Retrieval-augmented generation

Last updated

Retrieval Augmented Generation (RAG) is a technique that grants generative artificial intelligence models information retrieval capabilities. It modifies interactions with a large language model (LLM) so that the model responds to user queries with reference to a specified set of documents, using this information to augment information drawn from its own vast, static training data. This allows LLMs to use domain-specific and/or updated information. [1] Use cases include providing chatbot access to internal company data or giving factual information only from an authoritative source. [2]

Contents

Process

The RAG process is made up of four key stages. First, all the data must be prepared and indexed for use by the LLM. Thereafter, each query consists of a retrieval, augmentation, and generation phase. [1]

Indexing

Typically, the data to be referenced is converted into LLM embeddings, numerical representations in the form of large vectors. RAG can be used on unstructured (usually text), semi-structured, or structured data (for example knowledge graphs). [1] These embeddings are then stored in a vector database to allow for document retrieval.

Overview of RAG process, combining external documents and user input into an LLM prompt to get tailored output RAG diagram.svg
Overview of RAG process, combining external documents and user input into an LLM prompt to get tailored output

Retrieval

Given a user query, a document retriever is first called to select the most relevant documents that will be used to augment the query. [3] This comparison can be done using a variety of methods, which depend in part on the type of indexing used. [1]

Augmentation

The model feeds this relevant retrieved information into the LLM via prompt engineering of the user's original query. [2] Newer implementations (as of 2023) can also incorporate specific augmentation modules with abilities such as expanding queries into multiple domains and using memory and self-improvement to learn from previous retrievals. [1]

Generation

Finally, the LLM can generate output based on both the query and the retrieved documents. [4] Some models incorporate extra steps to improve output, such as the re-ranking of retrieved information, context selection, and fine-tuning. [1]

Improvements

Improvements to the basic process above can be applied at different stages in the RAG flow.

Encoder

These methods center around the encoding of text as either dense or sparse vectors. Sparse vectors, used to encode the identity of a word, are typically dictionary length and contain almost all zeros. Dense vectors, used to encode meaning, are much smaller and contain far fewer zeros. Several enhancements can be made to the way similarities are calculated in the vector stores (databases).

Retriever-centric methods

These methods focus on improving the quality of hits from the vector database:

Language model

Language model in Deepmind's 2021 Retro for RAG.svg
Retro language model for RAG. Each Retro block consists of Attention, Chunked Cross Attention, and Feed Forward layers. Black-lettered boxes show data being changed, and blue lettering shows the algorithm performing the changes.

By redesigning the language model with the retriever in mind, a 25-time smaller network can get comparable perplexity as its much larger counterparts. [12] Because it is trained from scratch, this method (Retro) incurs the high cost of training runs that the original RAG scheme avoided. The hypothesis is that by giving domain knowledge during training, Retro needs less focus on the domain and can devote its smaller weight resources only to language semantics. The redesigned language model is shown here.

It has been reported that Retro is not reproducible, so modifications were made to make it so. The more reproducible version is called Retro++ and includes in-context RAG. [13]

Chunking

Chunking involves various strategies for breaking up the data into vectors so the retriever can find details in it.

Rag-doc-styles.png
Different data styles have patterns that correct chunking can take advantage of.

Three types of chunking strategies are:

Challenges

If the external data source is large, retrieval can be slow. The use of RAG does not completely eliminate the general challenges faced by LLMs, including hallucination. [3]

Related Research Articles

Information retrieval (IR) in computing and information science is the task of identifying and retrieving information system resources that are relevant to an information need. The information need can be specified in the form of a search query. In the case of document retrieval, queries can be based on full-text or other content-based indexing. Information retrieval is the science of searching for information in a document, searching for documents themselves, and also searching for the metadata that describes data, and for databases of texts, images or sounds.

Question answering (QA) is a computer science discipline within the fields of information retrieval and natural language processing (NLP) that is concerned with building systems that automatically answer questions that are posed by humans in a natural language.

Latent semantic analysis (LSA) is a technique in natural language processing, in particular distributional semantics, of analyzing relationships between a set of documents and the terms they contain by producing a set of concepts related to the documents and terms. LSA assumes that words that are close in meaning will occur in similar pieces of text. A matrix containing word counts per document is constructed from a large piece of text and a mathematical technique called singular value decomposition (SVD) is used to reduce the number of rows while preserving the similarity structure among columns. Documents are then compared by cosine similarity between any two columns. Values close to 1 represent very similar documents while values close to 0 represent very dissimilar documents.

