Class: ValkeyObjects::KNN
- Inherits:
-
Object
- Object
- ValkeyObjects::KNN
- Defined in:
- lib/valkey/objects.rb
Overview
generic KNN processor
Instance Method Summary collapse
- #[](q) ⇒ Object
-
#hood(query) ⇒ Object
Find k nearest neighbors to the query string Returns array of hashes with :string and :distance keys.
-
#initialize(collection) ⇒ KNN
constructor
Initialize with a collection of strings.
- #rank(query) ⇒ Object
Constructor Details
#initialize(collection) ⇒ KNN
Initialize with a collection of strings
79 80 81 |
# File 'lib/valkey/objects.rb', line 79 def initialize(collection) @collection = collection.flatten.compact end |
Instance Method Details
#[](q) ⇒ Object
106 107 108 |
# File 'lib/valkey/objects.rb', line 106 def [] q rank(q)[0][:string] end |
#hood(query) ⇒ Object
Find k nearest neighbors to the query string Returns array of hashes with :string and :distance keys
85 86 87 88 89 90 91 92 93 |
# File 'lib/valkey/objects.rb', line 85 def hood(query) # Calculate distances for all strings distances = @collection.map do |str| { string: str, distance: levenshtein_distance(query, str) } end # Sort by distance (ascending) and take top k distances.sort_by { |item| item[:distance] }.take([10, @collection.size].min) end |
#rank(query) ⇒ Object
95 96 97 98 99 100 101 102 103 104 |
# File 'lib/valkey/objects.rb', line 95 def rank query neighbors = hood(query) neighbors.map do |item| max_len = [query.length, item[:string].length].max similarity = max_len.zero? ? 1.0 : 1.0 - (item[:distance].to_f / max_len) dist = levenshtein_distance(item[:string], query) { string: item[:string], similarity: similarity.round(4), distance: dist.round(4) } end.sort_by { |h| h[:distance] && -h[:similarity] } end |