OpenSearch:General-purpose vs. Industry Algorithm editions

Industry Algorithm edition for e-commerce

In the e-commerce industry, search performance directly impacts Gross Merchandise Volume (GMV) conversion, making it a significant technical challenge for businesses. Built on years of industry experience, the Industry Algorithm Edition provides a vertical solution specifically for e-commerce. It leverages advanced natural language processing (NLP) technology from Alibaba DAMO Academy to address industry-specific pain points and requirements. This edition offers dedicated e-commerce query analysis capabilities, built-in industry sort expressions, and specialized algorithms. It lowers the barrier to entry with one-stop configuration, speeds up the onboarding process, and delivers superior search performance.

Feature comparison

Feature	General-purpose edition	Industry algorithm edition
One-stop configuration	After creating an application, you must manually configure query analysis, sort policy, and drop-down suggestion models.	Designed for common e-commerce search scenarios, this edition provides templates for application schema and index schema. These templates enable one-stop configuration and reduce the learning curve for new users.
Query analysis	Provides features for general industries, such as synonym expansion, stop word filtering, spelling correction, term weight analysis, and category prediction.	Provides an enhanced analyzer and query analysis features for general e-commerce and its popular verticals, such as fashion, maternal and infant care, automotive, and dining. It builds more accurate indexes and recognizes user query intent by analyzing product and user characteristics, delivering better performance than the General-purpose Edition.
Sort policy	Requires manual configuration and tuning of the sort policy for specific business scenarios.	Based on templates for the application schema and index schema, this edition provides commonly used rough sort and fine sort expressions for e-commerce, meeting most ranking requirements without additional configuration.
Feature iteration	System default dictionaries for the analyzer and query analysis are updated periodically.	Continuously iterates and optimizes tokenization and query analysis capabilities based on evolving terms and products in the e-commerce industry, ensuring timely and effective service.

Query analysis performance comparison

The Industry Algorithm Edition offers deeper query analysis optimizations than the General-purpose Edition. It fixes poorly handled queries and enriches its dictionaries by incorporating best practices from the e-commerce industry. The following tables provide examples of these improvements.

• Tokenization (space-delimited)

Query	General-purpose edition	Industry algorithm edition
Hanbencui Nutrilite	Han ben cui Nutrilite	Hanbencui Nutrilite
small gold tube concealer	small-gold tube concealer	small gold-tube concealer
Supor air fryer	Supor air fry pot	Supor air-fryer
Coco movie guitar	Coco movie story guitar	Coco movie guitar
street style suit set	street style suit set	street-style suit set
Xige shampoo	Xi ge shampoo	Xige shampoo
Judydoll mono eyeshadow	Judy doll mono eye-shadow	Judydoll mono eyeshadow

• Named Entity Recognition (NER)

Query	General-purpose edition	Industry algorithm edition
apple 11 case for women	apple: category, 11: number, case: common term, for women: audience	apple: brand, 11: number, case: category, for women: audience
Hanbencui Nutrilite	Hanbencui: common term, Nutrilite: brand	Hanbencui: brand, Nutrilite: brand
vegetable seedling	vegetable seedling: new term	vegetable seedling: category
cosmetic contact lenses	cosmetic contact lenses: common term	cosmetic contact lenses: category

• Synonyms

Query	General-purpose edition	Industry algorithm edition
wide-leg pants for women new 2021	wide-leg pants for women new 2021	wide-leg pants (leg pants, wide-foot pants, flare pants) for women new 2021
Hanbencui Nutrilite	Hanbencui Nutrilite	Hanbencui Nutrilite (nutrilite)
women's sneakers running shoes	sneakers for women running shoes (runners)	sneakers for women running shoes (runners, joggers, walking shoes)
Supor air fryer	Supor air fryer	Supor (supor, supoer, suboer) air fryer
dishwasher Siemens	dishwasher Siemens	dishwasher Siemens (siemens)

• Spelling Correction

Query	General-purpose edition	Industry algorithm edition
wireles bra	wireles bra	wireless bra
putizhu	putizhu	Bodhi beads
baisiwa	baisiwa	white stockings
red baobao women's bag	red baobao women's bag	red bag women's bag
sushe refrigerator	sushe refrigerator	dormitory refrigerator
chuanglian european style	chuanglian european style	curtains european style
maternity wear dama	maternity wear dama	maternity wear plus size
u-shaped pillow nami	u-shaped pillow nami	u-shaped pillow nano
masager waist	masager waist	massager waist
yami	yami	Amway
yingyang	yingyang	nutrition
jiankang	jiankang	health
toothpast	toothpast	toothpaste

Usage notes

• For more information about how to create an Industry Algorithm Edition application for e-commerce, see Industry Algorithm Edition for E-commerce.

