Improve sudachi settings

app/chewy/accounts_index.rb

@@ -17,6 +17,16 @@ class AccountsIndex < Chewy::Index
         type: 'stemmer',
         language: 'possessive_english',
       },
+
+      my_posfilter: {
+        type: 'sudachi_part_of_speech',
+        stoptags: [
+          '助詞',
+          '助動詞',
+          '補助記号,句点',
+          '補助記号,読点',
+        ],
+      },
     },
 
     analyzer: {
@@ -32,6 +42,15 @@ class AccountsIndex < Chewy::Index
         ),
       },
 
+      sudachi_analyzer: {
+        filter: %w(
+          my_posfilter
+          sudachi_normalizedform
+        ),
+        type: 'custom',
+        tokenizer: 'sudachi_tokenizer',
+      },
+
       verbatim: {
         tokenizer: 'standard',
         filter: %w(lowercase asciifolding cjk_width),
@@ -49,6 +68,13 @@ class AccountsIndex < Chewy::Index
         min_gram: 1,
         max_gram: 15,
       },
+
+      sudachi_tokenizer: {
+        resources_path: '/etc/elasticsearch/sudachi',
+        split_mode: 'A',
+        type: 'sudachi_tokenizer',
+        discard_punctuation: 'true',
+      },
     },
   }
 
@@ -63,6 +89,6 @@ class AccountsIndex < Chewy::Index
     field(:domain, type: 'keyword', value: ->(account) { account.domain || '' })
     field(:display_name, type: 'text', analyzer: 'verbatim') { field :edge_ngram, type: 'text', analyzer: 'edge_ngram', search_analyzer: 'verbatim' }
     field(:username, type: 'text', analyzer: 'verbatim', value: ->(account) { [account.username, account.domain].compact.join('@') }) { field :edge_ngram, type: 'text', analyzer: 'edge_ngram', search_analyzer: 'verbatim' }
-    field(:text, type: 'text', analyzer: 'whitespace', value: ->(account) { account.searchable_text }) { field :stemmed, type: 'text', analyzer: 'natural' }
+    field(:text, type: 'text', analyzer: 'sudachi_analyzer', value: ->(account) { account.searchable_text }) { field :stemmed, type: 'text', analyzer: 'natural' }
   end
 end