Query Decomposition

QueryDecomposition Class Documentation

Overview

The QueryDecomposition class is used to decompose a multi-hop question into multiple single-hop questions using a LLM model. The class uses a default decomposition prompt from the Visconde paper. And default prompt is derived from few-shot prompts from the strategyQA dataset.

Usage

Initialize

To use the QueryDecomposition class, you first need to create an instance of the class.

from RAGchain.utils.query_decompose import QueryDecomposition

decomposer = QueryDecomposition()

You can put additional parameter model_name and api_base for using custom model.

Decompose

After the QueryDecomposition instance has been initialized, you can use the decompose method to decompose a query into multiple single-hop questions. The decompose method takes a query (a string) as input and returns a list of decomposed queries.

query = "Is it legal for a licensed child driving Mercedes-Benz to be employed in US?"
decomposed_queries = decomposer.decompose(query)
print(decomposed_queries)
>>> ["What is the minimum driving age in the US?", "What is the minimum age for someone to be employed in the US?"]

If the input query is not multi-hop question, so it doesn't need any decomposition, it returns empty list.

query = "What is the atomic number of hydrogen?" # it is single-hop question
decomposed_queries = decomposer.decompose(query)
print(len(decomposed_queries))
>>> 0