… are from solving this problem and provide suggestions to future research, in this paper, we
… code summarization models more correctly and comprehensively, we need to pay much …

… chose not to include Hu et al. in our baselines because the work done by LeClair et al.
built upon their work and was shows to have higher performance, and is much closer to our …

… set and the online code retrieval during testing, our approach may cost much more time to
… However, we only need 89 ms in average to generate the summary for each testing code …

… We present our evaluation in two experiments. First, we train different neural code summarization
… CCE loss, as we established in Section 2, is by far the main means of computing loss …

… , we propose a brand-new framework for code summarization based on meta-learning and
code … As far as we know, the offline training dataset for the neural model is always used as …

… By far the most popular metric for measuring word overlap in source … Second, a vast
majority of neural code summarization approaches are based on this, or a similar variant of this …

… summary using state-of-the-art neural summarization. The intended purpose of the … are
not reliable predictors of how much a summary helps them understand code. This aligns with …

… 6.9, which is much lower than the standard deviation of tokens of codes. Considering that …
models (eg, CodeNN is the first to apply a neural network to code summarization), or they are …

… We then use a fully-connected layer to squash these long vectors back to the specified vector
… We use four baseline neural code summarization techniques. We then augment each with …

… , we observe that code summarization models can be easily overloaded with source code.
A … [53] conduct a comprehensive analysis of ive state-of-the-art neural code summarization …