langchain_experimental.text_splitter.SemanticChunker

class langchain_experimental.text_splitter.SemanticChunker(embeddings: Embeddings, buffer_size: int = 1, add_start_index: bool = False, breakpoint_threshold_type: Literal['percentile', 'standard_deviation', 'interquartile'] = 'percentile', breakpoint_threshold_amount: Optional[float] = None, number_of_chunks: Optional[int] = None, sentence_split_regex: str = '(?<=[.?!])\\s+')

Split the text based on semantic similarity.

Taken from Greg Kamradt’s wonderful notebook: https://github.com/FullStackRetrieval-com/RetrievalTutorials/blob/main/tutorials/LevelsOfTextSplitting/5_Levels_Of_Text_Splitting.ipynb

All credits to him.

At a high level, this splits into sentences, then groups into groups of 3 sentences, and then merges one that are similar in the embedding space.

Methods

__init__(embeddings[, buffer_size, ...])
atransform_documents(documents, **kwargs)	Asynchronously transform a list of documents.
create_documents(texts[, metadatas])	Create documents from a list of texts.
split_documents(documents)	Split documents.
split_text(text)
transform_documents(documents, **kwargs)	Transform sequence of documents by splitting them.

Parameters

• embeddings (Embeddings) –

• buffer_size (int) –

• add_start_index (bool) –

• breakpoint_threshold_type (Literal['percentile', 'standard_deviation', 'interquartile']) –

• breakpoint_threshold_amount (Optional[float]) –

• number_of_chunks (Optional[int]) –

• sentence_split_regex (str) –

__init__(embeddings: Embeddings, buffer_size: int = 1, add_start_index: bool = False, breakpoint_threshold_type: Literal['percentile', 'standard_deviation', 'interquartile'] = 'percentile', breakpoint_threshold_amount: Optional[float] = None, number_of_chunks: Optional[int] = None, sentence_split_regex: str = '(?<=[.?!])\\s+')

Parameters

• embeddings (Embeddings) –

• buffer_size (int) –

• add_start_index (bool) –

• breakpoint_threshold_type (Literal['percentile', 'standard_deviation', 'interquartile']) –

• breakpoint_threshold_amount (Optional[float]) –

• number_of_chunks (Optional[int]) –

• sentence_split_regex (str) –

async atransform_documents(documents: Sequence[Document], **kwargs: Any) → Sequence[Document]

Asynchronously transform a list of documents.

Parameters

• documents (Sequence[Document]) – A sequence of Documents to be transformed.

• kwargs (Any) –

Returns

A list of transformed Documents.

Return type

Sequence[Document]

create_documents(texts: List[str], metadatas: Optional[List[dict]] = None) → List[Document]

Create documents from a list of texts.

Parameters

• texts (List[str]) –

• metadatas (Optional[List[dict]]) –

Return type

List[Document]

split_documents(documents: Iterable[Document]) → List[Document]

Split documents.

Parameters

documents (Iterable[Document]) –

Return type

List[Document]

split_text(text: str) → List[str]

Parameters

text (str) –

Return type

List[str]

transform_documents(documents: Sequence[Document], **kwargs: Any) → Sequence[Document]

Transform sequence of documents by splitting them.

Parameters

• documents (Sequence[Document]) –

• kwargs (Any) –

Return type

Sequence[Document]

Examples using SemanticChunker

• This is a long document we can split up.