Requêtes approximatives et avec caractères génériques

Imagine searching for 'apple' but typing 'aple'. Traditional exact searches fail! This lesson introduces techniques to make your search more forgiving and powerful.

We'll explore how to handle typos and search for patterns, ensuring users find what they need even with slight inaccuracies.

Fuzzy search helps you find documents that are 'similar' to your search term, even if there are minor spelling mistakes or variations.

It's incredibly useful for:

• Correcting user typos
• Matching variations in spelling
• Improving user experience by being more tolerant

Elasticsearch's fuzzy matching is often based on the Levenshtein distance algorithm. This algorithm measures the minimum number of single-character edits (insertions, deletions, or substitutions) required to change one word into the other.

For example, the Levenshtein distance between 'apple' and 'aple' is 1 (one deletion). Between 'apple' and 'apply' is 2 (one substitution, one insertion).

You can add fuzziness to a match query. The fuzziness parameter controls the maximum Levenshtein distance allowed.

Try running this example. It will find 'apple' even if you search for 'aple':

The fuzziness parameter can be set to AUTO (recommended) or a number (0, 1, or 2).

• AUTO: Elasticsearch calculates the allowed edit distance based on the term's length. Shorter terms allow fewer edits.
• prefix_length: You can specify a number of initial characters that must exactly match. This can improve performance and relevance.

Wildcard queries allow you to search for patterns within text using special characters. They're useful when you know only part of a term or want to match a family of terms.

Unlike fuzzy queries that correct typos, wildcards help you broaden your search based on specific patterns.

Wildcard queries use two main operators:

• * (asterisk): Matches zero or more characters. For example, appl* would match 'apple', 'application', 'appliance'.
• ? (question mark): Matches any single character. For example, appl? would match 'apply' but not 'apple'.

Let's see a wildcard query in practice. This query will find documents where the product_code field starts with 'ABC' and has any characters following it.

Remember, wildcard queries are typically case-sensitive on keyword fields and can be slow on text fields.

While powerful, wildcard queries, especially those with leading wildcards (e.g., *term), can be computationally expensive and slow.

• They don't use the inverted index efficiently.
• They have to scan many terms to find matches.
• Consider using match_phrase_prefix or completion suggesters for 'autocomplete' type functionality instead.

Choosing between fuzzy and wildcard depends on your goal:

• Fuzzy Queries: Best for handling minor typos and spelling variations. Ideal when you expect a close but not exact match.
• Wildcard Queries: Best for pattern matching and when you know parts of a term but not the whole thing. Be mindful of performance, especially with leading wildcards.

Which of the following statements are TRUE regarding fuzzy and wildcard queries in Elasticsearch?

You've mastered two powerful techniques for more flexible search:

• Fuzzy Queries: Leverage the Levenshtein distance to find matches despite typos, using fuzziness (e.g., AUTO).
• Wildcard Queries: Search for patterns using * (zero or more chars) and ? (single char). Use with caution due to potential performance impacts.

These methods make your search more fault-tolerant and user-friendly!