Cascaded Fast and Slow Models for Efficient Semantic Code Search
This addresses the need for practical and scalable code search tools for developers, representing a strong incremental improvement over existing methods.
The paper tackles the problem of inefficient and ineffective semantic code search by proposing a cascaded framework with fast retrieval and slow re-ranking models, achieving a state-of-the-art average MRR of 0.7795 across 6 programming languages compared to the previous 0.713.
The goal of natural language semantic code search is to retrieve a semantically relevant code snippet from a fixed set of candidates using a natural language query. Existing approaches are neither effective nor efficient enough towards a practical semantic code search system. In this paper, we propose an efficient and accurate semantic code search framework with cascaded fast and slow models, in which a fast transformer encoder model is learned to optimize a scalable index for fast retrieval followed by learning a slow classification-based re-ranking model to improve the performance of the top K results from the fast retrieval. To further reduce the high memory cost of deploying two separate models in practice, we propose to jointly train the fast and slow model based on a single transformer encoder with shared parameters. The proposed cascaded approach is not only efficient and scalable, but also achieves state-of-the-art results with an average mean reciprocal ranking (MRR) score of 0.7795 (across 6 programming languages) as opposed to the previous state-of-the-art result of 0.713 MRR on the CodeSearchNet benchmark.