SAShanza Abbas

Shanza Abbas

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Shanza Abbas has a total of 6 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A Review of NLIDB With Deep Learning: Findings, Challenges and Open Issues

    by Shanza Abbas, Muhammad Umair Khan, Scott Uk-Jin Lee, Asad Abbas, Ali Kashif Bashir

    Relational databases are storage for a massive amount of data. Knowledge of structured query language is a prior requirement to access that data. That is not possible for all non-technical personals, leading to the need for a system that translates text to SQL query itself rather than the user. Text to SQL task is also crucial because of its economic and industrial value. Natural Language Interface to Database (NLIDB) is the system that supports the text-to-SQL task. Developing the NLIDB system is a long-standing problem. Previously they were built based on domain-specific ontologies via pipelining methods. Recently a rising variety of Deep learning ideas and techniques brought this area to the attention again. Now end to end Deep learning models is being proposed for the task. Some publicly available datasets are being used for experimentation of the contributions, making the comparison process convenient. In this paper, we review the current work, summarize the research trends, and highlight challenging issues of NLIDB with Deep learning models. We discussed the importance of datasets, prediction model approaches and open challenges. In addition, methods and techniques are also summarized, along with their influence on the overall structure and performance of NLIDB systems. This paper can help future researchers start having prior knowledge of findings and challenges in NLIDB with Deep learning approaches.


  • Open Access

    Columns Occurrences Graph to Improve Column Prediction in Deep Learning Nlidb

    by Shanza Abbas, Muhammad Umair Khan, Scott Uk-Jin Lee, Asad Abbas

    Natural language interfaces to databases (NLIDB) has been a research topic for a decade. Significant data collections are available in the form of databases. To utilize them for research purposes, a system that can translate a natural language query into a structured one can make a huge difference. Efforts toward such systems have been made with pipelining methods for more than a decade. Natural language processing techniques integrated with data science methods are researched as pipelining NLIDB systems. With significant advancements in machine learning and natural language processing, NLIDB with deep learning has emerged as a new research trend in this area. Deep learning has shown potential for rapid growth and improvement in text-to-SQL tasks. In deep learning NLIDB, closing the semantic gap in predicting users’ intended columns has arisen as one of the critical and fundamental problems in this research field. Contributions toward this issue have consisted of preprocessed feature inputs and encoding schema elements afore of and more impactful to the targeted model. Various significant work contributed towards this problem notwithstanding, this has been shown to be one of the critical issues for the task of developing NLIDB. Working towards closing the semantic gap between user intention and predicted columns, we present an approach for deep learning text-to-SQL tasks that includes previous columns’ occurrences scores as an additional input feature. Overall exact match accuracy can also be improved by emphasizing the improvement of columns’ prediction accuracy, which depends significantly on column prediction itself. For this purpose, we extract the query fragments from previous queries’ data and obtain the columns’ occurrences and co-occurrences scores. Column occurrences and co-occurrences scores are processed as input features for the encoder–decoder-based text to the SQL model. These scores contribute, as a factor, the probability of having already used columns and tables together in the query history. We experimented with our approach on the currently popular text-to-SQL dataset Spider. Spider is a complex data set containing multiple databases. This dataset includes query–question pairs along with schema information. We compared our exact match accuracy performance with a base model using their test and training data splits. It outperformed the base model’s accuracy, and accuracy was further boosted in experiments with the pretrained language model BERT.


  • Open Access

    Wake Lock Leak Detection in Android Apps Using Multi-Layer Perceptron

    by Muhammad Umair Khan, Scott Uk-Jin Lee, Zhiqiang Wu, Shanza Abbas

    With the proliferation of mobile devices, the popularity of Android applications (apps) has increased exponentially. Efficient power consumption in a device is essential from the perspective of the user because users want their devices to work all day. Developers must properly utilize the application programming interfaces (APIs) provided by Android software development kit to optimize the power consumption of their app. Occasionally, developers fail to relinquish the resources required by their app, resulting in a resource leak. Wake lock APIs are used in apps to manage the power state of the Android smartphone, and they frequently consume more power than necessary if not used appropriately (also called energy leak). In this study, we use a multi-layer perceptron (MLP) to detect wake lock leaks in Android apps because the MLP can solve complex problems and determine similarities in graphs. To detect wake lock leaks, we extract the call graph as features from the APK and embed the instruction and neighbor information in the node’s label of the call graph. Then, the encoded data are input to an MLP model for training and testing. We demonstrate that our model can identify wake lock leaks in apps with 99% accuracy.


  • Open Access

    Detecting Wake Lock Leaks in Android Apps Using Machine Learning

    by Muhammad Umair Khan, Scott Uk-Jin Lee, Shanza Abbas, Asad Abbas, Ali Kashif Bashir

    The popularity of Android devices has increased exponentially with an increase in the number of mobile devices. Millions of online apps are used in these devices. Energy consumption of a device is a major concern for end-users, who want a long usage time on a single battery charge. The energy consumed by the app must be optimized by developers, and the available APIs must be used carefully. A wake-lock is used in apps to control the power state of the Android device and often leads to energy leakage. In this study, we detected wake-lock leaks in Android apps using machine learning. We pre-processed apps by extracting wake-lock related APIs to obtain the structural information of wake-lock usage and oversampled the data using the synthetic minority oversampling technique (SMOTE) to balance the dataset. The machine learning algorithms used to detect wake-lock leaks were first optimized using grid search to determine the best parameters. These parameters were then used in training to detect wake-lock leaks in these apps. We employed various machine learning algorithms and divided them into simple and ensemble algorithms to evaluate their efficacy. The support vector machine (SVM) and stochastic gradient boosting (SGB) were the most effective, producing 97 % and 98 % accuracy, respectively.