Converting Doccano NER task export (JSONL file) to a CONLL03-formatted file

doccano2conll2003.py

# coding: utf-8

import re
from builtins import frozenset

from typing import List, Tuple, Set


def find_occurrences(s, c):
    return [i for i, letter in enumerate(s) if letter == c]


def annotation2sequences(original_text: str, label_spans: List[List], excluded_labels: Set = frozenset([]),
                         already_tokenized: bool = True, tokenize_fn=None):
    original_text = original_text.rstrip()

    # building tokens' spans
    if already_tokenized:
        token_spans = [[0]]

        for m in re.finditer(" +", original_text):
            token_spans[-1].append(m.start())
            token_spans.append([m.end()])

        token_spans[-1].append(len(original_text))
    else:
        token_spans = tokenize_fn(original_text)

    # splitting into sentences
    newlines_positions = find_occurrences(original_text, "\n") + [len(original_text)]
    sentence_boundaries = [0]

    for token_number, (l, r) in enumerate(token_spans):
        if r == newlines_positions[0]:
            sentence_boundaries.append(token_number + 1)
            newlines_positions = newlines_positions[1:]
        elif r > newlines_positions[0]:
            raise Exception

    # print(sentence_boundaries)

    tokens = [original_text[fr: to] for fr, to in token_spans]
    labels = ["O"] * len(token_spans)
    curr_token_idx = 0

    # everything that intersects -- gets tagged as the corresponding label
    for span in label_spans:

        prefix = "B-"
        label_from, label_to, tag = span

        while token_spans[curr_token_idx][1] <= label_from:
            curr_token_idx += 1

        while token_spans[curr_token_idx][0] < label_to:
            if not tag in excluded_labels:
                labels[curr_token_idx] = prefix + tag
                prefix = "I-"
            curr_token_idx += 1

    assert len(tokens) == len(labels)

    results = []

    for i in range(len(sentence_boundaries) - 1):
        results.append((i + 1, list(zip(
            tokens[sentence_boundaries[i]:sentence_boundaries[i + 1]],
            labels[sentence_boundaries[i]:sentence_boundaries[i + 1]]))))

    return results


def seq2conll2003(list_of_lists_of_pairs: List[List[Tuple[str, str]]]):
    result = ""
    sentence_number = 1
    for docid, sentences in list_of_lists_of_pairs:
        for sentid, pairs in sentences:
            result += f"# {sentence_number} {docid} {sentid}\n"
            result += "\n".join([f"{t}\t-\t-\t{l}" for t, l in pairs])
            result += "\n\n"
            sentence_number += 1
    return result.strip()


if __name__ == "__main__":

    import json
    from apertium_tokenizer import ApertiumSimpleTokenizer

    tokenizer = ApertiumSimpleTokenizer()
    custom_tokenize = lambda sentence: list(tokenizer.span_tokenize(sentence))
    all_results_with_custom_tokenization = []

    for line_idx, line in enumerate(open("Anton.jsonl", "r+", encoding="utf-8"), 1):
        if line_idx > 3:
            break

        ann_data = json.loads(line.strip())
        idx, orig_text, label_spans = ann_data["id"], ann_data["text"], ann_data["label"]

        annotated_enumerated_sentences = annotation2sequences(orig_text, label_spans,
                                                              already_tokenized=False,
                                                              excluded_labels={"PERIOD", "MEASURE",
                                                                               "IDENTIFIER", "ACRONYM"},
                                                              tokenize_fn=custom_tokenize)
        all_results_with_custom_tokenization.append((line_idx, annotated_enumerated_sentences))

    with open("sample.conll2003-formatted.txt", "w+", encoding="utf-8") as wf:
        wf.write(seq2conll2003(all_results_with_custom_tokenization))

LICENSE

MIT License

Copyright (c) 2023 Anton Alekseev

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.