This paper analyses the capabilities of different techniques to build a semantic representation of educational digital resources. Educational digital resources are modeled using the Learning Object Metadata (LOM) standard, and these semantic representations can be obtained from different LOM fields, like the title, description, among others, in order to extract the features/characteristics from the digital resources. The feature extraction methods used in this paper are the Best Matching 25 (BM25), the Latent Semantic Analysis (LSA), Doc2Vec, and the Latent Dirichlet allocation (LDA). The utilization of the features/descriptors generated by them are tested in three types of educational digital resources (scientific publications, learning objects, patents), a paraphrase corpus and two use cases: in an information retrieval context and in an educational recommendation system. For this analysis are used unsupervised metrics to determine the feature quality proposed by each one, which are two similarity functions and the entropy. In addition, the paper presents tests of the techniques for the classification of paraphrases. The experiments show that according to the type of content and metric, the performance of the feature extraction methods is very different; in some cases are better than the others, and in other cases is the inverse.

https://www.mdpi.com/2079-3197/8/2/30/pdf

Comparison and Evaluation of Different Methods for the Feature Extraction from Educational Contents

Microblog hot topic discovery is one of the research hotspots in the field of text mining. The distance function of traditional K-means leads to low clustering accuracy, which leads to poor hot topic discovery. Three definitions are proposed in this paper: title words and body words, positional contribution-based weight and fusion similarity-based distance. The short text clustering algorithm based on BTM and GloVe similarity linear fusion (BG & SLF-Kmeans) is further proposed. BTM and GloVe are used to model the preprocessed microblog short texts. JS divergence is adopted to calculate the text similarity based on BTM topic modeling. WMD of improved word weight (IWMD) is used to calculate the text similarity based on GloVe word vector modeling. Finally, the two similarities are linearly fused and used as the distance function to realize K-means clustering. Specific word sets of 6 hot topics can be obtained, and microblog hot topics can be discovered. The experimental results show that BG & SLF-Kmeans significantly improves clustering accuracy compared with TF-IDF & K-means, BTM & K-means, and BTF & SLF-Kmeans.

/pdf/btm-and-glove-similarity-linear-fusion-based-short-text-3gmj687glf.pdf

BTM and GloVe Similarity Linear Fusion-Based Short Text Clustering Algorithm for Microblog Hot Topic Discovery

Covid-19 has been a dangerous outbreak for the world that has lasted more than 2 years. Covid-19 has evolved or developed into several new variations, such as delta which is more dangerous than its initial variant. Vaccines became the world's solution to defend against Covid-19. In Indonesia, at the early stages of implementing mass vaccination programs, people had been involved in many pros and cons, to support or against the program. On social media such as Twitter, public opinions about vaccines are very diverse. This study investigates public sentiment towards the early stage of vaccination program conducted by the government. The classification method used in the sentiment analysis is the Support Vector Machine (SVM), among the positive, negative and neutral classes, with word embeddings extraction features. Data was collected and labeled by 12 crowd sourced annotators. The training and parameter tuning process was carried out to find the model that produced the best accuracy of validation data. From 400 testing data, the application of this optimal model resulted in an F1-score of 65% and an accuracy of 69%, higher than several machine learning methods in the same study.

Support Vector Machine Method with Word2vec for Covid-19 Vaccine Sentiment Classification on Twitter

Twitter was one of the most influential social media among users. It might give an either positive or negative impact. One of the negative impacts was the presence of radicalism content. In Indonesia radicalism was often connected to the issue of SARA (ethnicity, religion, race, and intergroup relations). It remained a public issue, requiring an analysis to process information related to radicalism. The research aimed to classify radical contents. The classification based on the types of radicalism and non-radicalism. Data were classified using LSTM. In finding higher accuracy, word2vec was used to transform words into vectors. The accuracy showed using LSTM method was compared with that obtained using SVM and k-NN. The two latest methods were the methods used by previous researchers regarding Indonesian radical contents of Twitter. Referring to the findings, LSTM showed higher accuracy 81.60%.

Classification of Radicalism Content from Twitter Written in Indonesian Language using Long Short Term Memory

Abstract With the globalization of data, online social media plays an active role in spreading information and classifying people, and thinking about how to break the solidification of algorithms becomes critical. Current algorithmic research in the social media space often focuses on a single service or language, mainly due to the lack of a way to connect the different bubbles. The panel speakers described their various research activities in which they presented different perspectives on how to break the bubble. This article provides a summary of this interactive panel.

https://sciendo.com/pdf/10.2478/dim-2020-0020

Breaking Social Media Bubbles for Information Globalization: A Cross-Cultural and Cross-Language User-Centered Sense-Making Approach

Development of information technology makes the production of data increase dramatically. We can get lots of data from the internet, including data reviews about a product or service. The more data obtained, the system is needed to process it. Sentiment analysis is a text processing of Natural Language Processing (NLP) that can help someone to see the quality of service offered, including hotel services. This paper uses hotel review data to carry out sentiment analysis obtained from the Traveloka website. The data classified using the Long Short-Term Memory (LSTM) algorithm. To get better results, the authors use word embedding to convert words into vectors. This study aims to compare the performance of several word embedding, while word embedding compared is word2vec Continuous Bag of Words CBOW, word2vec skip-gram, doc2vec, and glove. From the experiment conducted, the results show that the glove method has the highest accuracy of 95.52% while the word2vec skip-gram model has the lowest accuracy of 91.81%, so it concluded that the glove method is the best word embedding method for hotel review data.

Widyawan

Papers

Word Embedding Comparison for Indonesian Language Sentiment Analysis