Natural Language Processing
M. Heydari; A. Albadvi; M. Khazeni
Abstract
Background and Objectives: The lack of a suitable tool for the analysis of conversational texts in Persian language has made various analyzes of these texts, including Sentiment Analysis, difficult. In this research, it has we tried to make the understanding of these texts easier for the machine by providing ...
Read More
Background and Objectives: The lack of a suitable tool for the analysis of conversational texts in Persian language has made various analyzes of these texts, including Sentiment Analysis, difficult. In this research, it has we tried to make the understanding of these texts easier for the machine by providing PSC, Persian Slang Convertor, a tool for converting conversational texts into formal ones, and by using the most up-to-date and best deep learning methods along with the PSC, the sentiment learning of short Persian language texts for the machine in a better way.Methods: Be made More than 10 million unlabeled texts from various social networks and movie subtitles (as dialogue texts) and about 10 million news texts (as official texts) have been used for training unsupervised models and formal implementation of the tool. 60,000 texts from the comments of Instagram social network users with positive, negative, and neutral labels are considered as supervised data for training the emotion classification model of short texts. The latest methods such as LSTM, CNN, BERT, ELMo, and deep processing techniques such as learning rate decay, regularization, and dropout have been used. LSTM has been utilized in the research, and the best accuracy has been achieved using this method.Results: Using the official tool, 57% of the words of the corpus of conversation were converted. Finally, by using the formalizer, FastText model and deep LSTM network, the accuracy of 81.91 was obtained on the test data.Conclusion: In this research, an attempt was made to pre-train models using unlabeled data, and in some cases, existing pre-trained models such as ParsBERT were used. Then, a model was implemented to classify the Sentiment of Persian short texts using labeled data.
Artificial Intelligence
S. Nemati
Abstract
Background and Objectives: Twitter is a microblogging platform for expressing assessments, opinions, and sentiments on different topics and events. While there have been several studies around sentiment analysis of tweets and their popularity in the form of the number of retweets, predicting the sentiment ...
Read More
Background and Objectives: Twitter is a microblogging platform for expressing assessments, opinions, and sentiments on different topics and events. While there have been several studies around sentiment analysis of tweets and their popularity in the form of the number of retweets, predicting the sentiment of first-order replies remained a neglected challenge. Predicting the sentiment of tweet replies is helpful for both users and enterprises. In this study, we define a novel problem; given just a tweet's text, the goal is to predict the overall sentiment polarity of its upcoming replies.Methods: To address this problem, we proposed a graph convolutional neural network model that exploits the text's dependencies. The proposed model contains two parallel branches. The first branch extracts the contextual representation of the input tweets. The second branch extracts the structural and semantic information from tweets. Specifically, a Bi-LSTM network and a self-attention layer are used in the first layer for extracting syntactical relations, and an affective knowledge-enhanced dependency tree is used in the second branch for extracting semantic relations. Moreover, a graph convolutional network is used on the top of these branches to learn the joint feature representation. Finally, a retrieval-based attention mechanism is used on the output of the graph convolutional network for learning essential features from the final affective picture of tweets.Results: In the experiments, we only used the original tweets of the RETWEET dataset for training the models and ignored the replies of the tweets in the training process. The results on three versions of the RETWEET dataset showed that the proposed model outperforms the LSTM-based models and similar state-of-the-art graph convolutional network models. Conclusion: The proposed model showed promising results in confirming that by using only the content of a tweet, we can predict the overall sentiment of its replies. Moreover, the results showed that the proposed model achieves similar or comparable results with simpler deep models when trained on a public tweet dataset such as ACL 2014 dataset while outperforming both simple deep models and state-of-the-art graph convolutional deep models when trained on the RETWEET dataset. This shows the proposed model's effectiveness in extracting structural and semantic relations in the tweets.
