Research Article
A Proposed Keyword-Based Feature Extraction Approach for Labeling and Classifying Egyptian Mobile Apps Arabic Slang User Requirements Reviews
@INPROCEEDINGS{10.1007/978-3-031-33614-0_2, author={Rabab Emad Saudy and Alaa El Din El-Ghazaly and Eman S. Nasr and Mervat H. Gheith}, title={A Proposed Keyword-Based Feature Extraction Approach for Labeling and Classifying Egyptian Mobile Apps Arabic Slang User Requirements Reviews}, proceedings={Big Data Technologies and Applications. 11th and 12th EAI International Conference, BDTA 2021 and BDTA 2022, Virtual Event, December 2021 and 2022, Proceedings}, proceedings_a={BDTA}, year={2023}, month={5}, keywords={Mobile Applications Mobile Applications Arabic Slang Reviews User Requirements Functional Requirements Non-Functional Requirements Sentimental Requirements Term Frequency -- Inverse Document Frequency Bag of Words Natural Language Processing Classifier Chains Machine Learning Deep Learning Logistic Regression Random Forest Neural Network}, doi={10.1007/978-3-031-33614-0_2} }- Rabab Emad Saudy
Alaa El Din El-Ghazaly
Eman S. Nasr
Mervat H. Gheith
Year: 2023
A Proposed Keyword-Based Feature Extraction Approach for Labeling and Classifying Egyptian Mobile Apps Arabic Slang User Requirements Reviews
BDTA
Springer
DOI: 10.1007/978-3-031-33614-0_2
Abstract
Mobile applications (apps) review feature is supplied by most mobile apps platforms, which authorize users to evaluate, comment, and rate apps after utilizing it. User reviews are identified as an oriental source to enhance mobile applications (apps) and raise the importance for users. With the acute rise in the quantity of reviews, how to functionally and efficiently analyze and mining the user reviews and recognize serious user requirements from them to enhance the mobile apps. In this paper, we suggest an automatic approach for identifying and classifying requirements into Functional Requirements (FR), Non-Functional Requirements (NFR) and Sentimental Requirements (SR) from Egyptian Mobile Apps Arabic Slang Reviews (MASR), utilizing a group of techniques Term Frequency – Inverse Document Frequency (TF-IDF), Bag of Words (BOW) and Natural Language Processing (NLP) techniques with keywords selection. We suggest applying Classifier Chains (CC) approach to convert classifying multi-labeled data problem into one or more problems of single labeling, and utilizing the hybrid stack classification model, which combines Machine Learning (ML) and Deep Learning (DL) approaches consist of Logistic Regression (LR), Random Forest (RF), and Multi-Layer Perceptron Neural Network (MLP-NN). The hybrid stack classification model accomplishes high accuracy results for classifying Egyptian MASR user requirements as follow: (99.7%) for classifying Performance, (99.5%) for classifying Dissatisfied Users, (98.8%) for classifying Others, and (98.1%) for classifying Security, (97.9%) for classifying Usability, and (97.4%) for classifying Feature Requests.