<span class="mw-page-title-main">Autoencoder</span> Neural network that learns efficient data encoding in an unsupervised manner

An autoencoder is a type of artificial neural network used to learn efficient codings of unlabeled data. An autoencoder learns two functions: an encoding function that transforms the input data, and a decoding function that recreates the input data from the encoded representation. The autoencoder learns an efficient representation (encoding) for a set of data, typically for dimensionality reduction, to generate lower-dimensional embeddings for subsequent use by other machine learning algorithms.

In statistical classification, the Fisher kernel, named after Ronald Fisher, is a function that measures the similarity of two objects on the basis of sets of measurements for each object and a statistical model. In a classification procedure, the class for a new object can be estimated by minimising, across classes, an average of the Fisher kernel distance from the new object to each known member of the given class.

Query expansion (QE) is the process of reformulating a given query to improve retrieval performance in information retrieval operations, particularly in the context of query understanding. In the context of search engines, query expansion involves evaluating a user's input and expanding the search query to match additional documents. Query expansion involves techniques such as:

Ontology learning is the automatic or semi-automatic creation of ontologies, including extracting the corresponding domain's terms and the relationships between the concepts that these terms represent from a corpus of natural language text, and encoding them with an ontology language for easy retrieval. As building ontologies manually is extremely labor-intensive and time-consuming, there is great motivation to automate the process.

Term discrimination is a way to rank keywords in how useful they are for information retrieval.

Vector space model or term vector model is an algebraic model for representing text documents as vectors such that the distance between vectors represents the relevance between the documents. It is used in information filtering, information retrieval, indexing and relevancy rankings. Its first use was in the SMART Information Retrieval System.

Multimedia information retrieval is a research discipline of computer science that aims at extracting semantic information from multimedia data sources. Data sources include directly perceivable media such as audio, image and video, indirectly perceivable sources such as text, semantic descriptions, biosignals as well as not perceivable sources such as bioinformation, stock prices, etc. The methodology of MMIR can be organized in three groups:

  1. Methods for the summarization of media content. The result of feature extraction is a description.
  2. Methods for the filtering of media descriptions
  3. Methods for the categorization of media descriptions into classes.

Multimodal learning is a type of deep learning that integrates and processes multiple types of data, referred to as modalities, such as text, audio, images, or video. This integration allows for a more holistic understanding of complex data, improving model performance in tasks like visual question answering, cross-modal retrieval, text-to-image generation, aesthetic ranking, and image captioning.

<span class="mw-page-title-main">Semantic parsing</span>

Semantic parsing is the task of converting a natural language utterance to a logical form: a machine-understandable representation of its meaning. Semantic parsing can thus be understood as extracting the precise meaning of an utterance. Applications of semantic parsing include machine translation, question answering, ontology induction, automated reasoning, and code generation. The phrase was first used in the 1970s by Yorick Wilks as the basis for machine translation programs working with only semantic representations. Semantic parsing is one of the important tasks in computational linguistics and natural language processing.

Paraphrase or paraphrasing in computational linguistics is the natural language processing task of detecting and generating paraphrases. Applications of paraphrasing are varied including information retrieval, question answering, text summarization, and plagiarism detection. Paraphrasing is also useful in the evaluation of machine translation, as well as semantic parsing and generation of new samples to expand existing corpora.

In natural language processing, a sentence embedding refers to a numeric representation of a sentence in the form of a vector of real numbers which encodes meaningful semantic information.

<span class="mw-page-title-main">Transformer (deep learning architecture)</span> Deep learning architecture for modelling sequential data

A transformer is a deep learning architecture developed by researchers at Google and based on the multi-head attention mechanism, proposed in the 2017 paper "Attention Is All You Need". Text is converted to numerical representations called tokens, and each token is converted into a vector via lookup from a word embedding table. At each layer, each token is then contextualized within the scope of the context window with other (unmasked) tokens via a parallel multi-head attention mechanism, allowing the signal for key tokens to be amplified and less important tokens to be diminished.

<span class="mw-page-title-main">Contrastive Language-Image Pre-training</span> Technique in neural networks for learning joint representations of text and images

Contrastive Language-Image Pre-training (CLIP) is a technique for training a pair of neural network models, one for image understanding and one for text understanding, using a contrastive objective. This method has enabled broad applications across multiple domains, including cross-modal retrieval, text-to-image generation, aesthetic ranking, and image captioning.

