Abstract:AbstractCode summarization aims at generating a code comment given a block of source code and it is normally performed by training machine learning algorithms on existing code block-comment pairs. Code comments in practice have different intentions. For example, some code comments might explain how the methods work, while others explain why some methods are written. Previous works have shown that a relationship exists between a code block and the category of a comment associated with it. In this article, we aim to investigate to which extent we can exploit this relationship to improve code summarization performance. We first classify comments into six intention categories and manually label 20,000 code-comment pairs. These categories include “what,” “why,” “how-to-use,” “how-it-is-done,” “property,” and “others.” Based on this dataset, we conduct an experiment to investigate the performance of different state-of-the-art code summarization approaches on the categories. We find that the performance of different code summarization approaches varies substantially across the categories. Moreover, the category for which a code summarization model performs the best is different for the different models. In particular, no models perform the best for “why” and “property” comments among the six categories. We design a composite approach to demonstrate that comment category prediction can boost code summarization to reach better results. The approach leverages classified code-category labeled data to train a classifier to infer categories. Then it selects the most suitable models for inferred categories and outputs the composite results. Our composite approach outperforms other approaches that do not consider comment categories and obtains a relative improvement of 8.57% and 16.34% in terms of ROUGE-L and BLEU-4 score, respectively.

Classifying Python Code Comments Based on Supervised Learning

Why My Code Summarization Model Does Not Work: Code Comment Improvement with Category Prediction

Why My Code Summarization Model Does Not Work

How to Identify Class Comment Types? A Multi-language Approach for Class Comment Classification

Analyzing Code Comments to Boost Program Comprehension

Automating Just-In-Time Comment Updating

CPC: Automatically Classifying and Propagating Natural Language Comments via Program Analysis

Performance Comparison of Binary Machine Learning Classifiers in Identifying Code Comment Types: An Exploratory Study

An Intra-Class Relation Guided Approach for Code Comment Generation.

Speculative Analysis for Quality Assessment of Code Comments

Associating Natural Language Comment and Source Code Entities

Dopamin: Transformer-based Comment Classifiers through Domain Post-Training and Multi-level Layer Aggregation

Learning Code Context Information to Predict Comment Locations

Neural Comment Generation for Source Code with Auxiliary Code Classification Task

Taxonomy of inline code comment smells

CodeAttention: Translating Source Code to Comments by Exploiting the Code Constructs

Beyond code: Is there a difference between comments in visual and textual languages?

Code Comments: A Way of Identifying Similarities in the Source Code

From Code to Natural Language: Type-Aware Sketch-Based Seq2Seq Learning

Comments on Comments: Where Code Review and Documentation Meet

Research on tendency classification algorithm for online movie comment