Tracking Hate in Social Media: Evaluation, Challenges and Approaches

Modha, Sandip; Mandl, Thomas; Majumder, Prasenjit; Patel, Daksh

doi:10.1007/s42979-020-0082-0

Cited by 10 publications

(7 citation statements)

References 26 publications

(39 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the event though, of extending machine learning usage on various tasks, some comparison with other machine learning algorithms might need to be performed, especially when considering the applicability of the approach to every possible seismic event occurrence. Even, research demonstrates that neural networks could have some comparative advantages, while used for classifying short text of specific topics [28,36] there are many algorithms that perform very well. Support vector machines (SVMs) are widely used for text classification [2,3,12,32,35] while there is demonstrated credibility in topics regarding natural disasters [30,31].…”

Section: Discussion: Conclusionmentioning

confidence: 99%

Social Network Users Create Seismic Intensity Maps: An Automatic Approach of the Methodology

Arapostathis

2021

SN COMPUT. SCI.

View full text Add to dashboard Cite

Main purpose of current article is to present current state-of-the-art, regarding a methodology for producing seismic intensity maps directly and exclusively from twitter data. The methodology had been presented previously, however, the vast majority of the processing procedures were based on manual work. In current state, an automatic approach is presented, aspiring thus to minimize the processing time needed which is quite enormous considering manual analytic processes. In specific, word patterns that had been identified, were used, to initially select tweet texts that contain macroseismic observations and for classifying them according to the corresponding values of the European macroseismic intensity scale (EMS98). As a next step machine learning was employed through the use of a long short term memory-recurrent neural network (LSTM-RNN) model, trained for removing ambiguity of potentially wrongly classified data. Sequentially, the classified information was geo-referenced by detecting geolocations contained within each tweet's text, while the precision of each georeference was indicated with a related value. Two georules improve the accuracy of the processed output, completing thus the extraction of macroseismic observations. Sequentially, the processed dataset was interpolated to create seismic intensity prediction surfaces at a raster format. The applied technique was the simple kriging, widely used for the creation of seismic intensity maps. As case study, the earthquake event of SW of Lesvos (M w = 6.3, 2017, Greece) was used. The output, was consisted of four seismic intensity maps, produced from macroseismic observations extracted from twitter, posted within 2, 6, 24 and total data period, respectively, since the earthquake event occurrence. The results were assessed through a comparison with official seismic intensity maps of the same earthquake published by seismological institutes and reality. Through the assessment, the main conclusions include significant steps towards the potential operational use of the methodology, as the produced output was close to reality while in some cases was able to be more precise than other approaches. Moreover, the time needed for data processing and for creating the final maps, has been limited drastically. Those facts emerge the potentials of integrating social network data approaches even in the most specialized scientific fields, like seismic intensity mapping.

show abstract

Section: Discussion: Conclusionmentioning

confidence: 99%

Social Network Users Create Seismic Intensity Maps: An Automatic Approach of the Methodology

Arapostathis

2021

SN COMPUT. SCI.

View full text Add to dashboard Cite

show abstract

“…In this section, we will discuss the previous state of the art methods proposed for detection of hate speech. The use of BERT and other transfer learning algorithms, and deep neural models based on LSTMs and CNNs tend to perform similar but better than traditional classifiers such as SVM [11]. The number of papers, trying to automate Hate speech detection, that have been published in Web of Science has been increasing exponentially [12].…”

Section: Related Workmentioning

confidence: 99%

Multilingual Hate Speech and Offensive Content Detection using Modified Cross-entropy Loss

Mitra¹,

Sankhala²

2022

Preprint

View full text Add to dashboard Cite

The number of increased social media users has led to a lot of people misusing these platforms to spread offensive content and use hate speech. Manual tracking the vast amount of posts is impractical so it is necessary to devise automated methods to identify them quickly. Large language models are trained on a lot of data and they also make use of contextual embeddings. We fine-tune the large language models to help in our task. The data is also quite unbalanced; so we used a modified cross-entropy loss to tackle the issue. We observed that using a model which is fine-tuned in hindi corpora performs better. Our team (HNLP) achieved the macro F1-scores of 0.808, 0.639 in English Subtask A and English Subtask B respectively. For Hindi Subtask A, Hindi Subtask B our team achieved macro F1-scores of 0.737, 0.443 respectively in HASOC 2021.

show abstract

“…The accuracy measure in the case of multi-class classification is calculated according to (19). Where tp i + tn i is the number of all instances which are predicted correctly,i is the index of the class , l is the total number of class labels, and f n i + tp i + f p i + tn i is the total number of instances in the given dataset.…”

Section: F Alsen Egativeratementioning

confidence: 99%

“…In particular, SVM, NB, DTs, RF, and LR methods, etc. which are used extensively with high accuracy in wide application fields that include sentiment analysis, such as cyberhate detection [19] movie and product reviews [20], [21], abusive language detection [22], cyberbullying identification [23], and social media [24]. In addition to classical ML algorithms as presented earlier, there are likewise DL algorithms such as CNN, FFNN, LSTM, GRU, and RNN, which are presently preferred for sentiment classification.…”

mentioning

confidence: 99%

Sentence-Level Classification Using Parallel Fuzzy Deep Learning Classifier

et al. 2021

View full text Add to dashboard Cite

At present, with the growing number of Web 2.0 platforms such as Instagram, Facebook, and Twitter, users honestly communicate their opinions and ideas about events, services, and products. Owing to this rise in the number of social platforms and their extensive use by people, enormous amounts of data are produced hourly. However, sentiment analysis or opinion mining is considered as a useful tool that aims to extract the emotion and attitude from the user-posted data on social media platforms by using different computational methods to linguistic terms and various Natural Language Processing (NLP). Therefore, enhancing text sentiment classification accuracy has become feasible, and an interesting research area for many community researchers. In this study, a new Fuzzy Deep Learning Classifier (FDLC) is suggested for improving the performance of data-sentiment classification. Our proposed FDLC integrates Convolutional Neural Network (CNN) to build an effective automatic process for extracting the features from collected unstructured data and Feedforward Neural Network (FFNN) to compute both positive and negative sentimental scores. Then, we used the Mamdani Fuzzy System (MFS) as a fuzzy classifier to classify the outcomes of the two used deep (CNN+FFNN) learning models in three classes, which are: Neutral, Negative, and Positive. Also, to prevent the long execution time taking by our hybrid proposed FDLC, we have implemented our proposal under the Hadoop cluster. An experimental comparative study between our FDLC and some other suggestions from the literature is performed to demonstrate our offered classifier's effectiveness. The empirical result proved that our FDLC performs better than other classifiers in terms of true positive rate, true negative rate, false positive rate, false negative rate, error rate, precision, classification rate, kappa statistic, F1-score and time consumption, complexity, convergence, and stability.

show abstract

Tracking Hate in Social Media: Evaluation, Challenges and Approaches

Cited by 10 publications

References 26 publications

Social Network Users Create Seismic Intensity Maps: An Automatic Approach of the Methodology

Social Network Users Create Seismic Intensity Maps: An Automatic Approach of the Methodology

Multilingual Hate Speech and Offensive Content Detection using Modified Cross-entropy Loss

Sentence-Level Classification Using Parallel Fuzzy Deep Learning Classifier

Contact Info

Product

Resources

About