Loading

B2C Commerce: Strict Japanese Search/Index Matching

Publiceringsdatum: Apr 23, 2026
Beskrivning

Overview

Starting from B2C Commerce version 26.4, a new feature switch, "Enable Strict Japanese Search/Index Matching", has been introduced. Enabling this feature resolves the previous issue in Japanese search where the search index and query fail to match for specific phrases.

Background: Previous Index/Query Mismatch Issue

Product search on a site is primarily driven by the following two processes:

  • Indexing: Builds a search index based on the product data in the catalog.
  • Search (Query): Processes the search phrase entered by the user using a Tokenizer, Stemming Processor, etc. to generate the actual query used for searching.
    • Tokenizer: Splits the search phrase into tokens.
    • Stemming Processor: Converts conjugated verb forms into their base forms and removes particles that only serve grammatical functions (such as て, に, を, は, etc.) from the split tokens, among other things.

In previous Japanese search behavior, while there were no issues during index creation, a problem occurred during query generation where the stemming processor would execute conversion processes without considering the surrounding context.

Example of the Problem: "Koinobori"

Suppose there is a product in the catalog named "こいのぼり 小物" (Koinobori Komono).

  • Index: Created as こい | のぼり | 小物.
  • Search Processing: When a user searched for "こいのぼり" (koinobori), the following incorrect processing occurred:
    • Segmentation: こいのぼり → こい | のぼり
    • Conversion:
      • こい → こい
      • のぼり → のぼる (noboru)
    • Note: Because the surrounding context was not considered, "nobori" was recognized as a conjugated form of the verb "noboru" (to climb) rather than the noun "nobori" (banner), resulting in an incorrect conversion.
    • Generated Query: こい | のぼる

Because of this, a discrepancy arose between the index (のぼり) and the query (のぼる). This caused an issue where the product would not appear in the search results, even though the user's search phrase perfectly matched a part of the product name.

Lösning

This feature can be enabled from the following menu in Business Manager:

  • Location: Administration > Global Preferences > Feature Switches
  • Name: Enable Strict Japanese Search/Index Matching
  • Note: Feature switch settings are not subject to data replication. Please first enable the feature and verify its behavior in the Staging or Development instance, and then manually enable it in the Business Manager of your Production instance.

The Effect of Strict Japanese Search/Index Matching

When this feature is enabled, token splitting and stemming processor operations during searches will correctly execute while taking context into account. This guarantees that when a search phrase perfectly matches a portion of a product's searchable attributes, the product will correctly appear in the search results.

Example of Improvement: "Koinobori"

Because the stemming processor contextually recognizes "nobori" as part of the phrase "koinobori," it will no longer incorrectly convert it to "noboru."

  • Segmentation: こいのぼり → こい | のぼり
  • Conversion: 
    • こい → こい
    • のぼり → のぼり
  • Generated Query: こい | のぼり

As a result, the query and the index perfectly match, and searching the phrase "koinobori" will correctly display the "Koinobori Komono" product in the results.

Other Improvement Examples

In addition to "koinobori," there are several other phrases that previously suffered from incorrect conversion processing in Japanese search. By enabling this feature, correct queries will be generated for these as well:

Search PhrasePrevious Incorrect QueryCorrect Query (After Enabling Feature)
多機能ペン (Multifunction pen)多い | 機能 | ペン多 | 機能 | ペン
二重ガーゼ (Double gauze)二 | 重い | ガーゼ二 | 重 | ガーゼ
9枚入り (Pack of 9)9 | 枚 | 入る9 | 枚 | 入り

Regarding Search Dictionaries (Synonyms, Stemming Exceptions, etc.)

Previously, the index/query mismatch issue could sometimes be worked around by configuring search dictionaries, such as adding synonyms or stemming exceptions.

Example: Registering "のぼり, こいのぼり" as synonyms, or adding "のぼり" as a stemming exception.

Once this feature is enabled, these workaround search dictionary entries become unnecessary. While leaving the settings as-is should not hinder search operations, it is recommended from an administrative perspective to delete these redundant entries in conjunction with enabling the feature.

If your changes involve deleting dictionary entries, a data replication (rebuilding the search index) will be required. Please use one of the following procedures in such cases:

Pattern A: Enabling the feature and deleting unnecessary search dictionaries simultaneously

  1. (Staging) Enable the feature.
  2. (Staging) Modify the search dictionaries as needed (e.g., delete synonyms and stemming exceptions that are no longer needed).
  3. (Staging) Rebuild the search index.
  4. (Staging) Verify the search behavior.
  5. (Production) Enable the feature.
  6. (Staging → Production) Perform data replication, including the "Search Indexes".

Pattern B: Enabling the feature first and gradually deleting unnecessary search dictionaries later

  1. (Staging) Enable the feature.
  2. (Staging) Verify the search behavior.
  3. (Production) Enable the feature.
  4. (Staging) At a later date: Modify the search dictionaries as needed (e.g., delete synonyms and stemming exceptions that are no longer needed).
  5. (Staging) Rebuild the search index.
  6. (Staging) Verify the search behavior.
  7. (Staging → Production) Perform data replication, including the "Search Indexes".

Note: Further Enhancing Search Accuracy

To further improve the quality of Japanese search, we recommend using "Japanese - improved" stemming in addition to enabling the "Strict Japanese Search/Index Matching" feature. For more details, please refer to the knowledge article below:

B2C Commerce: Japanese Stemming Options

Knowledge-artikelnummer

005316843

 
Laddar
Salesforce Help | Article