• You can upgrade an exclusive General-purpose Edition application to the Industry Algorithm Edition. This upgrade is irreversible.

• The Industry Algorithm Edition is available only for exclusive applications.

• To upgrade a shared application, you must first change its specification to exclusive. Ensure that both the online application and the instance specifications are exclusive before upgrading it to the Industry Algorithm Edition.

• In the application schema, you must add the field names corresponding to the field tags to prevent errors.

• In the index schema, you must add the index names corresponding to the index tags to proceed.

Upgrade an exclusive instance

To upgrade an exclusive General-purpose Edition application to the Industry Algorithm Edition for e-commerce, follow these steps:

1. On the instance management page, find the General-purpose Edition application that you want to upgrade. In the Actions column, click More and select Upgrade to Industry Algorithm Edition. You are redirected to the purchase page.

On the purchase page, confirm the current configuration information. For Industry Type, select E-commerce. Read and select the service agreement, and then click Buy Now.

After the payment is successful, a message confirms that the new instance will be provisioned in 1 to 5 minutes. Click management console to return to the instance management page.

2. After you upgrade to the Industry Algorithm Edition, find the corresponding instance on the instance management page and click Details. Then, click Adapt to Industry Algorithm Edition to modify the application.

3. Modify the application by following these steps: Feature Selection --> Application Schema --> Index Schema --> Data Source.

1) Feature selection: The template features include query analysis, sort policy, and other services.

The query analysis features include e-commerce spelling correction, term weight analysis, stop word filtering, synonym expansion, tokenization, and Named Entity Recognition (NER). The sort policy features include text relevance. The other services features include drop-down suggestion. All template features are selected by default. You can customize the selection based on your business needs.​​

2) Application schema: You can automatically generate the application schema.

In the field tag bar at the top of the page, click the required tags, such as Product Title, Brand Name, or Category Name. The system automatically generates the corresponding fields in the application schema table below.​​

Note: Field tags are used to quickly generate the index schema, query analysis rules, and sort policy. You must add the field names that correspond to the field tags in the application schema.

3) Index schema:

​In the index tag bar at the top of the page, click the required tags, such as Category Name Index, Default Index, or Title Index. The system automatically generates the corresponding indexes in the index field list below.

​Note: Index tags are used to quickly generate query analysis rules and sort policy. You must add the index names that correspond to the index tags in the index schema.

4) Data source: (This example uses an RDS data source.)

​Click Add Data Source on the left. In the Add Data Source dialog box, select the data source type (supported types include RDS, MaxCompute, and POLARDB), the database, and the data table. Then, click OK.​

4. Sort search results:

​On the Sort Search Results page, check the status of each sort expression in the Rough sort configuration and Fine sort configuration sections. If any sort expression is invalid, click Edit to modify it or click Delete to remove it.

After you confirm that the status of all sort expressions is valid, click Finish.​

5. Modify query analysis:

Go to the Query analysis configuration page. The system prompts you that changes to the application schema affect the query intent understanding configuration and that you must modify the relevant rules.

When you modify the query analysis, note the following:

• Change the index type referenced by the original query analysis to Enhanced query analysis for e-commerce.

• If NER in the query analysis references an intervention dictionary, the dictionary's analyzer type must also be changed to Industry - General analyzer for e-commerce.

6. After you save the configuration, choose whether to synchronize incremental data and whether to automatically publish the offline application.

• If you choose not to synchronize, the offline application performs a full data push. Any incremental data that enters the application during version creation is not synchronized to the new version.

• If you choose not to automatically publish, you must manually deploy the changes later by navigating to the details page of the offline application and clicking Publish.

7. Complete the configuration and test:

​After the configuration is complete, return to the instance management page. Confirm that the application has been upgraded to the Industry Algorithm Edition and that the instance status is Active. You can then click search test to verify its performance.