Multi-language Loomi Search+

This guide introduces the Multi-language Loomi Search+ feature.

Prerequisite knowledge

We recommend that you review the following guides:

• Loomi Search+: This guide introduces Loomi Search+ and the key vector search concepts.
• Recall studio: This guide helps get familiar with the Recall studio application.

What is Multi-language Loomi Search+

Multi-language Loomi Search+ extends existing Loomi Search+ capabilities to interpret queries in all supported languages.

This feature uses a powerful retrieval process that combines semantic search with advanced machine learning techniques to improve product discovery across different query types.

Multi-language Loomi Search+ brings AI-powered search capabilities to global brands requiring diverse languages and localized results while maintaining search quality and relevance.

Language support

Multi-language Loomi Search+ supports 38 languages with dedicated vector models. Each language model is individually tuned to ensure optimal search performance and relevance.

Europe: Albanian, Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Norwegian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swedish, Ukrainian

Middle East: Arabic, Hebrew, Turkish

Asia: Chinese (Simplified), Chinese (Traditional), Indonesian, Japanese, Korean, Thai

Feature benefits

Multi-language Loomi Search+ delivers all standard Loomi Search+ benefits while adding specific value for global ecommerce:

Scale AI-powered search precision globally

Deploy consistent search quality across all supported languages with minimal tuning required for great performance and shopper experience. Close localization gaps between regional stores.

Optimize results for ecommerce performance

Built on Bloomreach's commerce-trained models, Multi-language Loomi Search+ optimizes results to:

• Improve revenue per visitor (RPV) performance with more accurate search intent understanding, handling of long-tail and natural language queries.
• Reduce null results due to better contextual understanding.
• Display more relevant search results using the best models for product relevance, combining LLM-based and semantic search.

Business use cases

Multi-language Loomi Search+ is typically a great fit if you want to:

• Improve query understanding of intent and compound words.
• Reduce null results.
• Better handle long tail queries and diacritics.
• Handle search transliteration in multiple languages.

Multi-language Loomi Search+ isn't optimized for precision-focused use cases, but can provide improved results using high vector temperature settings. Consider the following examples illustrated using Product grid insights:

Improve query understanding

A fashion brand launching in European markets uses Multi-language Loomi Search+ to serve shoppers in French, German, Italian, and Spanish without building a separate search infrastructure.

Shoppers searching for "robe d'été" find summer dresses with the same accuracy as English "summer dress" queries.

Reduce null results

A furniture retailer expands to sell to German shoppers, but finds 40% of German searches return zero results. Queries like "Almirah-Aufbewahrungsschrank aus geschnitztem Holz und Spiegel" (Carved Wood and Mirror Almirah Storage Cabinet) earlier failed with a keyword match.

Multi-language Loomi Search+ interprets and matches such queries to relevant "Almirah Storage Cabinet" products, cutting null results and recovering potential revenue.

Better handle long tail queries and diacritics

A global retailer launches a multilingual e-commerce site targeting the Polish market. Their Polish shoppers often search for "roza na pniu" (rose tree) without proper Polish accents (róża na pniu).

Meanwhile, German shoppers use highly specific, multi-word phrases like “dunkelblauer Wintermantel mit Kapuze und Gürtel” (dark blue winter coat with hood and belt).

With Multi-language Search+, these queries, whether long-tail, missing diacritics, or combining compound words, produce relevant results. This fastens product discovery and improves conversion rates.

Handle search transliteration in multiple languages

A retailer has an English product catalog, but occasionally receives French searches that return zero results. Queries like "armoire en bois" (wooden cabinet) showed no results without Multi-language Loomi Search+.

When toggled ON, Multi-language Loomi Search+ transliterates and matches such queries to relevant "wooden cabinet" products, opening new conversion opportunities.

How does Multi-language Loomi Search+ work

Like Loomi Search+, Multi-language Loomi Search+ combines semantic keyword matching with LLM-powered vector search to understand different query types across languages.

This means shoppers will find relevant running shoes with similar precision as English queries searching for "chaussures de course" in French or "ランニングシューズ" in Japanese..

While Loomi search+ combines keywords and vector search to find the most relevant products, its dynamic ranking ensures the products most likely to sell reach the top. Review the complete search mechanism.

Key features and controls

Multi-language Loomi Search+ provides the same configuration options as Loomi Search+ through Recall studio, where you can toggle it ON per catalog and adjust recall settings per language.

Recall algorithm controls

Multi-language Search+ supports the following keyword precision modes:

• Text match
• Product type
• Category

LLM-based precision is not yet supported. You can also configure query relaxation, spell correction, and facet precision per language.

📘

Note

Multi-language Loomi Search+ has the same API controls as Loomi Search+. The API response includes information on when Multi-language Loomi Search+ relaxes a query. You can display this to the shopper to avoid confusion when they see related products instead of exact matches.

Vector search temperature

Multi-language Search+ offers different temperature settings optimized per language. Work with your Bloomreach representative to tune the vector temperature and adjust the precision as needed.

Access and permissions

The feature is available to all Premium package customers via Recall studio. To access Recall studio, ensure that you have the Recall and ranking studio editor permission.

Feed requirements

We strongly recommend having a single language catalog with searchable fields in one of the supported languages. Avoid multilingual catalog data.

FAQs

How does Multi-language Loomi Search+ differ from Multi-language semantic search?

Multi-language Loomi Search+ uses a search model that combines keyword search and LLM–based vector search capabilities, whereas Multi-language semantic search uses a semantic model to understand queries.

Can I A/B test Multi-language Loomi Search+?

Yes, you can A/B test Multi-language Loomi Search+ the same as Loomi Search+.

What happens if a shopper searches in an unsupported language?

The model relaxes the search. Typically, it lowers the vector temperature to show search results. The vector models use the transliteration capabilities to capture context from other languages, but you can expect that the results will not be as precise according to the shopper's search intent.

What is the impact of Multi-language Loomi Search+ on my existing merchandising activities?

We suggest optimizing your merchandising activities in the following ways:

• Review existing merchandising rules optimized for keyword search (signals, boosting, burying, slotting).
• Simplify or reduce manual rules, as Loomi Search+ can optimize results more autonomously.
• Use a diligent A/B testing strategy to toggle specific rules ON/OFF and evaluate their effect. Experiment with a combination of Product grid insights and API parameters to see the effect of merchandising rules.