Edit model card
YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

This model can be used for sentence compression (aka extractive sentence summarization).

It predicts for each word, whether the word can be dropped from the sentence without severely affecting its meaning.

The resulting sentences are often ungrammatical, but they still can be useful.

The model is rubert-tiny2 fine-tuned on the dataset from the paper Sentence compression for Russian: dataset and baselines (the data can be found here).

Example usage:

import torch
from transformers import AutoModelForTokenClassification, AutoTokenizer
model_name = 'cointegrated/rubert-tiny2-sentence-compression'
model = AutoModelForTokenClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)


def compress(text, threshold=0.5, keep_ratio=None):
    """ Compress a sentence by removing the least important words.
    Parameters:
        threshold: cutoff for predicted probabilities of word removal
        keep_ratio: proportion of words to preserve
    By default, threshold of 0.5 is used.
    """
    with torch.inference_mode():
        tok = tokenizer(text, return_tensors='pt').to(model.device)
        proba = torch.softmax(model(**tok).logits, -1).cpu().numpy()[0, :, 1]
    if keep_ratio is not None:
        threshold = sorted(proba)[int(len(proba) * keep_ratio)]
    kept_toks = []
    keep = False
    prev_word_id = None
    for word_id, score, token in zip(tok.word_ids(), proba, tok.input_ids[0]):
        if word_id is None:
            keep = True
        elif word_id != prev_word_id:
            keep = score < threshold
        if keep:
            kept_toks.append(token)
        prev_word_id = word_id
    return tokenizer.decode(kept_toks, skip_special_tokens=True)


text = 'Кроме того, можно взять идею, рожденную из сердца, и выразить ее в рамках одной '\
    'из этих структур, без потери искренности идеи и смысла песни.'
    
print(compress(text))
print(compress(text, threshold=0.3))
print(compress(text, threshold=0.1))
# можно взять идею, рожденную из сердца, и выразить ее в рамках одной из этих структур.
# можно взять идею, рожденную из сердца выразить ее в рамках одной из этих структур.
# можно взять идею рожденную выразить структур.

print(compress(text, keep_ratio=0.5))
# можно взять идею, рожденную из сердца выразить ее в рамках структур.
Downloads last month
89
Safetensors
Model size
29.1M params
Tensor type
I64
·
F32
·