Implement simple workflow for parsing spaCy files

.gitignore

@@ -0,0 +1,4 @@
+assets
+corpus/spacy
+__pycache__/
+project.lock

README.md

@@ -0,0 +1,43 @@
+<!-- SPACY PROJECT: AUTO-GENERATED DOCS START (do not remove) -->
+
+# 🪐 spaCy Project: Dataset builder to HuggingFace Hub
+
+This project contains utility scripts for uploading a dataset to HuggingFace
+Hub.  We want to separate the spaCy dependencies from the loading script, so
+we're parsing the spaCy files independently.
+
+The process goes like this: we download the raw corpus from Google Cloud
+Storage (GCS), convert the spaCy files into a readable IOB format, and parse
+that using our loading script (i.e., `tlunified-ner.py`).
+
+We're also shipping the IOB file so that it's easier to access.
+
+
+## 📋 project.yml
+
+The [`project.yml`](project.yml) defines the data assets required by the
+project, as well as the available commands and workflows. For details, see the
+[spaCy projects documentation](https://spacy.io/usage/projects).
+
+### ⏯ Commands
+
+The following commands are defined by the project. They
+can be executed using [`spacy project run [name]`](https://spacy.io/api/cli#project-run).
+Commands are only re-run if their inputs have changed.
+
+| Command | Description |
+| --- | --- |
+| `setup-data` | Prepare the Tagalog corpora used for training various spaCy components |
+| `upload-to-hf` | Upload dataset to HuggingFace Hub |
+
+### 🗂 Assets
+
+The following assets are defined by the project. They can
+be fetched by running [`spacy project assets`](https://spacy.io/api/cli#project-assets)
+in the project directory.
+
+| File | Source | Description |
+| --- | --- | --- |
+| `assets/corpus.tar.gz` | URL | Annotated TLUnified corpora in spaCy format with train, dev, and test splits. |
+
+<!-- SPACY PROJECT: AUTO-GENERATED DOCS END (do not remove) -->

project.yml

@@ -0,0 +1,42 @@
+title: "Dataset builder to HuggingFace Hub"
+description: |
+  This project contains utility scripts for uploading a dataset to HuggingFace
+  Hub.  We want to separate the spaCy dependencies from the loading script, so
+  we're parsing the spaCy files independently.
+
+  The process goes like this: we download the raw corpus from Google Cloud
+  Storage (GCS), convert the spaCy files into a readable IOB format, and parse
+  that using our loading script (i.e., `tlunified-ner.py`).
+
+  We're also shipping the IOB file so that it's easier to access.
+
+directories: ["assets", "corpus/spacy", "corpus/iob"]
+
+vars:
+  version: 1.0
+
+assets:
+  - dest: assets/corpus.tar.gz
+    description: "Annotated TLUnified corpora in spaCy format with train, dev, and test splits."
+    url: "https://storage.googleapis.com/ljvmiranda/calamanCy/tl_tlunified_gold/v${vars.version}/corpus.tar.gz"
+
+commands:
+  - name: "setup-data"
+    help: "Prepare the Tagalog corpora used for training various spaCy components"
+    script:
+      - mkdir -p corpus/spacy
+      - tar -xzvf assets/corpus.tar.gz -C corpus/spacy
+      - python -m spacy_to_iob corpus/spacy/ corpus/iob/
+    outputs:
+      - corpus/iob/train.iob
+      - corpus/iob/dev.iob
+      - corpus/iob/test.iob
+
+  - name: "upload-to-hf"
+    help: "Upload dataset to HuggingFace Hub"
+    script:
+      - ls
+    deps:
+      - corpus/iob/train.iob
+      - corpus/iob/dev.iob
+      - corpus/iob/test.iob

