mirror of
https://github.com/hwchase17/langchain.git
synced 2025-10-31 16:08:59 +00:00
3152d25811
Co-authored-by: cbornet <cbornet@hotmail.com> Co-authored-by: Mason Daugherty <mason@langchain.dev>
68 lines
2.2 KiB
Python
68 lines
2.2 KiB
Python
"""Spacy text splitter."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import Any
|
|
|
|
from langchain_text_splitters.base import TextSplitter
|
|
|
|
try:
|
|
# Type ignores needed as long as spacy doesn't support Python 3.14.
|
|
import spacy # type: ignore[import-not-found, unused-ignore]
|
|
from spacy.lang.en import English # type: ignore[import-not-found, unused-ignore]
|
|
from spacy.language import Language # type: ignore[import-not-found, unused-ignore]
|
|
|
|
_HAS_SPACY = True
|
|
except ImportError:
|
|
_HAS_SPACY = False
|
|
|
|
|
|
class SpacyTextSplitter(TextSplitter):
|
|
"""Splitting text using Spacy package.
|
|
|
|
Per default, Spacy's `en_core_web_sm` model is used and
|
|
its default max_length is 1000000 (it is the length of maximum character
|
|
this model takes which can be increased for large files). For a faster, but
|
|
potentially less accurate splitting, you can use `pipeline='sentencizer'`.
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
separator: str = "\n\n",
|
|
pipeline: str = "en_core_web_sm",
|
|
max_length: int = 1_000_000,
|
|
*,
|
|
strip_whitespace: bool = True,
|
|
**kwargs: Any,
|
|
) -> None:
|
|
"""Initialize the spacy text splitter."""
|
|
super().__init__(**kwargs)
|
|
self._tokenizer = _make_spacy_pipeline_for_splitting(
|
|
pipeline, max_length=max_length
|
|
)
|
|
self._separator = separator
|
|
self._strip_whitespace = strip_whitespace
|
|
|
|
def split_text(self, text: str) -> list[str]:
|
|
"""Split incoming text and return chunks."""
|
|
splits = (
|
|
s.text if self._strip_whitespace else s.text_with_ws
|
|
for s in self._tokenizer(text).sents
|
|
)
|
|
return self._merge_splits(splits, self._separator)
|
|
|
|
|
|
def _make_spacy_pipeline_for_splitting(
|
|
pipeline: str, *, max_length: int = 1_000_000
|
|
) -> Language:
|
|
if not _HAS_SPACY:
|
|
msg = "Spacy is not installed, please install it with `pip install spacy`."
|
|
raise ImportError(msg)
|
|
if pipeline == "sentencizer":
|
|
sentencizer: Language = English()
|
|
sentencizer.add_pipe("sentencizer")
|
|
else:
|
|
sentencizer = spacy.load(pipeline, exclude=["ner", "tagger"])
|
|
sentencizer.max_length = max_length
|
|
return sentencizer
|