozilla.org/MPL/2.0/.
+#
+# Authors: Milan Straka, Martin Popel <surname@ufal.mff.cuni.cz>
+#
+# This file extracts the load_conllu portion of conll18_ud_eval.py from
+# https://github.com/CoNLL-UD-2018/UDPipe-Future (version 1.1), with formatting
+# changes only and without the unrelated scoring and command-line code.
+
+from __future__ import division
+from __future__ import print_function
+
+import sys
+import unicodedata
+
+
+# CoNLL-U column names
+ID, FORM, LEMMA, UPOS, XPOS, FEATS, HEAD, DEPREL, DEPS, MISC = range(10)
+
+# Content and functional relations
+CONTENT_DEPRELS = {
+    "nsubj", "obj", "iobj", "csubj", "ccomp", "xcomp", "obl", "vocative",
+    "expl", "dislocated", "advcl", "advmod", "discourse", "nmod", "appos",
+    "nummod", "acl", "amod", "conj", "fixed", "flat", "compound", "list",
+    "parataxis", "orphan", "goeswith", "reparandum", "root", "dep"
+}
+
+FUNCTIONAL_DEPRELS = {
+    "aux", "cop", "mark", "det", "clf", "case", "cc"
+}
+
+UNIVERSAL_FEATURES = {
+    "PronType", "NumType", "Poss", "Reflex", "Foreign", "Abbr", "Gender",
+    "Animacy", "Number", "Case", "Definite", "Degree", "VerbForm", "Mood",
+    "Tense", "Aspect", "Voice", "Evident", "Polarity", "Person", "Polite"
+}
+
+
+# UD Error is used when raising exceptions in this module
+class UDError(Exception):
+    pass
+
+
+# Load given CoNLL-U file into internal representation
+def load_conllu(file):
+    # Internal representation classes
+    class UDRepresentation:
+        def __init__(self):
+            self.characters = []
+            self.tokens = []
+            self.words = []
+            self.sentences = []
+
+    class UDSpan:
+        def __init__(self, start, end):
+            self.start = start
+            self.end = end
+
+    class UDWord:
+        def __init__(self, span, columns, is_multiword):
+            self.span = span
+            self.columns = columns
+            self.is_multiword = is_multiword
+            self.parent = None
+            self.functional_children = []
+            self.columns[FEATS] = "|".join(sorted(
+                feat for feat in columns[FEATS].split("|")
+                if feat.split("=", 1)[0] in UNIVERSAL_FEATURES
+            ))
+            self.columns[DEPREL] = columns[DEPREL].split(":")[0]
+            self.is_content_deprel = self.columns[DEPREL] in CONTENT_DEPRELS
+            self.is_functional_deprel = self.columns[DEPREL] in FUNCTIONAL_DEPRELS
+
+    ud = UDRepresentation()
+
+    # Load the CoNLL-U file
+    index, sentence_start = 0, None
+    while True:
+        line = file.readline()
+        if not line:
+            break
+        line = line.rstrip("\r\n")
+
+        # Handle sentence start boundaries
+        if sentence_start is None:
+            # Skip comments
+            if line.startswith("#"):
+                continue
+            # Start a new sentence
+            ud.sentences.append(UDSpan(index, 0))
+            sentence_start = len(ud.words)
+        if not line:
+            # Add parent and children UDWord links and check there are no cycles
+            def process_word(word):
+                if word.parent == "remapping":
+                    raise UDError("There is a cycle in a sentence")
+                if word.parent is None:
+                    head = int(word.columns[HEAD])
+                    if head < 0 or head > len(ud.words) - sentence_start:
+                        raise UDError("HEAD '{}' points outside of the sentence".format(word.columns[HEAD]))
+                    if head:
+                        parent = ud.words[sentence_start + head - 1]
+                        word.parent = "remapping"
+                        process_word(parent)
+                        word.parent = parent
+
+            for word in ud.words[sentence_start:]:
+                process_word(word)
+            # func_children cannot be assigned within process_word because it is
+            # called recursively and may result in adding one child twice.
