Extracted Summary Sentences Match

Hub Blog Docs

Loading ...

🏅Guardrails Certified

Last updated 4 months ago

🇬🇧 EN

🧵 STRING

🤖 ML

📰 FACTUALITY

🚫 BRAND RISK

📋 SUMMARIZATION

Overview

Developed by	Guardrails AI
Date of development	Feb 15, 2024
Validator type	Summarization
Blog
License	Apache 2
Input/Output	Output

Description

This validator checks that sentences in a generated summary match the original text by performing a cosine similarity in the embedding space.

Intended use

This validator is only useful when performing summarization. If the summary is correct but is an abstractive summary, this validator will give false negatives.

Requirements

Dependencies:
- faiss
- openai
- sqlalchemy
- guardrails-ai>=0.4.0

Installation

$ guardrails hub install hub://guardrails/extracted_summary_sentences_match

Usage Examples

Validating string output via Python

In this example, we apply the validator to a string output generated by an LLM.

# Import Guard and Validator
from guardrails.hub import ExtractedSummarySentencesMatch
from guardrails import Guard

# Initialize Validator
val = ExtractedSummarySentencesMatch(
    threshold=0.8,
    filepaths="/path/to/original/documents"
)

# Setup Guard
guard = Guard.from_string(validators=[val, ...])

guard("Summarized text")  # Validator passes
guard("Inaccurate summary")  # Validator fails

Validating JSON output via Python

In this example, we apply the validator to a string field of a JSON output generated by an LLM.

# Import Guard and Validator
from pydantic import BaseModel
from guardrails.hub import ExtractedSummarySentencesMatch
from guardrails import Guard

# Initialize Validator
val = ExtractedSummarySentencesMatch(
    threshold=0.8,
    filepaths="/path/to/original/documents"
)

# Create Pydantic BaseModel
class Summary(BaseModel):
    title: str
    summary: str = Field(
	description="Summary of article", validators=[val]
    )

# Create a Guard to check for valid Pydantic output
guard = Guard.from_pydantic(output_class=Summary)

# Run LLM output generating JSON through guard
guard.parse("""
{
    "title": "Using Guardrails Hub",
    "summary": "To use Guardrails Hub, use the hub cli to download validators."
}
""")

API Reference

__init__(self, on_fail="noop")

Initializes a new instance of the Validator class.

Parameters:

threshold (float): The minimum cosine similarity to be considered similar. Default to 0.7.
on_fail (str, Callable): The policy to enact when a validator fails. If str, must be one of reask, fix, filter, refrain, noop, exception or fix_reask. Otherwise, must be a function that is called when the validator fails.

validate(self, value, metadata={}) -> ValidationOutcome

Validates the given value using the rules defined in this validator, relying on the metadata provided to customize the validation process. This method is automatically invoked by guard.parse(...), ensuring the validation logic is applied to the input data.

Note:

This method should not be called directly by the user. Instead, invoke guard.parse(...) where this method will be called internally for each associated Validator.
When invoking guard.parse(...), ensure to pass the appropriate metadata dictionary that includes keys and values required by this validator. If guard is associated with multiple validators, combine all necessary metadata into a single dictionary.

Parameters:

value (Any): The input value to validate.

metadata (dict): A dictionary containing metadata required for validation. Keys and values must match the expectations of this validator.

Key	Type	Description	Default
`filepaths`	list[str]	A list of strings that specifies the filepaths for any documents that should be used for asserting the summary's similarity.	N/A
`document_store`	Optional[DocumentStoreBase]	The document store to use during validation. Defaults to EphemeralDocumentStore.	None
`vector_db`	Optional[VectorDBBase]	A vector database to use for embeddings. Defaults to Faiss.	None
`embedding_model`	Optional[EmbeddingBase]	The embeddig model to use. Defaults to OpenAIEmbedding.	None

Guardrails API

The validator playground is available to authenticated users. Please log in to use it.