There is a lot of content on getting started with Retrieval Augmented Generation (RAG), and there are numerous posts on various forms of “advanced” RAG. LangChain lists 8 different PDF loaders, 4 different text splitters, and a variety of techniques you can add such as Parent-Document Retrieval and Multi-Vector indexing. It is easy to be paralyzed with doubt – which of these should you use? If you’re wondering which of these techniques is best for your domain, the long answer is – you should understand which techniques may help your documents and run experiments on your data to see which of those help. If you're in a hurry, the short answer is that how you extract information matters most Embeddings and advanced splitting and indexing strategies don't help (or actively hurt) retrieval. To understand the importance of different factors and prioritize the implementation of corresponding techniques within (the open-source knowledge base we’re developing), we ran some experiments on different configurations. Dewy We’re posting these to hopefully save you some time by using these results as a starting point and possibly using some of the methodology from this post for your own measurements. Experimentation There are a variety of frameworks for evaluating RAG and LLM usage, such as and – at the time of writing, LangSmith supports evaluation but doesn’t consider the context, making it less suited for evaluating RAG. TruLens RAGAS The general pattern is to record the question and answer (and often the retrieved contexts) and then use an LLM to grade the responses. Typically, you may ask for grades along different metrics – such as “Are the retrieved contexts relevant to the question?” Or “Is the generated answer grounded in the retrieved context?” This approach lets you measure the performance across a large set of generated responses, and fits well with the existing tracing systems like . For this article, I used RAGAS. LangSmith The data set I’m using is the . This data set includes a PDF representing the source content (the document or documents to index and retrieve) and a set of 160 questions and answers. History of Alexnet For your experiments, it’s best to have your LLM application record actual questions and possible golden answers – generated by a human or using a larger model and/or larger context. All of the frameworks mentioned support running the experiments out-of-band. This allows you to record the actual questions and answers your users are asking, and then run an evaluation after the fact to assess the quality of the answers you’ve served up. This is very useful for monitoring your LLM application. RAGAS provides a variety of . metrics For these experiments, I’m only measuring the following metrics related to retrieval. Whether the context items relevant to the ground-truth answer are ranked higher than those irrelevant. : Context Precision The proportion of sentences in the ground-truth answer attributed to the retrieved context. This uses the ground-truth answer as a proxy for the "correct" context. : Context Recall To simplify comparisons in this post, we use the geometric mean of precision and recall. This is often called the “f1 score”, but since the RAGAS metrics context recall and precision are slightly different, we’ll call this the “context f1.” See the code for the experiments in GitHub. Results For our experiments, we ran using 3 different extractors ( , , and ), 5 different chunking and indexing strategies ( , , , and with both recursive and semantic chunking), 2 different embeddings (OpenAI and ) and 2 different search strategies (similarity and maximum-marginal relevance). We ran the experiment and evaluation on the cross-product of all of these factors. pypdf pymupdf unstructured recursive chunking parent-document chunking semantic chunking questions-answered multi-vector embedding ada-002 bge-en-small The results are available in a , and the used to produce the charts is available in Google Colab. zip file analysis notebook For each of the choices, we show a heatmap comparing the median of the relative differences in Context F1 between the choices. Extraction Our suspicion was that extraction would play a significant role in the results. After all, if the wrong text is extracted from a PDF, or it is extracted in a confusing manner there isn’t much that can be done to retrieve the right content. The results largely supported this. In this visualization, you can see that switching from to produces an 87% increase in f1 score, and changing from to produces a 72% increase. pymupdf unstructured pypdf unstructured Chunking & Indexing On the indexing front, the results were somewhat surprising. We expected more sophisticated chunking strategies like to outperform the basic chunking. But the results suggest that generally outperforms the other techniques. semantic recursive recursive As you can see, indexing produces context f1 scores 81% better than , and 56% better than parent indexing. recursive semantic Generating the questions-answered is a relatively expensive process (applying an LLM during indexing) and didn’t seem to significantly improve upon the baseline chunking. In the future, we’d like to rerun these experiments on larger datasets to see if these advanced indexing techniques become more important in different cases. Until then, this suggests that for simple RAG use cases, simple indexing strategies like work well. recursive Search Many vector store retrievers support basic similarity search as well as a “maximum marginal relevance” or MMR. This is a technique to retrieve more diverse chunks with the hopes of producing richer contexts. We expected MMR to outperform similarity, but in our experiments, we saw the opposite - MMR reduced f1 scores by 13%. Our hypothesis is that because the dataset has a single PDF, there isn’t much benefit to increased diversity. This is another experiment we’d like to re-run on different datasets in the future. Embeddings On embeddings, we expected the larger (and better) embeddings from OpenAI ada-002 with 1536 dimensions to outperform which only has 512 dimensions. bge-small-en-v1.5 While there is technically a difference, it is insignificant. Conclusion While it is important to consider (and measure!) the benefits of these techniques on your own documents and questions, the above analysis suggests that the most important thing you can do for your RAG application is to ensure you’re extracting the right content. After that, relatively simple chunking, indexing, and retrieval provide good results. In the future, we’d like to re-run these experiments with a larger dataset and by feeding the entire set of documents through the models rather than comparing advanced RAG techniques to baseline answers using naive RAG. Don’t be paralyzed by all the options – nearly any choice you use will allow answering questions over the domain-specific data and significantly improve the experience of using your GenAI application. Providing the right context reduces hallucinations and lets you use GenAI to answer questions about domain-specific information. makes this easy. It provides a RAG service with built-in indexing and retrieval built on the lessons from these experiments and more, allowing you to add documents and retrieve relevant contexts with a and then get back to improving your GenAI application! Dewy simple API Also published here

This story contains new, firsthand information uncovered by the writer.

Here's Why Extraction Matters the Most

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

100 Days of AI Day 7: Building Your Own ChatGPT with Langchain

A Complete Overview of Embeddings in LangChain

A Comprehensive Guide to LangChain: Building Powerful Applications with Large Language Models

A Practical 5-Step Guide to Do Semantic Search on Your Private Data With the Help of LLMs

An Introduction to LangChain Use Cases with DeepInfra

100 Days of AI Day 7: Building Your Own ChatGPT with Langchain

A Complete Overview of Embeddings in LangChain

A Comprehensive Guide to LangChain: Building Powerful Applications with Large Language Models

A Practical 5-Step Guide to Do Semantic Search on Your Private Data With the Help of LLMs

An Introduction to LangChain Use Cases with DeepInfra

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps