}
Searching with advanced query syntax
You can create queries that are more specific than simple keyword or natural language queries by using advanced query syntax or operators. This includes ranges, Booleans, wildcards, and more. By using operators, you can give your query more context and further refine the search results.
Amazon Kendra supports the following operators.
• Boolean: Logic to limit or broaden the search. For example, amazon AND sports limits the search to only search for documents containing both terms.
• Parentheses: Reads nested query terms in order of precedence. For example, (amazon AND sports) NOT rainforest reads (amazon AND sports) before NOT rainforest.
• Ranges: Date or numeric range values. Ranges can be inclusive, exclusive, or unbounded. For example, you can search for documents that were last updated between January 1st 2020 and December 31st 2020, inclusive of these dates.
• Fields: Uses a specific field to limit the search. For example, you can search for documents that have 'United States' in the field 'location'.
• Wildcards: Partially match a string of text. For example, Cloud* could match CloudFormation. Amazon Kendra currently only supports trailing wildcards.
You can use a combination of any of the above operators.
Note that excessive use of operators or highly complex queries could impact query latency. Wildcards are some of the most expensive operators in terms of latency. A general rule is the more terms and operators that you use, the greater potential impact on latency. Other factors that affect latency include the average size of documents indexed, the size of your index, any filtering on search results, and the overall load on your Amazon Kendra index.
Boolean
You can combine or exclude words using the Boolean operators AND, OR, NOT.
The following are examples of using Boolean operators.
amazon AND sports
Returns search results that contain both the terms 'amazon' and 'sports' in the text, such as Amazon Prime video sports or other similar content.
sports OR recreation
Returns search results that contain the terms 'sports' or 'recreation', or both, in the text.
amazon NOT rainforest
Returns search results that contain the term 'amazon' but not the term 'rainforest' in the text. This is to search for documents about the company Amazon, not the Amazon Rainforest.
Parentheses
You can query nested words in order of precedence by using parentheses. The parentheses indicate to Amazon Kendra how a query should be read.
The following are examples of using parentheses operators.
Searching with advanced query syntax
(amazon AND sports) NOT rainforest
Returns search results that contain both the terms 'amazon' and 'sports' in the text, but not the term 'rainforest'. This is to search Amazon Prime video sports or other similar content, not adventure sports in the Amazon Rainforest. The parentheses help indicate that amazon AND sports should be read before NOT rainforest. The query should not be read as amazon AND (sports NOT rainforest).
(amazon AND (sports OR recreation)) NOT rainforest
Returns search results that contain the terms 'sports' or 'recreation', or both, and the term 'amazon'.
But it does not include the term 'rainforest'. This is to search Amazon Prime video sports or recreation, not adventure sports in the Amazon Rainforest. The parentheses help indicate that sports OR recreation should be read before combining with 'amazon', which is read before NOT rainforest.
The query should not be read as amazon AND (sports OR (recreation NOT rainforest)).
Ranges
You can use a range of values to filter the search results. You specify an attribute and the range values.
This can be date or numeric type.
Date ranges must be in the following formats:
• Epoch
• YYYY
• YYYY-mm
• YYYY-mm-dd
• YYYY-mm-dd'T'HH
You can also specify whether to include or exclude the lower and higher values of the range.
The following are examples of using range operators.
_processed_date:>2019-12-31 AND _processed_date:<2021-01-01
Returns documents that were processed in 2020—greater than December 31st 2019 and less than January 1st 2021.
_processed_date:>=2020-01-01 AND _processed_date:<=2020-12-31
Returns documents that were processed in 2020—greater than or equal to January 1st 2020 and less than or equal to December 31st 2020.
_document_likes:<1
Returns documents with zero likes or no user feedback—less than 1 like.
You can specify whether a range should be treated as inclusive or exclusive of the given range values.
Inclusive
_last_updated_at:[2020-01-01 TO 2020-12-31]
Returns documents last updated in 2020—includes the days December 1st 2020 and December 31st 2020.
Exclusive
_last_updated_at:{2019-12-31 TO 2021-01-01}
Returns documents last updated in 2020—excludes the days December 31st 2019 and January 1st 2021.
Searching with advanced query syntax
For unbounded ranges that are neither inclusive or exclusive, simply use the < and > operators. For example, _last_updated_at:>2019-12-31 AND _last_updated_at:<2021-01-01
Fields
You can limit your search to only return documents that meet a value in a specific field. The field can be of any type.
The following are examples of using field-level context operators.
status:"Incomplete" AND financial_year:2021
Returns documents for the 2021 financial year with their status as incomplete.
(sports OR recreation) AND country:"United States" AND level:"professional"
Returns documents that discuss professional sports or recreation in the United States.
Wildcards
You can broaden your search to account for variants of words and phrases using the wildcard operator.
This is useful when searching for name variants. Amazon Kendra currently only supports trailing wildcards. The number of prefix characters for a trailing wildcard must be greater than two.
The following are examples of using wildcard operators.
Cloud*
Returns documents that contain variants such as CloudFormation and CloudWatch.
kendra*aws
Returns documents that contain variants such as kendra.amazonaws.
kendra*aws*
Returns documents that contain variants such as kendra.amazonaws.com
Invalid query syntax
Amazon Kendra issues a warning if there are problems with your query syntax or your query is currently not supported by Amazon Kendra. For more information, see the API documentation for query warnings.
The following queries are examples of invalid query syntax.
_last_updated_at:<2021-12-32
Invalid date. Day 32 does not exist in the Gregorian calendar, which is used by Amazon Kendra.
_view_count:ten
Invalid numeric value. Digits must be used to represent numeric values.
nonExistentField:123
Invalid field search. The field must exist in order to use field search.
Product:[A TO D]
Invalid range. Numeric values or dates must be used for ranges.
OR Hello
Invalid Boolean. Operators must be used with terms and placed between terms.