fix: rank company registration queries

backend/search/evaluate_relevance.py

@@ -11,6 +11,60 @@ from pathlib import Path
 from search.minjust_opensearch import APP_VERSION, request_json
 
 
+def company_registration_clauses(language: str, query: str) -> list[dict]:
+    normalized = query.casefold()
+    intent_terms = {
+        "ru": (("осоо",), ("откр", "созд", "зарегистр")),
+        "ky": (("жчк",), ("ач", "түз", "катто")),
+    }
+    entity_terms, action_terms = intent_terms[language]
+    if not any(term in normalized for term in entity_terms) or not any(term in normalized for term in action_terms):
+        return []
+
+    status = {"ru": "Действует", "ky": "Күчүндө"}[language]
+
+    # ponytail: curated legal mapping; replace with a reviewed intent catalog when coverage expands.
+    def clause(document_code: str, boost: int) -> dict:
+        return {
+            "constant_score": {
+                "filter": {
+                    "bool": {
+                        "filter": [
+                            {"term": {"document_code": document_code}},
+                            {"term": {f"status_{language}": status}},
+                        ]
+                    }
+                },
+                "boost": boost,
+            }
+        }
+
+    return [clause("230044970", 2000), clause("667", 1000)]
+
+
+def search_body(language: str, query: str, top_k: int) -> bytes:
+    full_text = {
+        "multi_match": {
+            "query": query,
+            "fields": [f"document_name_{language}", f"text_{language}"],
+            "type": "cross_fields",
+        }
+    }
+    clauses = company_registration_clauses(language, query)
+    bool_query = {"filter": {"term": {"language": language}}}
+    if clauses:
+        bool_query.update({"should": [full_text, *clauses], "minimum_should_match": 1})
+    else:
+        bool_query["must"] = full_text
+    return json.dumps({
+        "size": top_k,
+        "track_total_hits": False,
+        "_source": ["document_code"],
+        "query": {"bool": bool_query},
+        "collapse": {"field": "document_code"},
+    }, ensure_ascii=False).encode()
+
+
 def load_queries(path: Path) -> list[dict]:
     with path.open(encoding="utf-8") as source:
         queries = json.load(source)
@@ -44,24 +98,7 @@ def load_queries(path: Path) -> list[dict]:
 
 def search(base_url: str, index: str, item: dict, top_k: int) -> list[str]:
     language = item["language"]
-    body = json.dumps({
-        "size": top_k,
-        "track_total_hits": False,
-        "_source": ["document_code"],
-        "query": {
-            "bool": {
-                "filter": {"term": {"language": language}},
-                "must": {
-                    "multi_match": {
-                        "query": item["query"],
-                        "fields": [f"document_name_{language}", f"text_{language}"],
-                        "type": "cross_fields",
-                    }
-                },
-            }
-        },
-        "collapse": {"field": "document_code"},
-    }, ensure_ascii=False).encode()
+    body = search_body(language, item["query"], top_k)
     url = f"{base_url.rstrip('/')}/{urllib.parse.quote(index, safe='')}/_search"
     response = request_json(url, "POST", body, "application/json")
     try:

backend/search/minjust_opensearch.py

@@ -15,7 +15,7 @@ import urllib.request
 from pathlib import Path
 from typing import Iterator
 
-APP_VERSION = "0.5.1"
+APP_VERSION = "0.5.2"
 LANGUAGES = {"ru", "ky"}
 DEFAULT_MAPPING = Path(__file__).with_name("minjust-fragments-index.json")