Skip to content
Permalink
master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Go to file
@nreimers
Latest commit 1a44af1 Feb 9, 2022 History
4 contributors

Users who have contributed to this file

@nreimers @RafaelWO @lukesalamone @Omarfos

Semantic Search

Semantic search seeks to improve search accuracy by understanding the content of the search query. In contrast to traditional search engines which only find documents based on lexical matches, semantic search can also find synonyms.

Background

The idea behind semantic search is to embed all entries in your corpus, whether they be sentences, paragraphs, or documents, into a vector space.

At search time, the query is embedded into the same vector space and the closest embeddings from your corpus are found. These entries should have a high semantic overlap with the query.

SemanticSearch

Symmetric vs. Asymmetric Semantic Search

A critical distinction for your setup is symmetric vs. asymmetric semantic search:

  • For symmetric semantic search your query and the entries in your corpus are of about the same length and have the same amount of content. An example would be searching for similar questions: Your query could for example be "How to learn Python online?" and you want to find an entry like "How to learn Python on the web?". For symmetric tasks, you could potentially flip the query and the entries in your corpus.
  • For asymmetric semantic search, you usually have a short query (like a question or some keywords) and you want to find a longer paragraph answering the query. An example would be a query like "What is Python" and you wand to find the paragraph "Python is an interpreted, high-level and general-purpose programming language. Python's design philosophy ...". For asymmetric tasks, flipping the query and the entries in your corpus usually does not make sense.

It is critical that you choose the right model for your type of task.

Suitable models for symmetric semantic search: Pre-Trained Sentence Embedding Models

Suitable models for asymmetric semantic search: Pre-Trained MS MARCO Models

Python

For small corpora (up to about 1 million entries) we can compute the cosine-similarity between the query and all entries in the corpus.

In the following example, we define a small corpus with few example sentences and compute the embeddings for the corpus as well as for our query.

We then use the util.cos_sim() function to compute the cosine similarity between the query and all corpus entries.

For large corpora, sorting all scores would take too much time. Hence, we use torch.topk to only get the top k entries.

For a simple example, see semantic_search.py:

.. literalinclude:: semantic_search.py

util.semantic_search

Instead of implementing semantic search by yourself, you can use the util.semantic_search function.

The function accepts the following parameters:

.. autofunction:: sentence_transformers.util.semantic_search

By default, up to 100 queries are processed in parallel. Further, the corpus is chunked into set of up to 500k entries. You can increase query_chunk_size and corpus_chunk_size, which leads to increased speed for large corpora, but also increases the memory requirement.

Speed Optimization

To get the optimal speed for the util.semantic_search method, it is advisable to have the query_embeddings as well as the corpus_embeddings on the same GPU-device. This significantly boost the performance.

Further, we can normalize the corpus embeddings so that each corpus embeddings is of length 1. In that case, we can use dot-product for computing scores.

corpus_embeddings = corpus_embeddings.to('cuda')
corpus_embeddings = util.normalize_embeddings(corpus_embeddings)

query_embeddings = query_embeddings.to('cuda')
query_embeddings = util.normalize_embeddings(query_embeddings)
hits = util.semantic_search(query_embeddings, corpus_embeddings, score_function=util.dot_score)

ElasticSearch

Starting with version 7.3, ElasticSearch introduced the possibility to index dense vectors and to use to for document scoring. Hence, we can use ElasticSearch to index embeddings along the documents and we can use the query embeddings to retrieve relevant entries.

An advantage of ElasticSearch is that it is easy to add new documents to an index and that we can store also other data along with our vectors. A disadvantage is the slow performance, as it compares the query embeddings with all stored embeddings. This has a linear run-time and might be too slow for large (>100k) corpora.

For further details, see semantic_search_quora_elasticsearch.py.

Approximate Nearest Neighbor

Searching a large corpus with millions of embeddings can be time-consuming if exact nearest neighbor search is used (like it is used by util.semantic_search).

In that case, Approximate Nearest Neighor (ANN) can be helpful. Here, the data is partitioned into smaller fractions of similar embeddings. This index can be searched efficiently and the embeddings with the highest similarity (the nearest neighbors) can be retrieved within milliseconds, even if you have millions of vectors.

However, the results are not necessarily exact. It is possible that some vectors with high similarity will be missed. That's the reason why it is called approximate nearest neighbor.

For all ANN methods, there are usually one or more parameters to tune that determine the recall-speed trade-off. If you want the highest speed, you have a high chance of missing hits. If you want high recall, the search speed decreases.

Three popular libraries for approximate nearest neighbor are Annoy, FAISS, and hnswlib. Personally I find hnswlib the most suitable library: It is easy to use, offers a great performance and has nice features included that are important for real applications.

Examples:

Retrieve & Re-Rank

For complex semantic search scenarios, a retrieve & re-rank pipeline is advisable: InformationRetrieval

For further details, see Retrieve & Re-rank.

Examples

In the following we list examples for different use-cases.

Similar Questions Retrieval

semantic_search_quora_pytorch.py [ Colab version ] shows an example based on the Quora duplicate questions dataset. The user can enter a question, and the code retrieves the most similar questions from the dataset using the util.semantic_search method. As model, we use distilbert-multilingual-nli-stsb-quora-ranking, which was trained to identify similar questions and supports 50+ languages. Hence, the user can input the question in any of the 50+ languages. This is a symmetric search task, as the search queries have the same length and content as the questions in the corpus.

Similar Publication Retrieval

semantic_search_publications.py [ Colab version ] shows an example how to find similar scientific publications. As corpus, we use all publications that have been presented at the EMNLP 2016 - 2018 conferences. As search query, we input the title and abstract of more recent publications and find related publications from our copurs. We use the SPECTER model. This is a symmetric search task, as the paper in the corpus consists of title & abstract and we search for title & abstract.

Question & Answer Retrieval

semantic_search_wikipedia_qa.py [ Colab Version ]: This example uses a model that was trained on the Natural Questions dataset. It consists of about 100k real Google search queries, together with an annotated passage from Wikipedia that provides the answer. It is an example of an asymmetric search task. As corpus, we use the smaller Simple English Wikipedia so that it fits easily into memory.

retrieve_rerank_simple_wikipedia.py [ Colab Version ]: This script uses the Retrieve & Re-rank strategy and is an example for an asymmetric search task. We split all Wikipedia articles into paragraphs and encode them with a bi-encoder. If a new query / question is entered, it is encoded by the same bi-encoder and the paragraphs with the highest cosine-similarity are retrieved (see semantic search). Next, the retrieved candidates are scored by a Cross-Encoder re-ranker and the 5 passages with the highest score from the Cross-Encoder are presented to the user. We use models that were trained on the MS Marco Passage Reranking dataset, a dataset with about 500k real queries from Bing search.