International Journal of Software Engineering and Knowledge Engineering

Topic detection is a hot issue that many researchers are interested in. The previous researches focused on the single data stream, they did not consider the topic detection from different data streams in a harmonious way, so they cannot detect closely related topics from different data streams. Recently, Twitter, along with other SNS such as Weibo, and Yelp, began backing position services in their texts. Previous approaches are either complex to be conducted or oversimplified that cannot achieve better performance on detecting spatial topics. In our paper, we introduce a probabilistic method which can precisely detect closely related bursty topics and their bursty periods across different data streams in a unified way. We also introduce a probabilistic method called Latent Spatial Events Model (LSEM) that can find areas as well as to detect the spatial events, it can also predict positions of the texts. We evaluate LSEM on different datasets and reflect that our approach outperforms other baseline approaches in different indexes such as perplexity, entropy of topic and KL-divergence, range error. Evaluation of our first proposed approach on different datasets shows that it can detect closely related topics and meaningful bursty time periods from different datasets.

GitHub and Stack Overflow are often used together for software development. GH-SO users, who use both GitHub and Stack Overflow, contribute to the development of various software projects in GitHub and share their knowledge and experience on software development in Stack Overflow. To widely understand the interests and working habits of GH-SO users on software development, it is important to investigate how GH-SO users utilize GitHub and Stack Overflow. In this paper, we present an exploratory study on GitHub commit and Stack Overflow post activities of GH-SO users. Specifically, we investigate the working habits of GH-SO users on GitHub commit and Stack Overflow post activities. We randomly selected 19,756 of GH-SO users as our target sample and collected 2,819,483 and 2,147,317 of commit activity data and post activity data of the GH-SO users. We then categorized the collected commit and post activity datasets into specific categories on programming languages and statistically analyzed the categorized commit and post activity datasets. As the results of our analysis, we found the following: (1) The overall commit and post activities of the GH-SO users share some similarity. (2) The commit activities gradually change while the post activities drastically change over time. (3) The commit activities of the GH-SO users are broadly distributed while the post activities are narrowly distributed and the commit activity can be better predictor for post activity. (4) The commit activity of the GH-SO users tends to be performed prior post activity. We believe that our findings can contribute to finding the ways to better support commit and post activities of GitHub and Stack Overflow users.

In the field of software engineering, practitioners’ share in the constructed knowledge cannot be underestimated and is mostly in the form of grey literature (GL). GL is a valuable resource though it is subjective and lacks an objective quality assurance methodology. In this paper, a quality assessment scheme is proposed for question and answer (Q&A) sites. In particular, we target stack overflow (SO) and stack exchange (SE) sites. We model the problem of author reputation measurement as a classification task on the author-provided answers. The authors’ mean, median, and total answer scores are used as inputs for class labeling. State-of-the-art language models (BERT and DistilBERT) with a softmax layer on top are utilized as classifiers and compared to SVM and random baselines. Our best model achieves $63.8\%$ accuracy in binary classification in SO design patterns tag and $71.6\%$ accuracy in SE software engineering category. Superior performance in SE software engineering can be explained by its larger dataset size. In addition to quantitative evaluation, we provide qualitative evidence, which supports that the system’s predicted reputation labels match the quality of provided answers.

With the advent of web 2.0, web application architectures have been evolved, and their complexity has grown enormously. Due to the complexity, testing of web applications is getting time-consuming and intensive process. In today’s web applications, users can achieve the same goal by performing different actions. To ensure that the entire system is safe and robust, developers try to test all possible user action sequences in the testing phase. Since the space of all the possibilities is enormous, covering all user action sequences can be impossible. To automate the test script generation task and reduce the space of the possible user action sequences, we propose a novel method based on long short-term memory (LSTM) network for generating test scripts from user clickstream data. The experiment results clearly show that generated hidden test sequences are user-like sequences, and the process of generating test scripts with the proposed model is less time-consuming than writing them manually.

Automatic test oracle generation is a bottleneck in realizing full automation of the entire software testing process. This study proposes a new method for automatically generating a test oracle for a new test input on the basis of several historical test cases by using a backpropagation neural network (BPNN) model. The new method is different from existing test oracle techniques. Specifically, our method has two steps. First, the values of variables are collected as training data when several historical test inputs are used to execute the program at different breakpoints. The test oracles (pass or fail) of these test cases are utilized to classify and label the training data. Second, a new test input is used to execute the program at different breakpoints, where the trained BPNN prediction model automatically generates its test oracle on the basis of the collected values of the variables involved. We conduct an experiment to validate our method. In the experiment, 113 faulty versions of seven types of programs are used as experimental objects. Results show that the average prediction accuracy rate of 74,651 test oracles is 95.8%. Although the failed test cases in the training data account for less than 5%, the overall average recall rate (prediction accuracy of test case execution failure) of all programs is 78.9%. Furthermore, the trained BPNN can reveal not only the impact of the values of variables but also the impact of the logical correspondence between variables in test oracle generation.

Knowledge representation learning (knowledge graph embedding) plays a critical role in the application of knowledge graph construction. The multi-source information knowledge representation learning, which is one class of the most promising knowledge representation learning at present, mainly focuses on learning a large number of useful additional information of entities and relations in the knowledge graph into their embeddings, such as the text description information, entity type information, visual information, graph structure information, etc. However, there is a kind of simple but very common information — the number of an entity’s relations which means the number of an entity’s semantic types has been ignored. This work proposes a multi-source knowledge representation learning model KRL-NER, which embodies information of the number of an entity’s relations between entities into the entities’ embeddings through the attention mechanism. Specifically, first of all, we design and construct a submodel of the KRL-NER LearnNER which learns an embedding including the information on the number of an entity’s relations; then, we obtain a new embedding by exerting attention onto the embedding learned by the models such as TransE with this embedding; finally, we translate based onto the new embedding. Experiments, such as related tasks on knowledge graph: entity prediction, entity prediction under different relation types, and triple classification, are carried out to verify our model. The results show that our model is effective on the large-scale knowledge graphs, e.g. FB15K.

Trigger-action (TA) programming is a programming paradigm that allows end-users to automate and connect IoT devices and online services using if-trigger-then-action rules. Early studies have demonstrated this paradigms usability, but more recent work has also highlighted complexities that arise in realistic scenarios. To facilitate end-users in TA programming, we propose AutoTAR, a context-aware conversational recommendation technique for recommending TA rules. AutoTAR leverages a TA knowledge graph to encode semantic features and abstract functionalities of rules, and then takes a two-phase method to recommend TA rules to end-users: during the context-aware recommendation phase, it elicits user preferences from programming context and recommends the top-N rules using a mixed content and collaborative technique; during the conversational recommendation phase, it justifies recommendations by iteratively raising questions and collecting feedback from end-users.

We evaluate AutoTAR on Mturk and real data collected from the IFTTT community. The results show that our method outperforms state-of-the-arts significantly — its context-aware recommendation outperforms RecRules by 26% on R@5 and 21% on NDCG@5; its conversational recommendation outperforms LARecommender (a conversational recommender with the LA model) by 67.64% on accuracy. In addition, AutoTAR is effective in solving three problems frequently occurring in TA rule recommendations, i.e., the cold-start problem, the repeat-consumption problem, and the incomplete-intent problem.