一种融合传统深度学习与LLM的API补全方法
By integrating traditional deep learning with LLM methods, and combining semantic analysis and inverted rank fusion algorithms, this approach addresses the shortcomings of existing API completion techniques in handling complex semantics and long-distance dependencies, achieving more efficient and accurate API recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV OF SCI & TECH
- Filing Date
- 2025-11-12
- Publication Date
- 2026-07-17
AI Technical Summary
Existing API completion technologies perform poorly when dealing with complex semantic relationships and long-distance dependencies. Traditional deep learning methods struggle to understand the logical connections between API calls, and LLM suffers from the problem of generating irrelevant or incorrect "illusions," affecting developer trust.
This paper integrates traditional deep learning and LLM methods. It generates preliminary API candidates through deep learning models, combines semantic analysis and optimization screening with large language model LLM, and uses the inverted rank fusion algorithm to fuse the ranking results to generate the final API completion list.
It significantly improves the accuracy and reliability of API recommendations, provides more accurate and efficient API call support, and reduces the risk of LLM generating irrelevant APIs.
Smart Images
Figure CN121300847B_ABST