requirements.txt

@@ -0,0 +1,5 @@
+spacy
+typer
+datasets
+huggingface_hub
+wasabi

spacy_to_iob.py

@@ -0,0 +1,50 @@
+from pathlib import Path
+
+import spacy
+import typer
+from spacy.tokens import DocBin
+from wasabi import msg
+
+DELIMITER = "-DOCSTART- -X- O O"
+
+
+def spacy_to_iob(
+    # fmt: off
+    spacy_indir: Path = typer.Argument(..., help="Path to the directory containing the spaCy files."),
+    iob_outdir: Path = typer.Argument(..., help="Path to the directory to save the IOB files."),
+    lang: str = typer.Option("tl", "-l", "--lang", help="Language code for the spaCy vocab."),
+    verbose: bool = typer.Option(False, "-v", "--verbose", help="Print additional information."),
+    delimiter: str = typer.Option(DELIMITER, "-d", "--delimiter", help="Delimiter between examples.")
+    # fmt: on
+):
+    """Convert spaCy files into IOB-formatted files."""
+    nlp = spacy.blank(lang)
+    for spacy_file in spacy_indir.glob(f"*.spacy"):
+        msg.text(f"Converting {str(spacy_file)}", show=verbose)
+        doc_bin = DocBin().from_disk(spacy_file)
+        docs = doc_bin.get_docs(nlp.vocab)
+
+        lines = []  # container for the IOB lines later on
+        for doc in docs:
+            lines.append(delimiter)
+            lines.append("\n\n")
+            for token in doc:
+                label = (
+                    f"{token.ent_iob_}-{token.ent_type_}"
+                    if token.ent_iob_ != "O"
+                    else "O"
+                )
+                line = f"{token.text}\t{label}"
+                lines.append(line)
+                lines.append("\n")
+            lines.append("\n")
+
+        iob_file = iob_outdir / f"{spacy_file.stem}.iob"
+        with open(iob_file, "w", encoding="utf-8") as f:
+            f.writelines(lines)
+
+        msg.good(f"Saved to {iob_file}")
+
+
+if __name__ == "__main__":
+    typer.run(spacy_to_iob)

tlunified-ner.py

@@ -0,0 +1,89 @@
+import os
+from typing import List
+
+import datasets
+
+logger = datasets.logging.get_logger(__name__)
+
+_DESCRIPTION = """"""
+_LICENSE = """GNU GPL v3.0"""
+_CLASSES = ["O", "B-PER", "I-PER", "B-ORG", "I-ORG", "B-LOC", "I-LOC"]
+_VERSION = "1.0"
+
+
+class TLUnifiedNERConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super(TLUnifiedNER, self).__init__(**kwargs)
+
+
+class TLUnifiedNER(datasets.GeneratorBasedBuilder):
+    """Contains an annotated version of the TLUnified dataset from Cruz and Cheng (2021)."""
+
+    VERSION = datasets.Version(_VERSION)
+
+    def _info() -> "datasets.DatasetInfo":
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "id": datasets.Value("string"),
+                    "tokens": datasets.Sequence(datasets.Value("string")),
+                    "ner_tags": datasets.Sequence(
+                        datasets.feature.ClassLabel(names=_CLASSES)
+                    ),
+                }
+            ),
+            supervised_keys=None,
+        )
+
+    def _split_generators(
+        self, dl_manager: "datasets.builder.DownloadManager"
+    ) -> List["datasets.SplitGenerator"]:
+        """Return a list of SplitGenerators that organizes the splits."""
+        # The file extracts into {train,dev,test}.spacy files. The _generate_examples function
+        # below will define how these files are parsed.
+        corpus_dir = "corpus/iob"
+        data_files = {
+            "train": os.path.join(corpus_dir, "train.iob"),
+            "dev": os.path.join(corpus_dir, "dev.iob"),
+            "test": os.path.join(corpus_dir, "test.iob"),
+        }
+
+        return [
+            # fmt: off
+            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": data_files["train"]}),
+            datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": data_files["dev"]}),
+            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": data_files["test"]}),
+            # fmt: on
+        ]
+
+    def _generate_examples(self, filepath: str):
+        """Defines how examples are parsed from the IOB file."""
+        logger.info("⏳ Generating examples from = %s", filepath)
+        with open(filepath, encoding="utf-8") as f:
+            guid = 0
+            tokens = []
+            ner_tags = []
+            for line in f:
+                if line.startswith("-DOCSTART-") or line == "" or line == "\n":
+                    if tokens:
+                        yield guid, {
+                            "id": str(guid),
+                            "tokens": tokens,
+                            "ner_tags": ner_tags,
+                        }
+                        guid += 1
+                        tokens = []
+                        ner_tags = []
+                else:
+                    # TLUnified-NER iob are separated by \t
+                    token, ner_tag = line.split("\t")
+                    tokens.append(token)
+                    ner_tags.append(ner_tag.rstrip())
+            # Last example
+            if tokens:
+                yield guid, {
+                    "id": str(guid),
+                    "tokens": tokens,
+                    "ner_tags": ner_tags,
+                }