+            for word in ud.words[sentence_start:]:
+                if word.parent and word.is_functional_deprel:
+                    word.parent.functional_children.append(word)
+
+            # Check there is a single root node
+            if len([word for word in ud.words[sentence_start:] if word.parent is None]) != 1:
+                raise UDError("There are multiple roots in a sentence")
+
+            # End the sentence
+            ud.sentences[-1].end = index
+            sentence_start = None
+            continue
+
+        # Read next token/word
+        columns = line.split("\t")
+        if len(columns) != 10:
+            raise UDError(
+                "The CoNLL-U line does not contain 10 tab-separated columns: '{}'".format(line)
+            )
+
+        # Skip empty nodes
+        if "." in columns[ID]:
+            continue
+
+        # Delete spaces from FORM, so gold.characters == system.characters even
+        # if one of them tokenizes the space. Use Unicode category Zs.
+        if sys.version_info < (3, 0) and isinstance(line, str):
+            columns[FORM] = columns[FORM].decode("utf-8")
+        columns[FORM] = "".join(filter(
+            lambda c: unicodedata.category(c) != "Zs", columns[FORM]
+        ))
+        if sys.version_info < (3, 0) and isinstance(line, str):
+            columns[FORM] = columns[FORM].encode("utf-8")
+        if not columns[FORM]:
+            raise UDError("There is an empty FORM in the CoNLL-U file")
+
+        # Save token
+        ud.characters.extend(columns[FORM])
+        ud.tokens.append(UDSpan(index, index + len(columns[FORM])))
+        index += len(columns[FORM])
+
+        # Handle multi-word tokens to save word(s)
+        if "-" in columns[ID]:
+            try:
+                start, end = map(int, columns[ID].split("-"))
+            except:
+                raise UDError("Cannot parse multi-word token ID '{}'".format(columns[ID]))
+
+            for _ in range(start, end + 1):
+                word_line = file.readline().rstrip("\r\n")
+                word_columns = word_line.split("\t")
+                if len(word_columns) != 10:
+                    raise UDError(
+                        "The CoNLL-U line does not contain 10 tab-separated columns: '{}'".format(
+                            word_line
+                        )
+                    )
+                ud.words.append(UDWord(ud.tokens[-1], word_columns, is_multiword=True))
+        # Basic tokens/words
+        else:
+            try:
+                word_id = int(columns[ID])
+            except:
+                raise UDError("Cannot parse word ID '{}'".format(columns[ID]))
+            if word_id != len(ud.words) - sentence_start + 1:
+                raise UDError(
+                    "Incorrect word ID '{}' for word '{}', expected '{}'".format(
+                        columns[ID], columns[FORM], len(ud.words) - sentence_start + 1
+                    )
+                )
+
+            try:
+                head_id = int(columns[HEAD])
+            except:
+                raise UDError("Cannot parse HEAD '{}'".format(columns[HEAD]))
+            if head_id < 0:
+                raise UDError("HEAD cannot be negative")
+
+            ud.words.append(UDWord(ud.tokens[-1], columns, is_multiword=False))
+
+    if sentence_start is not None:
+        raise UDError("The CoNLL-U file does not end with empty line")
+
+    return ud

tests/test_conllu_well_formed.py

@@ -0,0 +1,40 @@
+import hashlib
+import io
+from pathlib import Path
+
+import pyarrow.parquet as pq
+
+from scripts.vendor.conll18_ud_eval import UDError, load_conllu
+
+
+DATA_ROOT = Path(__file__).resolve().parents[1] / "data"
+CONFIGS = tuple(
+    f"{base}_{suffix}"
+    for base in ("prose", "verse_sentence", "verse_metre")
+    for suffix in ("1", "10", "100")
+)
+
+
+def test_all_conllu_documents_load_with_official_conll18_parser() -> None:
+    """Require every distinct published treebank document to parse as CoNLL-U."""
+    checked = set()
+    for config in CONFIGS:
+        for split in ("train", "validation", "test"):
+            path = DATA_ROOT / config / f"{split}-00000-of-00001.parquet"
+            parquet = pq.ParquetFile(path)
+            for batch in parquet.iter_batches(
+                columns=["id", "conllu"], batch_size=256, use_threads=False,
+            ):
+                for row in batch.to_pylist():
+                    digest = hashlib.sha256(row["conllu"].encode("utf-8")).digest()
+                    if digest in checked:
+                        continue
+                    try:
+                        load_conllu(io.StringIO(row["conllu"]))
+                    except UDError as error:
+                        raise AssertionError(
+                            f"Malformed CoNLL-U in {config}/{split}, row {row['id']}: {error}"
+                        ) from error
+                    checked.add(digest)
+
+    assert checked