Enhanced Best Practices for the Game Industry.
Customer background
The customer is a leading cultural community and video platform in China. Its game section provides players with various content, such as game downloads, game tests, community discussions, and walkthroughs. Game revenue constitutes a significant portion of the company's total revenue.
Search business background
The platform features various forms of content, including videos, wikis, walkthroughs, and user profiles. The search feature must support all of these categories.
The business focuses on scenarios such as game searches and community forum walkthrough searches. Search is a critical feature for driving business conversion.
The performance of the company's self-developed search engine did not meet business needs, resulting in a high rate of no-result queries and a low click-through rate (CTR).
Tokenization for the game industry is complex and specialized. It involves queries for game names, versions, characters, equipment, and streamers. New terms appear quickly. Search performance directly affects the end user experience.
Tokenization in game scenarios involves aliases, terms, and nicknames. Business optimization requires the operations team to be involved, but the self-developed search engine could not meet these operational needs.
The search guidance feature must support custom search requirements.
OpenSearch Industry-specific Enhanced Edition solutions
OpenSearch is a one-stop intelligent search development platform from Alibaba Cloud, built on a large-scale distributed search engine. It lets you integrate a high-quality search service with a single click and no development effort. The core of OpenSearch is a search engine refined by years of Alibaba's technical expertise, offering industry-leading search capabilities and algorithms. The platform is fully open and supports calling your own algorithm models to meet business needs across various industries and scenarios.
Custom tokenization based on customer data
For specific business realms, the text search tokenization algorithm automatically creates a custom tokenization model from customer data. This process can be completed within a day and generates a text tokenizer customized for that realm.
Tokenization performance comparison: General-purpose Edition VS. Industry-specific Enhanced Edition for games
2. Vector recall model for the game industry
Traditional text search uses algorithms such as tokenization, synonym matching, error correction, and word weight adjustments to improve semantic search performance. In contrast, the semantic vector recall model, which is based on deep learning, offers superior representational power. This allows it to better handle variations in user search queries, such as abbreviations, aliases, and spelling errors.
3. Custom Search Guidance
The drop-down suggestion feature is supported.
Intelligent query extraction based on user document content
Candidate queries for drop-down suggestions are retrieved using methods such as Chinese prefixes, full Pinyin, Pinyin initial abbreviations, a mix of Chinese characters and Pinyin, post-tokenization prefixes, and Chinese homophones or typos.
Top searches and hints are essential features for any complete search engine.
Prominent search box placement
This placement provides significant business value.
Customer value
The service was efficiently integrated and launched in just one week. It required minimal debugging and immediately addressed the customer's business needs.
The search performance of the Industry-specific Enhanced Edition for games was significantly better than the self-built search. Core metrics showed a 10% decrease in page views (PV) that returned no results and a 40% decrease in the no-result rate, which improved business conversion.
It resolved most of the issues caused by the previous self-built search. The Intelligent Semantic Understanding feature, designed for the game industry, enhanced the user experience and increased user stickiness.
Tests on important operational terms maintained by the customer showed that the search results met expectations. The operations team can perform tuning and maintenance directly from the console. This allows them to collaborate with product and technology teams, take ownership of search metrics, and continuously track search feature iterations.
For more information about the OpenSearch Industry-specific Enhanced Edition for games, see the product documentation.