get keyphrases from TestGenerator

[ableiweiss]

Is it get the keyphrases extracted by TestGenerator using the API?

[dosubot]

To get the keyphrases extracted by TestGenerator using the API, you can use the KeyphraseExtractor class and its extract method. Here is an example of how you can do this:

from ragas.testset.extractor import KeyphraseExtractor
from ragas.llms.base import BaseRagasLLM
from ragas.testset.docstore import Node

# Assuming you have an instance of BaseRagasLLM and a Node
llm = BaseRagasLLM()  # Initialize your LLM instance
node = Node(page_content="Your text content here")  # Initialize your Node with the text content

# Create an instance of KeyphraseExtractor
keyphrase_extractor = KeyphraseExtractor(llm=llm)

# Extract keyphrases
keyphrases = await keyphrase_extractor.extract(node=node, is_async=True)

print(keyphrases)

This code initializes the KeyphraseExtractor with an instance of BaseRagasLLM and a Node containing the text content. It then calls the extract method to get the keyphrases. The extract method returns a list of keyphrases extracted from the text content.

Additionally, the TestsetGenerator class provides a convenient way to initialize the KeyphraseExtractor when creating a document store. Here is an example of how you can use TestsetGenerator to set up the keyphrase extraction:

from ragas.testset.generator import TestsetGenerator
from ragas.llms import BaseRagasLLM
from ragas.embeddings.base import BaseRagasEmbeddings
from ragas.testset.docstore import InMemoryDocumentStore

# Initialize your LLM and embeddings instances
generator_llm = BaseRagasLLM()
critic_llm = BaseRagasLLM()
embeddings = BaseRagasEmbeddings()

# Create a TestsetGenerator instance
testset_generator = TestsetGenerator(
    generator_llm=generator_llm,
    critic_llm=critic_llm,
    embeddings=embeddings,
    docstore=InMemoryDocumentStore()
)

# Now you can use the testset_generator to handle keyphrase extraction

This approach ensures that the KeyphraseExtractor is properly integrated into the document store, making it easier to manage and use within your RAG pipeline ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[jjmachan]

@ableiweiss that is an interesting suggestion. How are you planning to use it?

We are working on a rewamp for testset generation #1016 will add it to that but if you could explain your reasoning and usecase it will help us a lot 🙂