Prompt engineering is the process of structuring an instruction that can be interpreted and understood by a generative artificial intelligence (AI) model. A prompt is natural language text describing the task that an AI should perform. A prompt for a text-to-text language model can be a query such as "what is Fermat's little theorem?", a command such as "write a poem in the style of Edgar Allan Poe about leaves falling", or a longer statement including context, instructions, and conversation history.

LangChain is a software framework that helps facilitate the integration of large language models (LLMs) into applications. As a language model integration framework, LangChain's use-cases largely overlap with those of language models in general, including document analysis and summarization, chatbots, and code analysis.

A vector database, vector store or vector search engine is a database that can store vectors along with other data items. Vector databases typically implement one or more Approximate Nearest Neighbor algorithms, so that one can search the database with a query vector to retrieve the closest matching database records.

Learned sparse retrieval or sparse neural search is an approach to Information Retrieval which uses a sparse vector representation of queries and documents. It borrows techniques both from lexical bag-of-words and vector embedding algorithms, and is claimed to perform better than either alone. The best-known sparse neural search systems are SPLADE and its successor SPLADE v2. Others include DeepCT, uniCOIL, EPIC, DeepImpact, TILDE and TILDEv2, Sparta, SPLADE-max, and DistilSPLADE-max.

References

  1. 1 2 3 4 5 6 Gao, Yunfan; Xiong, Yun; Gao, Xinyu; Jia, Kangxiang; Pan, Jinliu; Bi, Yuxi; Dai, Yi; Sun, Jiawei; Wang, Meng; Wang, Haofen (2023). "Retrieval-Augmented Generation for Large Language Models: A Survey". arXiv: 2312.10997 [cs.CL].
  2. 1 2 "What is RAG? - Retrieval-Augmented Generation AI Explained - AWS". Amazon Web Services, Inc. Retrieved 16 July 2024.
  3. 1 2 "Next-Gen Large Language Models: The Retrieval-Augmented Generation (RAG) Handbook". freeCodeCamp.org. 11 June 2024. Retrieved 16 July 2024.
  4. ↑ Lewis, Patrick; Perez, Ethan; Piktus, Aleksandra; Petroni, Fabio; Karpukhin, Vladimir; Goyal, Naman; Küttler, Heinrich; Lewis, Mike; Yih, Wen-tau; Rocktäschel, Tim; Riedel, Sebastian; Kiela, Douwe (2020). "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks". Advances in Neural Information Processing Systems. 33. Curran Associates, Inc.: 9459–9474. arXiv: 2005.11401 .
  5. ↑ "faiss". GitHub .
  6. ↑ Khattab, Omar; Zaharia, Matei (2020). ""ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT"". doi:10.1145/3397271.3401075.
  7. ↑ Formal, Thibault; Lassance, Carlos; Piwowarski, Benjamin; Clinchant, Stéphane (2021). ""SPLADE v2: Sparse Lexical and Expansion Model for Information Retrieval"". Arxiv. S2CID   237581550.
  8. ↑ Lee, Kenton; Chang, Ming-Wei; Toutanova, Kristina (2019). ""Latent Retrieval for Weakly Supervised Open Domain Question Answering"" (PDF).
  9. ↑ Lin, Sheng-Chieh; Asai, Akari (2023). ""How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval"" (PDF).
  10. ↑ Shi, Weijia; Min, Sewon (2024). "REPLUG: Retrieval-Augmented Black-Box Language Models". "REPLUG: Retrieval-Augmented Black-Box Language Models". pp. 8371–8384. arXiv: 2301.12652 . doi:10.18653/v1/2024.naacl-long.463.
  11. ↑ Ram, Ori; Levine, Yoav; Dalmedigos, Itay; Muhlgay, Dor; Shashua, Amnon; Leyton-Brown, Kevin; Shoham, Yoav (2023). ""In-Context Retrieval-Augmented Language Models"". Transactions of the Association for Computational Linguistics. 11: 1316–1331. arXiv: 2302.00083 . doi:10.1162/tacl_a_00605.
  12. ↑ Borgeaud, Sebastian; Mensch, Arthur (2021). ""Improving language models by retrieving from trillions of tokens"" (PDF).
  13. ↑ Wang, Boxin; Ping, Wei (2023). ""Shall We Pretrain Autoregressive Language Models with Retrieval? A Comprehensive Study"" (PDF).
This page is based on this Wikipedia article
Text is available under the CC BY-SA 4.0 license; additional terms may apply.
Images, videos and audio are available under their respective licenses.