contrib: improve search accuracy for names, funders, affiliations

inveniosoftware · Nov 26, 2024 · 17d4bf8 · 17d4bf8
1 parent 1cd09ef
commit 17d4bf8
Show file tree

Hide file tree

Showing 5 changed files with 51 additions and 42 deletions.
diff --git a/invenio_vocabularies/contrib/affiliations/config.py b/invenio_vocabularies/contrib/affiliations/config.py
@@ -13,7 +13,9 @@
 from invenio_i18n import lazy_gettext as _
 from invenio_records_resources.services import SearchOptions
 from invenio_records_resources.services.records.components import DataComponent
-from invenio_records_resources.services.records.params import SuggestQueryParser
+from invenio_records_resources.services.records.queryparser import (
+    CompositeSuggestQueryParser,
+)
 from werkzeug.local import LocalProxy
 
 from ...services.components import PIDComponent
@@ -24,23 +26,29 @@
 affiliation_edmo_country_mappings = LocalProxy(
     lambda: current_app.config["VOCABULARIES_AFFILIATIONS_EDMO_COUNTRY_MAPPING"]
 )
-localized_title = LocalProxy(lambda: f"title.{get_locale()}^20")
+localized_title = LocalProxy(lambda: f"title.{get_locale()}^7")
 
 
 class AffiliationsSearchOptions(SearchOptions):
     """Search options."""
 
-    suggest_parser_cls = SuggestQueryParser.factory(
+    suggest_parser_cls = CompositeSuggestQueryParser.factory(
         fields=[
-            "name^100",
-            "acronym.keyword^100",
-            "acronym^40",
+            # We boost the acronym fields, since they're smaller words and are more
+            # likely to be used in a query.
+            "acronym.keyword^50",
+            "acronym^10",
+            "name^10",
+            # Aliases can sometimes be shorter, so we boost them a bit.
+            "aliases^5",
             localized_title,
-            "id^20",
-            "aliases^20",
+            "id^2",
+            # Allow to search identifiers directly (e.g. ROR)
+            "identifiers.identifier",
+            "country",
+            "country_name",
+            "types",
         ],
-        type="most_fields",  # https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-multi-match-query.html#multi-match-types
-        fuzziness="AUTO",  # https://www.elastic.co/guide/en/elasticsearch/reference/current/common-options.html#fuzziness
     )
 
     sort_default = "bestmatch"

diff --git a/invenio_vocabularies/contrib/funders/config.py b/invenio_vocabularies/contrib/funders/config.py
@@ -13,7 +13,9 @@
 from invenio_i18n import lazy_gettext as _
 from invenio_records_resources.services import SearchOptions
 from invenio_records_resources.services.records.components import DataComponent
-from invenio_records_resources.services.records.params import SuggestQueryParser
+from invenio_records_resources.services.records.queryparser import (
+    CompositeSuggestQueryParser,
+)
 from werkzeug.local import LocalProxy
 
 from ...services.components import ModelPIDComponent
@@ -23,24 +25,29 @@
 funder_fundref_doi_prefix = LocalProxy(
     lambda: current_app.config["VOCABULARIES_FUNDER_DOI_PREFIX"]
 )
-localized_title = LocalProxy(lambda: f"title.{get_locale()}^20")
+localized_title = LocalProxy(lambda: f"title.{get_locale()}^7")
 
 
 class FundersSearchOptions(SearchOptions):
     """Search options."""
 
-    suggest_parser_cls = SuggestQueryParser.factory(
+    suggest_parser_cls = CompositeSuggestQueryParser.factory(
         fields=[
-            "name^100",
-            "acronym.keyword^100",
-            "acronym^40",
+            # We boost the acronym fields, since they're smaller words and are more
+            # likely to be used in a query.
+            "acronym.keyword^50",
+            "acronym^10",
+            "name^10",
+            # Aliases can sometimes be shorter, so we boost them a bit.
+            "aliases^5",
             localized_title,
-            "id^20",
-            "aliases^20",
-            "identifiers.identifier^10",
-        ],
-        type="most_fields",  # https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-multi-match-query.html#multi-match-types
-        fuzziness="AUTO",  # https://www.elastic.co/guide/en/elasticsearch/reference/current/common-options.html#fuzziness
+            "id^2",
+            # Allow to search identifiers directly (e.g. ROR)
+            "identifiers.identifier",
+            "country",
+            "country_name",
+            "types",
+        ]
     )
 
     sort_default = "bestmatch"

diff --git a/invenio_vocabularies/contrib/names/config.py b/invenio_vocabularies/contrib/names/config.py
@@ -15,7 +15,9 @@
     DataComponent,
     RelationsComponent,
 )
-from invenio_records_resources.services.records.params import SuggestQueryParser
+from invenio_records_resources.services.records.queryparser import (
+    CompositeSuggestQueryParser,
+)
 from werkzeug.local import LocalProxy
 
 from ...services.components import PIDComponent
@@ -26,16 +28,17 @@
 class NamesSearchOptions(SearchOptions):
     """Search options."""
 
-    suggest_parser_cls = SuggestQueryParser.factory(
+    suggest_parser_cls = CompositeSuggestQueryParser.factory(
         fields=[
-            "given_name^100",
-            "name^70",
-            "family_name^50",
-            "identifiers.identifier^20",
-            "affiliations.name^20",
+            "name^5",
+            # We boost the affiliation acronym fields, since they're short and more
+            # likely to be used in a query.
+            "affiliations.acronym.keyword^3",
+            "affiliations.acronym",
+            "affiliations.name",
+            # Allow to search identifiers directly (e.g. ORCID)
+            "identifiers.identifier",
         ],
-        type="most_fields",  # https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-multi-match-query.html#multi-match-types
-        fuzziness="AUTO",
     )
 
     sort_default = "bestmatch"

diff --git a/invenio_vocabularies/services/config.py b/invenio_vocabularies/services/config.py
@@ -14,17 +14,11 @@
 from invenio_i18n import lazy_gettext as _
 from invenio_records_resources.services import (
     Link,
-    LinksTemplate,
-    RecordService,
     RecordServiceConfig,
     SearchOptions,
     pagination_links,
 )
-from invenio_records_resources.services.base import (
-    ConditionalLink,
-    Service,
-    ServiceListResult,
-)
+from invenio_records_resources.services.base import ConditionalLink
 from invenio_records_resources.services.records.components import DataComponent
 from invenio_records_resources.services.records.params import (
     FilterParam,

diff --git a/tests/contrib/names/test_names_resource.py b/tests/contrib/names/test_names_resource.py
@@ -10,7 +10,6 @@
 """Test the name vocabulary resource."""
 
 import json
-from copy import deepcopy
 
 import pytest
 
@@ -193,8 +192,6 @@ def test_names_suggest_sort(client_with_credentials, example_multiple_names, h,
     # With affiliation
     res = client_with_credentials.get(f"{prefix}?suggest=john%20wwe", headers=h)
     assert res.status_code == 200
-    assert (
-        res.json["hits"]["total"] == 3
-    )  # Will find 3 johns but WWE affiliation should be at the top
+    assert res.json["hits"]["total"] == 1
     assert res.json["hits"]["hits"][0]["name"] == "Cena, John"
     assert res.json["hits"]["hits"][0]["affiliations"][0]["name"] == "WWE"