OpenSearch:E-commerce industry algorithm edition

Industry types

OpenSearch Industry Algorithm Edition provides two Industry Types: E-commerce and general-purpose. The main differences are as follows:

• E-commerce: Provides a fixed field tag template (for example, product title, brand name, and category name). All field tags are required. Your business data fields must map to the tags in the e-commerce template. This type is suitable for e-commerce search scenarios.

• general-purpose: Supports fully custom fields with no field tag restrictions. If your business fields do not fully match the e-commerce template, we recommend choosing this type. For configuration instructions, see Configure the general-purpose industry template.

Configuration process

1. In the OpenSearch Industry Algorithm Edition console, go to the Instance Management page and click Create Instance.

2. On the buy page, select Industry Algorithm Edition for Product Edition and E-commerce for Industry Type. For Cluster Preference, choose Exclusive Cluster - General-purpose, Exclusive Cluster - Compute-optimized, or Exclusive Cluster - Storage-optimized based on your business requirements. The other configuration settings are the same as those for a general-purpose industry template configuration.

3. After the purchase, confirm the order:

On the Confirm Order page, verify the configuration parameters (such as region, application name, edition type, Industry Type, storage capacity, and compute resources) and the fee details. Select the service agreement and submit the order.

4. After the service is activated, go to the Instance Management page, find your application in the application list, and click Configure. Complete the following configuration steps in order: Feature Selection → Application Schema → Index Schema → Data Source → Complete.

(1) Feature Selection:

The template features are grouped into three categories: Query Analysis, Sort Policy, and Other Services.

The Query Analysis category includes E-commerce Typo Correction, E-commerce Term Weighting, E-commerce Stop Words, E-commerce Synonyms, text vectorization, and E-commerce Entity Recognition. The Sort Policy category includes multimodal search, Text Relevance, and Vector Relevance. The Other Services category includes autocomplete. By default, all template features are selected, but you can customize the selection based on your business needs.

(2) Application Schema:

You can create an application schema in one of four ways: Use Data Source, Manually Create, Import Template, or Use File. The following example uses the Use Data Source method.

On the Application Schema page, select the Create from Data Source method. Field tags (Product Title, Brand Name, Category Name, Product Description, On Sale, Listing Time, and Sales Volume) are displayed at the top of the page. Click a tag to automatically add the corresponding field to the table below.

Note: Field tags are used to quickly generate the industry-enhanced index schema, query analysis rules, and sort policies. You must add the field names that correspond to the field tags in the Application Schema.

(3) Index Schema:

On the Index Schema page, the index tags include Category Name Index, Default Index, Vector Index, and Title Index. Click a tag to automatically generate the corresponding index. The index field list displays the name, tag, included fields, and analyzer for each index. For the vector index, set the index tag to Vector Index and the analyzer to Vector - General E-commerce Vector.

Note:

• Index tags are used to quickly generate query analysis rules and sort policies. When you configure the Industry Algorithm Edition for E-commerce for the first time, you must add the index names that correspond to the index tags in the Index Schema.

• For the index with the vector index tag, you must set the analyzer to "Vector - General E-commerce Vector".

• For subsequent configurations of the Industry Algorithm Edition for E-commerce, add the e-commerce vector analyzer. You do not need to configure a vector tag.

(4) Data Source: (The following example uses an RDS data source.)

Configure the data source information:

​In the left panel, click Add Data Source. In the pop-up Add Data Source dialog box, select a cloud data source type (RDS, MaxCompute, or POLARDB), then select a database and a data table, and click OK.​

Set up field mapping:

In the field mapping step, map the OpenSearch table fields to the RDS source fields. Click Add Data to add the fields that you want to synchronize. After you complete the mapping, click OK.

(5) Complete:

The page displays the message: Industry Algorithm Edition application configuration completed. You can then: 1. Configure Query Analysis by using the selected industry template feature packages. 2. Configure a Sort Policy by using the selected sort policy models. 3. Configure and use the autocomplete feature.

Query testing

Search test

After the index is built, you can go to the Feature Extensions > Search Test page to test the index configured with the e-commerce industry text analyzer. Enter a search clause in the query field and click Search to view the results.

# For example, to search the default index, enter the following search clause:
default:'beef'

Vector search

To test the vector search feature, follow these steps:

1. Go to Search Algorithm Center > Retrieval Configuration > Query Analysis Configuration. In the default query analysis rule sys_default that is generated for the Industry Algorithm Edition, enable the text vectorization feature. To do this, click sys_default to open the Edit Rule dialog box. In the Feature Configuration section, select Text Vectorization, set Vector Index to vector, and click OK.

2. On the Feature Extensions > Search Test page, add the required parameters to run the test. After you add the parameters in the Parameters field, click Search to view the vector search results.

# Required parameters
query=default:'beef'
qp=sys_default
raw_query=beef

You can view the vector results in the query analysis process. Click the Query Analysis Process link in the search results to view the outcome of each analysis step, such as spelling correction, normalization, tokenization, stop word removal, term weighting, and synonym matching, as well as the vector index and dimension information from text vectorization.

Note:

• The vector index takes effect only after you configure the text vectorization feature in Query Analysis.

• The index used for query analysis must use the "Industry - General Analyzer for E-commerce" analyzer, and the industry must be set to "E-commerce Enhanced". In the text vectorization settings, you must select the analyzer that is configured with "Vector - General E-commerce Vector".

Multimodal search

The Industry Algorithm Edition for E-commerce automatically generates a multimodal search policy named sys_strategy for your instance. To test the multimodal search feature, you need to configure specific parameters on the Feature Extensions > Search Test page. For example, add the following parameters in the Parameters field and click Search to view the multimodal search results.

# Required parameters
query=default:'beef'
qp=sys_default
raw_query=beef
search_strategy=sys_strategy

Usage notes

• You can upgrade a General-purpose Edition application to the Industry Algorithm Edition, but you cannot downgrade an Industry Algorithm Edition application to the General-purpose Edition.

• The Industry Algorithm Edition is available only for Exclusive Cluster applications.

• For a sample schema for the e-commerce industry template, see e-commerce_industry_template_schemaDemo.json.

• The "Vector - General E-commerce Vector" index cannot be used alone. It takes effect only after you configure the text vectorization feature in Query Analysis for an index that uses the "Industry - General Analyzer for E-commerce" analyzer.

• In the Application Schema, you must add the field names that correspond to the field tags. Otherwise, an error occurs.

​If you do not add any field tags, the system displays a prompt that says "Please select the necessary field tags", and you cannot proceed. If your business data fields do not match the field tag template for the e-commerce industry, we recommend that you use the General-purpose Industry Template Configuration. The general-purpose industry template supports custom fields and has no tag restrictions.

• In the Index Schema, you must add the index names corresponding to the index tags. Otherwise, you cannot proceed to the next step.

If you do not add the index tags, the Next button is disabled and you cannot continue the configuration.