retrieval-augmented-generation
- Retrieval methods in RAG - Haystack (24 Apr 2025)
A tour of retrieval methods for RAG, from baseline and classical techniques to sentence-window, auto-merging and maximum marginal relevance, as presented at PyCon Lithuania 2025. - Improving RAG Retrieval with Auto-Merging (12 Sep 2024)
How auto-merging retrieval improves RAG by returning the parent document once enough of its splits match a query, with a worked example built from Haystack components. - Benchmarking Haystack Pipelines for Optimal Performance (24 Jun 2024)
How to evaluate the performance of a RAG pipeline with Haystack, using the ARAGOG dataset to build the indexing, RAG and evaluation pipelines and compare the results. - Extract Metadata from Queries to Improve Retrieval (13 May 2024)
Use LLMs to extract metadata from queries to use as filters that improve retrieval in RAG applications. - Incorporate HyDE into Haystack RAG pipelines (28 Feb 2024)
How to build Hypothetical Document Embeddings into a Haystack RAG pipeline, from assembling the components to packaging them as a reusable HypotheticalDocumentEmbedder.
retrieval-augmented-generation Haystack viterbi sequence-prediction evaluation-metrics scikit-learn pos-tags named-entity-recognition embeddings conditional-random-fields classification word2vec triplet-loss syntactic-dependencies sentence-transformers relationship-extraction neural-networks information-retrieval fine-tuning coursera conference SyntaxNet NLTK LSTM wikidata transformers tokenization tf-idf text-summarisation semantic-web semantic-drift resources reference-post production portuguese political-science naive-bayes multi-label-classification monitoring mlops metadata-extraction maximum-entropy-markov-models logistic-regression llms language-models information-extraction imbalanced-data hyperparameter-optimization hidden-markov-models grid-search gensim generative-ai fasttext document-classification doc2vec deployment dependency-graph dataset data-challenge convolutional-neural-networks contrastive-learning books attention SPARQL RNN PyData KONVENS GRU