Learning to Mine Aligned Code and Natural Language Pairs from Stack Overflow
This work addresses the need for parallel data in code-related NLP tasks, offering a scalable solution for multiple programming languages, though it is incremental in improving data mining techniques.
The paper tackles the problem of mining high-quality aligned natural language and code pairs from Stack Overflow, proposing a novel method that uses hand-crafted and neural network-based features to classify pair quality, resulting in greatly expanded coverage and accuracy over existing methods, with experiments showing improvements in Python and Java.
For tasks like code synthesis from natural language, code retrieval, and code summarization, data-driven models have shown great promise. However, creating these models require parallel data between natural language (NL) and code with fine-grained alignments. Stack Overflow (SO) is a promising source to create such a data set: the questions are diverse and most of them have corresponding answers with high-quality code snippets. However, existing heuristic methods (e.g., pairing the title of a post with the code in the accepted answer) are limited both in their coverage and the correctness of the NL-code pairs obtained. In this paper, we propose a novel method to mine high-quality aligned data from SO using two sets of features: hand-crafted features considering the structure of the extracted snippets, and correspondence features obtained by training a probabilistic model to capture the correlation between NL and code using neural networks. These features are fed into a classifier that determines the quality of mined NL-code pairs. Experiments using Python and Java as test beds show that the proposed method greatly expands coverage and accuracy over existing mining methods, even when using only a small number of labeled examples. Further, we find that reasonable results are achieved even when training the classifier on one language and testing on another, showing promise for scaling NL-code mining to a wide variety of programming languages beyond those for which we are able to annotate data.