wiktor_ner.py

"""Module for parsing WiktorNER files."""

import json
from typing import List, Tuple

from src.annotations import Annotation, MorphosyntacticAnnotation, NerAnnotation
from src.input_parsers.interface import InputParser


class WiktorNERInputParser(InputParser):
    """Parser for WiktorNER files.

    Example WiktorNER file:
    {
        "filename": "greeting-5b1401",
        "text": "Hello Tom!",
        "tokens": [
            {
                "index": 1,
                "position": [0,5],
                "orth": "Hello",
                "lex": [
                    {
                        "lemma": "hello",
                        "mstag": "interj"
                    }
                ]
            },
            {
                "index": 2,
                "position": [6,9],
                "orth": "Tom",
                "lex": [
                    {
                        "lemma": "Tom",
                        "mstag": "noun"
                    }
                ]
            },
            {
                "index": 3,
                "position": [9,10],
                "orth": "!",
                "lex": [
                    {
                        "lemma": "!",
                        "mstag": "interp"
                    }
                ]
            }
        ],
        "entities": [
            {
                "text": "Tom",
                "type": "nam_prs_human",
                "tokens": [2],
                "positions": [6,9]
            }
        ]
    }
    """

    def parse(self, content: str) -> Tuple[str, List[Tuple[int, int, Annotation]]]:
        """Parse wiktorner file into text and annotations.

        Annotations are returned as a dictionary with channel name as a key and list of
        tuples.

        Args:
            co z  (str): Path to file containing CCL.

        Returns:
            Tuple[str, List[Tuple[int, int, Annotation]]]: Text and annotations.

        """
        content_parsed = json.loads(content)

        if "text" in content_parsed:
            text = content_parsed["text"]
        else:
            text = ""

        annotations = []

        # Morphosyntactic annotations
        if "tokens" in content_parsed:
            for token in content_parsed["tokens"]:
                if "position" in token:
                    token_start, token_end = token["position"]
                    if "lexemes" in token:
                        for lexeme in token["lexemes"]:
                            if "disamb" in lexeme and lexeme["disamb"] is True:
                                if "mstag" in lexeme:
                                    lemma = lexeme.get("lemma", None)
                                    annotations.append(
                                        (
                                            token_start,
                                            token_end,
                                            MorphosyntacticAnnotation(
                                                lexeme["mstag"], lemma
                                            ),
                                        )
                                    )

        # NER annotations
        if "entities" in content_parsed:
            for entity in content_parsed["entities"]:
                if "positions" in entity:
                    entity_start, entity_end = entity["positions"]
                    if "type" in entity:
                        annotations.append(
                            (entity_start, entity_end, NerAnnotation(entity["type"]))
                        )

        return text, annotations