## Process Overview This document breaks down the complete process for deploying multimodal search capabilities, covering four core modules: **Text Search Configuration**, **Image Search Configuration**, **Result Ranking**, and **Search Term Optimisation**. Each step is accompanied by a video tutorial. We recommend following the videos in sequence to efficiently build and optimise your multimodal search engine to meet your specific business requirements. ## 1. Text Search Configuration & Testing Once your item dataset is uploaded, navigate to **"Experience"** to test the search functionality. The system provides default search strategies while allowing you to create and compare multiple custom configurations. Choose from three preset modes — **Semantic, Keyword, or Balanced ** — along with custom settings to tailor your unique business scenarios.

## 2. Image Search Configuration & Testing The platform provides three specialized image search modes based on **visual details, object similarity, and custom commands**. In the configuration panel, you can also define the maximum number of results for flexible adaptation across diverse application scenarios.

## 3. Search Result Ranking & Re\-ranking We support four flexible sorting methods for search results to suit different business scenarios: * **Prioritized Recall Protection:** Ensures high\-quality baseline results. * **Recall Reordering:** Utilises AI semantic optimisation for precision ranking. * **Boosting and Burying Rules:** Dynamically elevates or suppresses specific items based on strategic priorities. * **Field\-Based Sorting:** Supports multi\-weight scoring and ascending/descending orders based on specific field logic.

## 4. Query Intelligence & Auxiliary Features Enhance search accuracy and user search experience with three intelligent features: * **Synonym Mapping:** Bridges the gap between different terminologies. * **Typo Correction:** Automatically rectifies user input errors. * **Query Completion:** Suggests relevant terms in real\-time.

This concludes the ** ** guide to multimodal search capabilities. To configure Recommendation and Q&A modules, or to perform API Integrations, please refer to the respective configuration sections in the help documentation. :::tip \*说明 * To learn more about Viking AI search product capabilities, see: [Viking AI search engine, an integrated intelligent engine for enterprise search and recommendation](https://www.volcengine.com/product/AI-Search-Rec) :::

Viking AI Search