SIEVE: Helping developers sift wheat from chaff via cross-platform analysis

Abstract

Software developers have benefited from various sources of knowledge such as forums, question-and-answer sites, and social media platforms to help them in various tasks. Extracting software-related knowledge from different platforms involves many challenges. In this paper, we propose an approach to improve the effectiveness of knowledge extraction tasks by performing crossplatform analysis. Our approach is based on transfer representation learning and word embeddings, leveraging information extracted from a source platform which contains rich domain-related content. The information extracted is then used to solve tasks in another platform (considered as target platform) with less domain-related contents. We first build a word embeddings model as a representation learned from the source platform, and use the model to improve the performance of knowledge extraction tasks in the target platform. We experiment with Software Engineering Stack Exchange and Stack Overflow as source platforms, and two different target platforms, i.e., Twitter and YouTube. Our experiments show that our approach improves performance of existing work for the tasks of identifying software-related tweets and helpful YouTube comments.

FindICI: Using machine learning to detect linguistic inconsistencies between code and natural language descriptions in infrastructure-as-code

“…There are no statistically significant differences among FastText and Word2vec models in terms of MCC. Our results confirm previous work (Sulistya et al 2020 ; Mikolov et al 2013 ) which assessed the superiority of Word2vec and FastText in a different context (i.e., text mining). In addition, our work agrees with the findings of previous work (Lau and Baldwin 2016 ) suggesting that Doc2vec creates document embeddings which align with lower frequency words when the documents are short and the corpus is relatively small.…”

“…In a recent study, Sulistya et al ( 2020 ) compared different word embedding learning methods for finding software-relevant tweets. Following their guidelines, we used the same hyper-parameter settings for each word embedding learning model (i.e., Word2vec , Doc2Vec , and FastText ).…”

“…To answer RQ2, we chose three widely used word embedding learning models, Word2vec , Doc2vec (Le and Mikolov 2014), and FastText (Joulin et al 2016). These embedding models are used by software engineering research for learning representations source codes and method names (Pradel and Sen 2018;Liu and et al 2019;Fakhoury et al 2018), and other natural language texts (Sulistya et al 2020). Word2vec is a two-layer neural network that processes text by creating vector representations from words.…”

“…Boosted by the emerging trend of learning-based approaches and word embedding in the software engineering research (Sulistya et al 2020;Pradel and Sen 2018;Liu and et al 2019;Omri and Sinz 2020;, we propose FINDICI, a novel approach to detect linguistic anti-patterns in IaC, focusing on name-body inconsistencies in IaC code units. We formulate name-body inconsistency detection as a binary classification problem and train a classifier that distinguishes between consistent and inconsistent code units.…”