dsets/tfidf_stats.py

""" 
Parts of the code is based on source code of memit

MIT License

Copyright (c) 2022 Kevin Meng

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.
"""

import json
from itertools import chain
from pathlib import Path

import numpy as np
import scipy.sparse as sp
import torch
from sklearn.feature_extraction.text import TfidfVectorizer

from dsets import AttributeSnippets

REMOTE_ROOT_URL = "https://rome.baulab.info"
REMOTE_IDF_URL = f"{REMOTE_ROOT_URL}/data/dsets/idf.npy"
REMOTE_VOCAB_URL = f"{REMOTE_ROOT_URL}/data/dsets/tfidf_vocab.json"


def get_tfidf_vectorizer(data_dir: str):
    """
    Returns an sklearn TF-IDF vectorizer. See their website for docs.
    Loading hack inspired by some online blog post lol.
    """

    data_dir = Path(data_dir)

    idf_loc, vocab_loc = data_dir / "idf.npy", data_dir / "tfidf_vocab.json"
    if not (idf_loc.exists() and vocab_loc.exists()):
        collect_stats(data_dir)

    idf = np.load(idf_loc)
    with open(vocab_loc, "r") as f:
        vocab = json.load(f)

    class MyVectorizer(TfidfVectorizer):
        TfidfVectorizer.idf_ = idf

    vec = MyVectorizer()
    vec.vocabulary_ = vocab
    vec._tfidf._idf_diag = sp.spdiags(idf, diags=0, m=len(idf), n=len(idf))

    return vec


def collect_stats(data_dir: str):
    """
    Uses wikipedia snippets to collect statistics over a corpus of English text.
    Retrieved later when computing TF-IDF vectors.
    """

    data_dir = Path(data_dir)
    data_dir.mkdir(exist_ok=True, parents=True)
    idf_loc, vocab_loc = data_dir / "idf.npy", data_dir / "tfidf_vocab.json"

    try:
        print(f"Downloading IDF cache from {REMOTE_IDF_URL}")
        torch.hub.download_url_to_file(REMOTE_IDF_URL, idf_loc)
        print(f"Downloading TF-IDF vocab cache from {REMOTE_VOCAB_URL}")
        torch.hub.download_url_to_file(REMOTE_VOCAB_URL, vocab_loc)
        return
    except Exception as e:
        print(f"Error downloading file:", e)
        print("Recomputing TF-IDF stats...")

    snips_list = AttributeSnippets(data_dir).snippets_list
    documents = list(chain(*[[y["text"] for y in x["samples"]] for x in snips_list]))

    vec = TfidfVectorizer()
    vec.fit(documents)

    idfs = vec.idf_
    vocab = vec.vocabulary_

    np.save(data_dir / "idf.npy", idfs)
    with open(data_dir / "tfidf_vocab.json", "w") as f:
        json.dump(vocab, f, indent=1)