Best Practices of Convolutional Neural Networks for Question Classification

Abstract

<jats:p>Question Classification (QC) is of primary importance in question answering systems, since it enables extraction of the correct answer type. State-of-the-art solutions for short text classification obtained remarkable results by Convolutional Neural Networks (CNNs). However, implementing such models requires choices, usually based on subjective experience, or on rare works comparing different settings for general text classification, while peculiar solutions should be individuated for QC task, depending on language and on dataset size. Therefore, this work aims at suggesting best practices for QC using CNNs. Different datasets were employed: (i) A multilingual set of labelled questions to evaluate the dependence of optimal settings on language; (ii) a large, widely used dataset for validation and comparison. Numerous experiments were executed, to perform a multivariate analysis, for evaluating statistical significance and influence on QC performance of all the factors (regarding text representation, architectural characteristics, and learning hyperparameters) and some of their interactions, and for finding the most appropriate strategies for QC. Results show the influence of CNN settings on performance. Optimal settings were found depending on language. Tests on different data validated the optimization performed, and confirmed the transferability of the best settings. Comparisons to configurations suggested by previous works highlight the best classification accuracy by those optimized here. These findings can suggest the best choices to configure a CNN for QC.</jats:p>

Automatic question-answer pairs generation and question similarity mechanism in question answering system

“…Systems like [35] pick up the keywords from the question and paragraph and match them using RNN. Pota et al, [27] used Convolution Neural Networks (CNNs) to classify the questions. The question classification plays a vital role in extracting the correct answer in the Question Answering System.…”

An Effective BERT-Based Pipeline for Twitter Sentiment Analysis: A Case Study in Italian

“…Here a single token can be viewed as a vector by using word embedding, hence a 2D matrix represents the generic sentence. The most famous CNN-based sentiment analysis model was introduced by [ 39 ], extensively used by [ 40 ] and enhanced by [ 41 ]. Furthermore, Chen et al [ 8 ] improved sentiment detection through a two-steps architecture, leveraging separated CNNs trained on sentences clustered according to the number of opinion targets contained.…”

“…Regarding the Italian scene, the number of annotated corpora is much lower, although recent work on the classification of texts focused on approaches based on deep learning [ 75 ], possibly starting from models pre-trained on large unlabeled resources. With regard to the scope of analysis of tweets collected by the social network Twitter, it is possible to identify several problems due to differences in structure and grammar compared to plain text.…”

A Subword-Based Deep Learning Approach for Sentiment Analysis of Political Tweets

“…The architecture and the techniques used for training its parameters are described below. Architecture arrangements and hyperparameters are settled by taking into account previous work, where optimal settings are found for sentence classification and question classification [11][12][13]. Moreover, particular settings are chosen, matching the peculiarities of the tweet classification problem.…”