Workshop

Component-based synthesis is one of the hottest research areas in automated software engineering. It aims to generate programs from a collection of components like Java library. However, the program space constituted by all the components in the library is fairly large, which leads to a vast number of candidate programs generated for a long time. The intractability of the program space affects the synthesis efficiency of the program and the size of the program generated. In this paper, we propose Itas, a framework of iterative program synthesis via API usage knowledge from the Internet, which can significantly improve the efficiency of program synthesis. Itas aims to constrain the program space by combining two main ideas. First, narrow down the program space from the outside via the guidance of API usage knowledge. Second, expand the program space from the inside via iterative strategy based on knowledge. For evaluation, we collect a set of programming tasks and compare our approach with a program synthesis tool on synthesizing these tasks. The experiment results show that Itas can significantly improve the efficiency of program synthesis, which can reduce the synthesis time by 97.1% than the original synthesizer.

Jiaxin Liu

National University of Defense Technology

Wei Dong

School of Computer, National University of Defense Technology, China

Binbin Liu

National University of Defense Technology

Workshop

We study the problem of automatically generating source code from different forms of user intents. Existing methods treating this problem as a language generating task of the neural network, known as Neural Program Synthesis(NPS). Most of these methods struggle with achieving high generating accuracy, one reason for that is the incompleteness and inaccuracy of user intents for a specific programming task. Inspired by the Swarm Intelligence(SI) and Collective Intelligence(CI) techniques, we proposed an automatic task-specific user intent merging framework combining both the bio-inspired algorithm in SI and CI merged from multiple developers. Empirically, we show that our approach is able to provide more accurate and adequate input for NPS, and our experiment on CI indicates that knowledge merging among isolated software developers in our approach has a significant influence on NPS.

Daiyan Wang

National University of Defense Technology

Wei Dong

School of Computer, National University of Defense Technology, China

Yating Zhang

National University of Defense Technology

Workshop

Software design patterns are solutions to common software problems that are proven to work adequately in particular scenarios. Deciding which design pattern to use for a given software problem often requires practical knowledge acquired with experience in a similar domain and can be highly subjective and error-prone. Further, for novice programmers, an automated approach would be a tremendous help as they usually lack practical knowledge required for deciding which design pattern to use for a particular software problem. The majority of research in software design pattern prediction involves using software structure and features in determining which design pattern to implement. However, there are circumstances where software designers would prefer to know which design pattern to be used by looking at the design problem during or before the implementation phase. Existing design pattern prediction tools cannot be utilized in this scenario due to the absence of code and class structures. To address this issue, this paper proposes a new approach that analyses the context of the software problem from text and predicts a suitable design pattern for the given problem context using feature learning, neural embedding, and classification. To evaluate our approach, we make use of Stack Overflow posts, where developers often discuss design problems and consequences that they should consider, which are related to two main design pattern elements. We evaluate our approach on a case study from Stack Overflow with more than 66,000 questions that discuss problems and consequences related to 23 design patterns. The experimental evaluation shows that our approach can predict design patterns from the text with 82% overall accuracy. This indicates that our approach can successfully be used to support software designers in determining the most suitable design pattern for their software implementation.

Laksri Wijerathna

Monash University

Aldeida Aleti

Monash University

Australia

Workshop

Baber Majid Bhatti

University of South Australia

Australia

Workshop

In this paper, we have defined an NLP task, for the automatic extraction of business process redesign suggestions from natural language text. In particular, we have employed a systematic protocol to define the task, which is composed of three elements and three sub-tasks. The elements are: a) a real-world process model, b) actual feedback in natural language text, and c) three-level classification of the feedback. The task is composed of two binary and one multi-class classification sub-tasks. The evaluation of the AutoEPRS-20 task is performed using six traditional supervised learning techniques. The results show that the third sub-task is more challenging that the two binary sub-tasks.

Amina Mustansir

PUCIT, University of the Punjab

Khurram Shahzad

PUCIT, University of the Punjab

Muhammad Kamran Malik

PUCIT, University of the Punjab

Workshop

Emotion detection is playing a very important role in our life. People express their emotions in different ways i.e face expression, gestures, speech, and text. This research focuses on detecting emotions from the Roman Urdu text. Previously, A lot of work has been done on different languages for emotion detection but there is limited work done in Roman Urdu. Therefore, there is a need to explore Roman Urdu as it is the most widely used language on social media platforms for communication. One major issue for the Roman Urdu is the absence of benchmark corpora for emotion detection from text because language assets are essential for different natural language processing (NLP) tasks. There are many useful applications of the emotional analysis of a text such as improving the quality of products, dialog systems, investment trends, mental health. In this research, to focus on the emotional polarity of the Roman Urdu sentence we develop a comprehensive corpus of 18k sentences that are gathered from different domains and annotate it with six different classes. We applied different baseline algorithms like KNN, Decision tree, SVM, and Random Forest on our corpus. After experimentation and evaluation, the results showed that the SVM model achieves a better F-measure score.

Adil Majeed

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Hasan Mujtaba

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Mirza Omer Beg

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Workshop

In this paper, we have proposed novel concept of mapping natural language customer feedback text to relevant business process model elements. Customer feedback mapped over business process model will provide augmented business process having customer perception. More specifically, in this work, we have proposed systematic approach for mapping feedback comment to relevant process model elements which comprises a)process model generation, b) preparation of real-world customer feedback corpus, c) BPRI framework based mapping guidelines and d) first novel human annotated customer feedback process model element mapping dataset. We have evaluated the effectiveness of six traditional text similarity measures for automatic mapping of customer feedback to process model elements. Based on the results, we concluded that automatic mapping identification is challenging task as six traditional similarity measures resulted zero recall score.

Sanam Ahmed

Punjab University College of Information Technology , University of the Punjab

Amina Mustansir

PUCIT, University of the Punjab

Workshop

Social media, today, demonstrates the rapid growth of modern society as it becomes the main platform for Internet users to communicate and express themselves. People around the world, use a number of devices and resources to access the Internet, set up social networks, conduct online business, e-commerce, e-surveys, etc. Currently, social media is not only a technology that provides information to consumers, it also encourages users to connect and share their views and perspectives. It leads to an increase in inspiration towards Opinion Mining (OM), which is important for both customers and companies in making decisions. Individuals like to see the opinions provided by other customers about a particular product or a service. Companies need to analyze their customer’s feedback to strengthen their business decisions. A lot of research has been performed in various languages in the field of Aspect Based OM (ABOM). However, there are still certain languages that need to be explored, such as Roman Urdu (RU). This paper presents a proposed reviews data-set (a RU data-set) of mobile reviews that has been manually annotated with multi-aspect sentiment labels at the sentence-level. It presents base-line results using different Machine Learning (ML) algorithms. The results demonstrate 71% F1-score for aspect detection and 64% for aspect-based polarity.

Rabail Zahid

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Muhammad Owais Idrees

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Hasan Mujtaba

National University of Computer and Emerging Sciences, Islamabad, Pakistan

Mirza Omer Beg

National University of Computer and Emerging Sciences, Islamabad, Pakistan