Vectara supports the ability to incorporate partial/exact/Boolean matching in search results and even to blend them with the neural model in what's called a "hybrid" retrieval model.
For example, you can use this in Vectara to:
- Allow Vectara to include exact keyword matches for occasions where a search term was absent from Vectara's training data (e.g. product SKUs)
- Turn off neural retrieval entirely, and instead use exact term matching
- Incorporate typical keyword modifiers like a
NOTfunction, exact phrase matching, and wildcard prefixes of terms
Enable Exact and Boolean Text Matching
By default, the exact and Boolean text matching is disabled and only neural
retrieval is used. You can enable the feature by specifying a value,
lambda, at query time. This value can range from
The default value of
0, thus turning off exact and Boolean text
A value of
1 would turn off neural retrieval instead, relying only on
Boolean and exact text matching. This can be useful if you're trying to
evaluate how a keyword system like one based on Elasticsearch or Solr may
compare to Vectara.
Vectara supports values in between as well, which tells Vectara to try to consider both neural and Boolean and exact text matching and then to blend the scores of the results of the two different scoring models. Users often see best results by setting this value somewhere between 0.01 and 0.1, and we typically recommend users start experimentation with a value of 0.025.
When interpreting query strings, Vectara treats the following syntax specially.
Words that are quoted must match exactly in that order. So, for example, the query "blue shoes" must match the word blue followed immediately by shoes.
A word fragment suffixed with a
* is treated as a prefix match, meaning that
it matches any word of which it is a prefix. For example,
Words prefixed with a
- sign are excluded from the results. So, to extend the
-Mississippi would exclude results referencing the Magnolia
-Miss* would exclude references to both Mississippi and Missouri.