A comparative analysis on question classification task based on deep learning approaches

Abstract

Question classification is one of the essential tasks for automatic question answering implementation in natural language processing (NLP). Recently, there have been several text-mining issues such as text classification, document categorization, web mining, sentiment analysis, and spam filtering that have been successfully achieved by deep learning approaches. In this study, we illustrated and investigated our work on certain deep learning approaches for question classification tasks in an extremely inflected Turkish language. In this study, we trained and tested the deep learning architectures on the questions dataset in Turkish. In addition to this, we used three main deep learning approaches (Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN)) and we also applied two different deep learning combinations of CNN-GRU and CNN-LSTM architectures. Furthermore, we applied the Word2vec technique with both skip-gram and CBOW methods for word embedding with various vector sizes on a large corpus composed of user questions. By comparing analysis, we conducted an experiment on deep learning architectures based on test and 10-cross fold validation accuracy. Experiment results were obtained to illustrate the effectiveness of various Word2vec techniques that have a considerable impact on the accuracy rate using different deep learning approaches. We attained an accuracy of 93.7% by using these techniques on the question dataset.

Stress Recognition in Code-Mixed Social Media Texts using Machine Learning

“…The research (Zulqarnain, et al, 2023) is focused on attention-aware deep learning approaches for an effective stress classification domain. The E-LSTM, which is an enhancement on traditional LSTM and combines pre-attention, resulted in superior performance when compared to other classification methods, including Naïve Bayesian, SVM, deep belief network, and standard LSTM.…”

An Improved Deep Neural Network Based on Combination of GRU and Auto Encoder for Sentiment Analysis

“…-Compute the update gate 𝒛 𝒕 , and reset gate 𝒓 𝒕 , using equations ( 8), (7) -Compute the candidate state ĥ 𝒕 , using equation ( 9) -while stopping criteria did not met do -while training for all instances do -Train a mini-batch dataset as network input.…”

“…In this proposed architecture, auto encoder layers were added with GRU network to enhance its ability for dimensionality reduction. Moreover, a learning algorithm of typical GRU architecture with AE layers is carried out two key phases: 1) To start, it effectively reduces the input features vector 𝑥, and provides the input raw features in the form of highly compressed presentation: 2) The output 𝑥̅ produced by equation ( 5) is then further computed 𝑥̅ as inputs to the GRU network, and evaluating by update gate, and reset gate based on the equations ( 8) and (7), represented in section 2.…”

“…Sentence identification is the primary task in many NLP applications, including internet text analysis [5], documents categorization, knowledge representation and spam filtering [6], which involve to allocates predetermined classes to consecutive words. In recent times, traditional machine learning based representation approaches have been widely used in sentence categorization, question-answering system [7] and some of them focus [8] on effective algorithms constructed features for creating effective sentiment classifiers [9].…”

Generic Question Classification for Dialogue Systems

“…In this case, the problem consists in finding a correspondence between questions and answers that are usually textual elements (sentences, summaries, etc.). Machine learning is very efficient for this type of problem [12,10,25], even though some rule-based approaches have also shown interesting results [14,6,15]. In this type of application, the problem consists in identifying fine-level class labels (and as a consequence a large number of classes) on a semantic basis, such labels being used as keys for accessing the answers [10].…”