Deep Semantic Code Search aims to explore a joint embedding space for code and description vectors and then use it for a code search application.
In these experiments, there are 2 parts:
- The first one uses an approach suggested in [1] and we train their architecture on our own python dataset.
- The second approach expands on the first one through methodology suggested in [2] and we achieve reasonably good results.
We can observe that some sort of semantic information is captured the results:
Implementation of [1] is within Joint Training Model and [2] is within Code Summarization Transfer Learning
For [1], our dataset is provided within Joint Training Model. For [2], the full dataset is available on Google Cloud Platform.
For how to access data on GCP, please follow this link https://cloud.google.com/storage/docs/access-public-data
[1] https://guxd.github.io/papers/deepcs.pdf
[2] https://towardsdatascience.com/semantic-code-search-3cd6d244a39c