This type of search utilizes mathematical vectors to represent documents or items in a high-dimensional space. It allows for similarity-based search, where similar items are retrieved based on their proximity in the vector space
This is also called "Semantic Search"
This aims to understand the context and meaning behind search queries, rather than just matching keywords. It considers the intent of the user and the relationships between words to deliver more relevant results.
"cats" can also refer to "felines" or "kittens."
Semantic search understands the meaning behind your query, so searching for 'cats' might also bring results about felines or kittens. It provides more relevant search results by considering the context and intent behind your search terms.
Keyword search finds items by matching their keywords with those in your query. We use the BM25F algorithm to gauge this match. It works by looking at how often a keyword appears in each item's details compared to its frequency in the entire dataset.
BM25F algorithm
The BM25F algorithm is an extension of the BM25 algorithm, which is a popular ranking function used in information retrieval systems such as search engines. BM25F stands for "Best Matching 25 with Field," and it is specifically designed to handle structured documents with multiple fields or attributes.
"apple pie recipes."
When we search for "apple pie recipes," the search engine finds web pages and articles containing those keywords, giving you a list of relevant results to explore.
A hybrid search blends the outcomes of a vector search and a keyword search. It works by doing both searches and then merging the results using a 'fusion' algorithm.
The fusion algorithm
that merges results from different search methods like vector and keyword searches. It determines how to combine and present these results to offer the most relevant information.
"best smartphones"
In a hybrid search for "best smartphones," the fusion algorithm combines vector and keyword search results to prioritize highly rated smartphones closely matching the query. This yields a curated list of top-rated smartphones that align with the user's preferences and search intent.