pipeline.py

from typing import Dict, List, Any
from nemo.collections.nlp.models import PunctuationCapitalizationModel


class PreTrainedPipeline:
    def __init__(self, path=""):
        # IMPLEMENT_THIS
        # Preload all the elements you are going to need at inference.
        # For instance your model, processors, tokenizer that might be needed.
        # This function is only called once, so do all the heavy processing I/O here"""
        self.model = PunctuationCapitalizationModel.from_pretrained("dchaplinsky/punctuation_uk_bert")

    def __call__(self, inputs: str) -> List[Dict[str, Any]]:
        """
        Args:
            inputs (:obj:`str`):
                a string containing some text
        Return:
            A :obj:`list`:. The object returned should be like [{"entity_group": "XXX", "word": "some word", "start": 3, "end": 6, "score": 0.82}] containing :
                - "entity_group": A string representing what the entity is.
                - "word": A substring of the original string that was detected as an entity.
                - "start": the offset within `input` leading to `answer`. context[start:stop] == word
                - "end": the ending offset within `input` leading to `answer`. context[start:stop] === word
                - "score": A score between 0 and 1 describing how confident the model is for this entity.
        """
        inputs = inputs.strip()
        labels = self.model.add_punctuation_capitalization([inputs], return_labels=True)[0].split()

        tokens = inputs.split()

        res: List[Dict[str, Any]] = []
        offset = 0
        for tok, lab in zip(tokens, labels):
            if lab != "OO":
                res.append({"entity_group": lab, "word": tok, "start": offset, "end": offset + len(tok), "score": 0.99})

            offset += len(tok) + 1

        return res