Quantifying the Natural Sentiment Strength of Polar Term Senses Using Semantic Gloss Information and Degree Adverbs

Darwich, Mohammad; Noah, Shahrul Azman Mohd; Omar, Nazlia; Osman, Nurul Aida; Ahmad, Ibrahim Said

doi:10.12720/jait.11.3.109-118

Cited by 24 publications

(2 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on the fact that rich vocabulary can also solve the problem of term sparsity to enrich short text features, Wang [12] et al used a non-linear sliding approach and a N-gram language model to obtain rich short text features that were fed into a convolutional neural network to improve the accuracy in short text classification. As well as Mohammad [13] et al quantified the sentiment intensity of human-defined lexical words and degree adverbs in the text to generate a classification model that matches human intuition. In addition to enriching short text features, a large number of fusion classification models have been proposed to improve classification accuracy in recent years.…”

Section: Related Workmentioning

confidence: 99%

An Improved Text Classification Model Based on Part of Speech Features

Liu

2023

Preprint

View full text Add to dashboard Cite

The most important thing in short text classification is to extract text features. The traditional method of static training word vector represents text features, which has some problems such as insufficient semantics and sparse features. However, the dynamic training word vector also lacks the incomplete expression of part-of-speech features. Therefore, this paper proposes a classification model PB-DPCNN (Part-of-speech based BERT-DPCNN). Part-of-speech tagging introduces part-of-speech features of text, introduces part-of-speech features to construct part-of-speech vectors with part-of-speech attributes, dynamically trains word vectors with BERT model, and uses the obtained word vectors with part-of-speech information as input layer information to improve text representation. The word vector format is fixed by embedding layer, and the long-distance pattern is captured by double-layer convolution of equal length to improve the richness of word embedding. The output layer is combined with the residual pooling to further enhance the lexical perception space, and the connection layer obtains the final text feature representation. Compared with other models in the THUCNews data set, the experimental results show that the model improves each classification index and has good text semantic recognition ability.

show abstract

Section: Related Workmentioning

confidence: 99%

An Improved Text Classification Model Based on Part of Speech Features

Liu

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Generally, there are two approaches to sentiment analysis: the unsupervised-learning (lexicon-based) and the supervised-learning (machine learning-based) approaches. The unsupervised-learning approach involves the use of a sentiment lexicon to determine the sentiment value of a text, while the supervised-learning approach involves the use of machine learning classifiers to classify the texts into different classes of sentiment (Darwich et al , 2020).…”

Section: Literature Reviewmentioning

confidence: 99%

Sequel movie revenue prediction model based on sentiment analysis

Ahmad

Bakar

Yaakub

et al. 2020

DTA

Self Cite

View full text Add to dashboard Cite

PurposeSequel movies are very popular; however, there are limited studies on sequel movie revenue prediction. The purpose of this paper is to propose a sentiment analysis based model for sequel movie revenue prediction and to propose a missing value imputation method for the sequel revenue prediction dataset.Design/methodology/approachA sequel of a successful movie will most likely also be successful. Therefore, we propose a supervised learning approach in which data are created from sequel movies to predict the box-office revenue of an upcoming sequel. The algorithms used in the prediction are multiple linear regression, support vector machine and multilayer perceptron neural network.FindingsThe results show that using four sequel movies in a franchise to predict the box-office revenue of a fifth sequel achieved better prediction than using three sequels, which was also better than using two sequel movies.Research limitations/implicationsThe model produced will be beneficial to movie producers and other stakeholders in the movie industry in deciding the viability of producing a movie sequel.Originality/valuePrevious studies do not give priority to sequel movies in movie revenue prediction. Additionally, a new missing value imputation method was introduced. Finally, sequel movie revenue prediction dataset was prepared.

show abstract