Knowledge Base Intelligent Search

---

To improve information retrieval efficiency, the SERVICEME platform provides two search methods: Quick Search and AI Search, both accessible through a unified entry point on the homepage. Each search mode is suitable for different types of queries, meeting diverse usage scenarios.

Intelligent Search Entry

Users can initiate a search directly from the search box at the top of the SERVICEME homepage:

• Enter keywords or questions in the search box and press Enter to access the search interface;
• The system will automatically recognize user intent and process the query using either Quick Search or AI Search as needed.

---

Quick Search — Direct Access to Information Sources via Keywords

Quick Search is based on a keyword matching mechanism. The system quickly locates relevant content across multiple data sources, including:

• Knowledge Base Document Retrieval: The system searches for documents in the knowledge base that contain the entered keywords;

  • For example, searching for Microsoft 365 will automatically look for related content in all knowledge documents;
  • Matched documents are sorted by relevance to prioritize the most relevant materials;
  • You can click "View More" below the search results to preview documents online.

• QnA Matching: The system will also automatically identify and recommend frequently asked questions and answers related to the keywords.

• Online Search: If the knowledge base does not contain relevant content, the system will also provide extended information sources through online search.

✅ Applicable Scenarios: Quick Search is suitable for quickly locating and viewing information about known keywords, terms, product names, function points, etc.

AI Search — Semantic Understanding, Intelligent Answer Generation

AI Search leverages natural language processing technology to deeply understand the real intent behind user queries, transforming them into internal system requests and providing accurate, natural answers by integrating multiple data sources.

Feature Highlights

• Automatic Question Parsing: Users can directly enter complete questions or descriptive statements (e.g., "How to configure Microsoft 365 email auto-forwarding?"), and AI will automatically parse the intent and convert it into a searchable knowledge request.

• Multi-source Integrated Retrieval: AI Search answers may come from:

  • Knowledge base documents
  • Configured QnA libraries
  • Online information (such as web content)

• Cited Source Annotation: Each AI answer will automatically include the source document of the referenced knowledge;

  • Clicking the citation will open the document on the left for content preview, making it easy to verify and read in depth.

• Intelligent Summary Output: After answering, users can click the "Mind Map/Outline" button on the right side of the answer. The system will automatically extract and summarize the current answer content:
  • Mind Map visualization displays the core points and structure;
  • Outline presents a clear content structure, making it easy for subsequent organization, reuse, or documentation.

✅ Applicable Scenarios: AI Search is suitable for scenarios where keywords are uncertain, or when systematic answers or more complex questions are needed.

---

Feature Comparison Overview

Feature Comparison	Quick Search	AI Search
Input Method	Keyword	Natural language question
Matching Mechanism	Keyword matching	Intent recognition + multi-source retrieval
Returned Content	Documents, QnA, online results	Question answers (with source citation)
Additional Capabilities	Document preview	Citation jump, mind map/outline generation
Applicable Scenarios	Quick content location	Obtain structured answers

---

Search Scope

• Content Searched: File name, file content, QnA questions, QnA answers

• Supported Languages: Simplified Chinese, Traditional Chinese, English (other languages can be added in multilingual settings)

• Supported File Types: TXT, DOCX, XLSX, XLS, PPTX, PDF

• Search Result Display: All, Documents, QnA, Online

• Search Filter Conditions: File space, date, metadata filtering, document match similarity, QnA match similarity

• Search Result Sorting: Match degree, time dimension