Application of Cloud Search service in app search scene

Search is ubiquitous, especially in mobile connectivity today. Search has played an important role in social, e-commerce, and video apps. As a gateway to information, search can help users find the information they want from a huge amount of information. The typical scenario for app search is as follows:

Find Users: According to the input matching user nickname, according to user attributes to sort the search results;

Search topic: According to the user input matches the topic, according to the topic attribute to the search result sorting;

Search Posts: According to the user input matches the content of the post, according to the content properties of the search results sorted;

Search by location or recommendation: according to the user input address information, converted to latitude and longitude, in a certain distance to match the topic, the content of the address information, according to their properties to sort the search results.

In fact, although the concept of search is popular, it is not easy to make a good search service. In detail, such as performance and search accuracy, there are difficult points to be ignored, which are often the key to improving the search experience. Customer difficulties are as follows:

database-based performance issues:

? The data are generally stored in the MySQL database, and there is an associated operation in the search, the query efficiency is very low;

? Database query performance is difficult to achieve in milliseconds, the user search experience is very bad.

database-based matching issues:

In the database, it is not possible to comprehensively consider the properties of the query condition after sorting;

? Some special characters cannot be filtered in the database, and the search process is inaccurate if there are special characters in the user's nickname.

Elasticsearch issues based on open source search engines:

? The default BM25 correlation scoring sort does not guarantee the rationality of search results;

? A simple word breaker improperly handles special characters.

For example: When searching for the user "sea" Why large v users "kkw in the eyes of the De star Sea" in the "Looking for that sea" behind?

For the customers who want to build app search, from the technical level, the implementation of the following scenarios. The cloud Search service is based on Elasticsearch and is able to complete terabytes of retrieval tasks and return results in milliseconds, which can be a good solution to the performance problems of traditional databases.

Overall implementation of the program

In the cloud Search service, we make the following optimizations for the customer's search pain points to help customers enhance their user experience, improve efficiency, and quickly achieve business growth.

Search solution Optimization in cloud Search service:

Support Custom Thesaurus , users in the word database to add hot search words, keywords, network commonly used words to improve the matching accuracy of the search input;

Support Custom weights , the user can define different query criteria according to the weight of the business to determine the final result of the order;

To support the more accurate (planning), the cloud Search service based on the search results of the click Behavior, automatically update the attribute weights, to achieve the final result sorting more precise effect.

In the cloud Search service, we provide additional solutions on word segmentation and thesaurus to help customers solve problems such as search accuracy.

Cloud Search service Participle, thesaurus scenario:

Because the user nickname and content, the meaning of the topic is different, for different fields can take a different word breaker, in order to improve the accuracy of matching. The specific effects are shown in the following table:

Take the use of simple traditional word breaker, can also support all searchable fields can handle traditional Chinese characters;

For hot search words, keywords, network commonly used words, often change with the time, therefore, we take the user custom Thesaurus update, to improve the accuracy of the search.

Overall, the search scenario for cloud search services in Internet apps has the following advantages:

one-touch operation, worry-free operations

All the main operations are a key to reach, professional team close to the operation and maintenance, rapid implementation of the search function to build modules;

flexible and resilient

On-demand application, online expansion, easily meet business growth needs;

Flexible word base management

Support custom thesaurus and pinyin word segmentation, support the hot update of the thesaurus without restarting, the configuration is effective, no impact on the end user;

Flexible word segmentation, flexible weights

Different fields use different word segmentation methods to improve the accuracy of search. Different conditions to set different weights, improve the accuracy of search;

efficient real-time retrieval

The millisecond retrieval results are returned and the new data is retrieved immediately, providing a good user experience.

Application of Cloud Search service in app search scene