Synonyms

If multiple words have an equivalent meaning in your dataset, you can create a list of synonyms. This will make your search results more relevant.

Words set as synonyms won't always return the same results. With the default settings, the movies dataset should return 547 results for great and 66 for fantastic. Let's set them as synonyms:

curl \
  -X PUT 'http://localhost:7700/indexes/movies/settings/synonyms' \
  -H 'Content-Type: application/json' \
  --data-binary '{
    "great": ["fantastic"], "fantastic": ["great"]
  }'

With the new settings, searching for great returns 595 results and fantastic returns 423 results. This is due to various factors like typos and splitting the query to find relevant documents. The search for great will allow only one typo (for example, create) and take into account all variations of great (for instance, greatest) along with fantastic.

All synonyms are lowercased and de-unicoded during the indexing process.

Consider a situation where Résumé and CV are set as synonyms.

{
  "Résumé": [
    "CV"
  ],
  "CV": [
    "Résumé"
  ]
}

A search for cv would return any documents containing cv or CV, in addition to any that contain Résumé, resumé, resume, etc., unaffected by case or accent marks.

Use this when you want one word to be synonymous with another, but not the other way around.

phone => iphone

A search for phone will return documents containing iphone as if they contained the word phone.

However, if you search for iphone, documents containing phone will be ranked lower in the results due to the typo rule.

To create a one-way synonym list, this is the JSON syntax that should be added to the settings.

{
  "phone": [
    "iphone"
  ]
}

The exact search query will always take precedence over its synonyms. The exactness ranking rule favors exact words over synonyms when ranking search results.

Taking the following set of search results:

[
  {
    "id": 0,
    "title": "Ghouls 'n Ghosts"
  },
  {
    "id": 1,
    "title": "Phoenix Wright: Spirit of Justice"
  }
]

If you configure ghost as a synonym of spirit, queries searching for spirit will return document 1 before document 0.

By associating one or more synonyms with each other, they will be considered the same in both directions.

shoe <=> boot <=> slipper <=> sneakers

When a search is done with one of these words, all synonyms will be considered as the same word and will appear in the search results.

To create a mutual association between four words, this is the JSON syntax that should be added to the settings.

{
  "shoe": [
    "boot",
    "slipper",
    "sneakers"
  ],
  "boot": [
    "shoe",
    "slipper",
    "sneakers"
  ],
  "slipper": [
    "shoe",
    "boot",
    "sneakers"
  ],
  "sneakers": [
    "shoe",
    "boot",
    "slipper"
  ]
}

Meilisearch treats multi-word synonyms as phrases.

Suppose you set San Francisco and SF as synonyms with a mutual association

{
  "san francisco": [
    "sf"
  ],
  "sf": [
    "san francisco"
  ]
}

If you input SF as a search query, Meilisearch will also return results containing the phrase San Francisco. However, depending on the ranking rules, they might be considered less relevant than those containing SF. The reverse is also true: if your query is San Francisco, documents containing San Francisco may rank higher than those containing SF.

A single term may have up to 50 synonyms. Meilisearch silently ignores any synonyms beyond this limit. For example, if you configure 51 synonyms for book, Meilisearch will only return results containing the term itself and the first 50 synonyms.

If any synonyms for a term contain more than one word, the sum of all words across all synonyms for that term cannot exceed 100 words. Meilisearch silently ignores any synonyms beyond this limit. For example, if you configure 40 synonyms for computer in your application, taken together these synonyms must contain fewer than 100 words.