From 5441c3a37552f787474d45ba7b17b97d2c7b2bda Mon Sep 17 00:00:00 2001
From: pasha <ppodolsky@me.com>
Date: Fri, 23 Jun 2023 16:59:34 +0300
Subject: [PATCH] feat: Rule-based noun detection

---
 .../src/components/default_tokenizers.rs      |  2 +-
 summa-core/src/components/mod.rs              |  2 +-
 .../query_parser/morphology/english.rs        | 32 +++++++++++++------
 .../src/components/query_parser/summa_ql.rs   |  4 +--
 4 files changed, 27 insertions(+), 13 deletions(-)

diff --git a/summa-core/src/components/default_tokenizers.rs b/summa-core/src/components/default_tokenizers.rs
index af80df72..065a0e0c 100644
--- a/summa-core/src/components/default_tokenizers.rs
+++ b/summa-core/src/components/default_tokenizers.rs
@@ -3,7 +3,7 @@ use tantivy::tokenizer::{LowerCaser, RawTokenizer, RemoveLongFilter, SimpleToken
 use super::summa_tokenizer::SummaTokenizer;
 
 /// List of stop words mixed for multiple languages
-const STOP_WORDS: [&str; 318] = [
+pub const STOP_WORDS: [&str; 318] = [
     "a",
     "an",
     "and",
diff --git a/summa-core/src/components/mod.rs b/summa-core/src/components/mod.rs
index e3c0c623..cecdcb37 100644
--- a/summa-core/src/components/mod.rs
+++ b/summa-core/src/components/mod.rs
@@ -13,7 +13,7 @@ mod summa_document;
 mod summa_tokenizer;
 
 pub use custom_serializer::NamedFieldDocument;
-pub use default_tokenizers::default_tokenizers;
+pub use default_tokenizers::{default_tokenizers, STOP_WORDS};
 pub use fruit_extractors::{build_fruit_extractor, FruitExtractor, IntermediateExtractionResult};
 pub use index_holder::{cleanup_index, IndexHolder};
 pub use index_registry::IndexRegistry;
diff --git a/summa-core/src/components/query_parser/morphology/english.rs b/summa-core/src/components/query_parser/morphology/english.rs
index f2aeec9a..c8997345 100644
--- a/summa-core/src/components/query_parser/morphology/english.rs
+++ b/summa-core/src/components/query_parser/morphology/english.rs
@@ -1,3 +1,7 @@
+use std::collections::HashSet;
+
+use regex::RegexSet;
+
 use crate::components::query_parser::morphology::Morphology;
 
 #[derive(Default, Clone)]
@@ -5,16 +9,26 @@ pub struct EnglishMorphology {}
 
 impl Morphology for EnglishMorphology {
     fn derive_tenses(&self, word: &str) -> Option<String> {
-        let is_singular = pluralize_rs::is_singular(word);
-        let is_plural = pluralize_rs::is_plural(word);
-
-        if is_singular {
-            Some(pluralize_rs::to_plural(word))
-        } else if is_plural {
-            Some(pluralize_rs::to_singular(word))
-        } else {
-            None
+        thread_local! {
+            static NOT_A_NOUN: (RegexSet, HashSet<&'static str>) = (RegexSet::new(&[
+                r"\d$",
+                r"ing$",
+            ]).expect("cannot compile regex"), HashSet::from_iter(crate::components::default_tokenizers::STOP_WORDS.into_iter()));
         }
+        NOT_A_NOUN.with(|(not_a_noun_regex, stop_words)| {
+            if stop_words.contains(word) || not_a_noun_regex.is_match(word) {
+                return None;
+            }
+            let is_singular = pluralize_rs::is_singular(word);
+            let is_plural = pluralize_rs::is_plural(word);
+            if is_singular {
+                Some(pluralize_rs::to_plural(word))
+            } else if is_plural {
+                Some(pluralize_rs::to_singular(word))
+            } else {
+                None
+            }
+        })
     }
 
     fn detect_ners(&self, _: &str) -> Vec<String> {
diff --git a/summa-core/src/components/query_parser/summa_ql.rs b/summa-core/src/components/query_parser/summa_ql.rs
index e20c16a5..fdc43b07 100644
--- a/summa-core/src/components/query_parser/summa_ql.rs
+++ b/summa-core/src/components/query_parser/summa_ql.rs
@@ -1178,8 +1178,8 @@ mod tests {
         );
         query_parser.query_parser_config.0.morphology_configs = morphology_configs;
         query_parser.query_parser_config.0.query_language = Some("en".to_string());
-        let query = query_parser.parse_query("red search engine");
-        assert_eq!(format!("{:?}", query), "Ok(BooleanQuery { subqueries: [(Should, DisjunctionMaxQuery { disjuncts: [TermQuery(Term(field=0, type=Str, \"red\")), TermQuery(Term(field=0, type=Str, \"reds\"))], tie_breaker: 0.3 }), (Should, TermQuery(Term(field=0, type=Str, \"search\"))), (Should, DisjunctionMaxQuery { disjuncts: [TermQuery(Term(field=0, type=Str, \"engine\")), TermQuery(Term(field=0, type=Str, \"engines\"))], tie_breaker: 0.3 })] })");
+        let query = query_parser.parse_query("red1 search engine going");
+        assert_eq!(format!("{:?}", query), "");
         let query = query_parser.parse_query("iso 34-1:2022");
         assert_eq!(format!("{:?}", query), "Ok(BooleanQuery { subqueries: [(Should, TermQuery(Term(field=0, type=Str, \"iso\"))), (Should, TermQuery(Term(field=0, type=Str, \"34\"))), (Should, TermQuery(Term(field=0, type=Str, \"1\")))] })");
     }