English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In the field of software engineering, applying language models to the tokensequence of source code is the state-of-art approach to build a coderecommendation system. The syntax tree of source code has hierarchicalstructures. Ignoring the characteristics of tree structures decreases the modelperformance. Current LSTM model handles sequential data. The performance ofLSTM model will decrease sharply if the noise unseen data is distributedeverywhere in the test suite. As code has free naming conventions, it is commonfor a model trained on one project to encounter many unknown words on anotherproject. If we set many unseen words as UNK just like the solution in naturallanguage processing, the number of UNK will be much greater than the sum of themost frequently appeared words. In an extreme case, just predicting UNK ateverywhere may achieve very high prediction accuracy. Thus, such solutioncannot reflect the true performance of a model when encountering noise unseendata. In this paper, we only mark a small number of rare words as UNK and showthe prediction performance of models under in-project and cross-projectevaluation. We propose a novel Hierarchical Language Model (HLM) to improve therobustness of LSTM model to gain the capacity about dealing with theinconsistency of data distribution between training and testing. The newlyproposed HLM takes the hierarchical structure of code tree into considerationto predict code. HLM uses BiLSTM to generate embedding for sub-trees accordingto hierarchies and collects the embedding of sub-trees in context to predictnext code. The experiments on inner-project and cross-project data setsindicate that the newly proposed Hierarchical Language Model (HLM) performsbetter than the state-of-art LSTM model in dealing with the data inconsistencybetween training and testing and achieves averagely 11.2\% improvement inprediction accuracy.

Improving the Robustness to Data Inconsistency between Training and Testing for Code Completion by Hierarchical Language Model