The overview page describes the product positioning and supported scenarios for the Intelligent Search algorithm product. It also allows developers to quickly explore and understand the product features, algorithm performance, and the underlying principles for various scenarios.
The overview page also displays statistics on the number of projects and resource usage for your account. It provides quick links to help documentation.
My projects
Total services: The total number of services created in your account.
Published services: Lists all services published in the current account.
Intelligent Document Parsing package usage: If you have purchased an Intelligent Document Parsing package, this section displays usage details, such as the total and remaining amount.
Scenarios
Search enhancement
This feature improves search performance in various domains, such as encyclopedias, healthcare, and news, using high-quality algorithms.
Quick experience:
Intelligent search results: Provides intelligent semantic search by combining traditional keyword search, vector retrieval, and Natural Language Processing (NLP) algorithms. It integrates proprietary capabilities for multi-language query analysis, such as tokenization, Named Entity Recognition (NER), error correction, rewriting, and classification. It also integrates pre-trained vector representation with multiple model structures (encoder-only and decoder-only) and hybrid retrieval with multi-factor sorting, such as text matching and deep semantic matching. This approach delivers industry-leading search performance compared to pure vector search.
Traditional keyword search results: Similar to Elasticsearch, this search is based solely on keyword matching.
Vector retrieval results: A search based on text embedding similarity.
Address search
This feature searches for information related to the input address text and returns relevant Point of Interest (POI) results.
Quick experience:
Intelligent search can parse non-standard addresses. It completes and normalizes them according to standard address hierarchy rules to provide standardized and structured address resolution.
Resume search
This feature intelligently connects candidates from a large talent pool with job positions. This helps ensure that the best talent is matched with the most suitable roles.
Quick experience:
You can enter a custom or recommended query to see the intelligent search results. The panel on the right shows the core algorithm logic for this scenario. It includes four main steps: core information extraction, query segment result extension, merging multiple extension results, and structured and relevance queries.
Step 1: Core information extraction. This step extracts keywords from the query.
Step 2: Query segment result extension. This step interprets and explains the extracted keywords.
Step 3: Merge multiple extension results. This step analyzes and merges the results from multiple query segments into a single long text that contains key information.
Step 4: Structured and relevance queries. The structured query semantically matches the extended query entity text with database fields to support structured searches. The relevance query analyzes the relevance of the extended query text to support unstructured text information searches.
Note: The information shown above is virtual data generated by the algorithm and does not represent real information.
Image search
This feature searches an image gallery using a text description and returns matching images and related structured information.
File search
This feature searches for relevant files using a text description and returns a list of matching files and related structured information.