Tutorials

Build Your First QA System

Get Started by creating a Retriever Reader pipeline.

Build a QA System Without Elasticsearch

Create a Retriever Reader pipeline that requires no external database dependencies.

Utilizing Existing FAQs for Question Answering

Create a smarter way to answer new questions using your existing FAQ documents.

Preprocessing Your Documents

Start converting, cleaning, and splitting Documents using Haystack’s preprocessing capabilities.

Fine-Tuning a Model on Your Own Data

Improve the performance of your Reader by performing fine-tuning.

Better Retrieval with Embedding Retrieval

Use Transformer based dense Retrievers to improve your system’s performance.

Generative QA with Retrieval-Augmented Generation

Try out a generative model in place of the extractive Reader.

How to Use Pipelines

Learn about the many ways which you can route queries through the nodes in a pipeline.

Generative QA with LFQA

Try out a generative model in place of the extractive Reader.

Question Generation

Generate a set of questions that can be answered by a given Document.

Query Classifier

Classify incoming queries so that they can be routed to the nodes that are best at handling them.

Document Classification at Index Time

Generate and attach classification labels to your Documents when indexing.

Make Your QA Pipelines Talk!

Convert text Answers into speech.

Text-To-Image Search Pipeline with Multimodal Retriever

Use a MultiModalRetriever to build a cross-modal search pipeline.

Evaluation of a QA System

Learn how to evaluate the performance of individual nodes as well as entire pipelines.

Training Your Own Dense Passage Retrieval Model

Learn about training a Dense Passage Retrieval model and the data needed to do so.

Question Answering on a Knowledge Graph

Experiment with a question answering system that draws upon knowledge graph.h

Open-Domain QA on Tables

Perform question answering on tabular data.

Generative Pseudo Labeling for Domain Adaptation

Use a Retriever and a query generator to perform unsupervised domain adaptation.