With the increase in social networks, more number of people are creating and sharing information than ever before, many of them have no relevance to reality. Due to this, fake news for various political and commercial purposes are spreading quickly. Online newspaper has made it challenging to identify trustworthy news sources. In this work, Hindi news articles from various news sources are collected. Preprocessing, feature extraction, classification and prediction processes are discussed in detail. Different machine learning algorithms such as Naïve Bayes, logistic regression and Long Short-Term Memory (LSTM) are used to detect the fake news. The preprocessing step includes data cleaning, stop words removal, tokenizing and stemming. Term frequency inverse document frequency(TF-IDF) is used for feature extraction. Naïve Bayes, logistic regression and LSTM classifiers are used and compared for fake news detection with probability of truth. It is observed that among these three classifiers, LSTM achieved best accuracy of 92.36%. 

Fake News Detection on Hindi News Dataset

Change is the only constant. In many sectors, a change is being witnessed that is getting increasingly rapid. This carries a plethora of new innovation possibilities with it. This necessitates well-founded data about trends, future developments and their consequences. This study seeks to catch the new directions, paradigms as predictors with an association of each topic which will be discovered through topic modeling techniques like LDA with BoW. For this, empirical analysis on 3269 research articles from the Journal of Applied Intelligence was done, which were gathered during a 30-year span. The inferred topics were then structured into a way suitable for performing predictive analysis. This is significant in the sense that it will help to predict what technology will be encountered in the future, as well as how far human's ability to innovate and discover things may lead this world to. The final model using TF-IDF scores has outperformed the baseline model by a margin of 41%. 

Prediction of Research Trends using LDA based Topic Modeling

Learning the syntax and structure of command-line commands is of utmost importance in the field of cyber security to identify valid and malicious sets of commands. It is hard to learn the syntax and structure of every command because of various reasons, such as the continuous evolution of commands, precise syntax requirement, huge volume of available commands, and no room for errors, etc. In this research work, we studied two approaches to learning the structure of the commands by detecting the random tokens in them, such as temp files, temp directories, numerical values, etc. In the first approach, we write hard-coded regular expressions to identify random tokens in a command whereas in the second approach we trained a second-order Markov model which detects the random tokens based on their probabilities. To validate the efficiency of these approaches, we clustered the commands using their word embeddings and sentence embeddings. For clustering, we explored KMeans, and DBScan with word embeddings and sentence clustering based on sentence embeddings. We evaluated the performance of clustering algorithms against three metrics, the Silhouette Coefficient, the Calinski-Harabasz Index, and the Davies-Bouldin Index. The results show that regular expression and the Markov model achieve the same scores for KMeans and DBScan based on word embeddings against three metrics, whereas when clustered using sentence embeddings, the Markov model performs better than regular expression. These results validate our idea of using the Markov model instead of regular expressions, to get similar scores or even better performance with less resource utilization, such as human effort, time to write regular expressions, and maintaining & storage of those regular expressions.

/pdf/learning-the-structure-of-commands-by-detecting-random-1njsc3fn.pdf

Learning the Structure of Commands by Detecting Random Tokens Using Markov Model

In the contemporary era of technology, individuals who utilize mobile phones and laptops have developed a preference for accessing news through online media. News organizations disseminate news and offer confirmation sources. However, the issue at hand is how to authenticate stories and articles shared on social networks such as WhatsApp groups, Facebook pages, Twitter, and other smaller blogs and social networking sites. It is hazardous for society to accept rumours disguised as news, especially in developing nations like India, where it is crucial to prevent rumours and specialize in honest and verified information. Classifying written articles as misleading or deceptive is not easy to automate, and even experts in a specific field must evaluate several factors before rendering a judgment regarding the validity of a message. This project proposes the use of a machine learning approach to automatically classify news articles. This endeavour explores numerous text characteristics that can be employed to differentiate fabricated news content from actual news.

/pdf/interactive-web-app-for-fake-news-detection-2xmvevpm.pdf

Interactive Web App for Fake News Detection

With the proliferation of internet, more individuals are creating and sharing knowledge due to the growth of social networks. However, many of these pieces of information have little to do with reality. This has led to the rapid dissemination of false information used for various political, personal and business objectives. Finding reliable news sources has become more difficult, thanks to online newspapers. This paper presents an approach to detect fake news headline in the Hindi language. For this work, dataset was manually collected and it comprises of more than 7300 news headlines. After pre-processing the dataset, features were extracted using Term Frequency (TF) and Term Frequency-Inverse Document Frequency (TF-IDF). 12 different machine learning models trained and tested for this research work include Logistic Regression, K-Nearest Neighbour, Support Vector Machine with linear kernel, Multinomial Naive Bayes, Decision Tree, Random Forest, XG Boost, Bagging, Gradient Boosting, Artificial Neural Network, Convolutional Neural Network, and Long Short-Term Memory. Based on the top-3 performers from traditional model, an innovative Soft Voting Based Hybrid (SVBH) model was proposed. Accuracy reported by the proposed model with TF weighing scheme is 98.26%, which is better as compared to traditional ones.

A Novel Soft Voting Based Hybrid Approach to Detect Fake News in Hindi

The exponential growth in fake news and its role in deteriorating general public trust and democratic standards certainly calls for some counter combat approaches. The prediction of chances of news to be fake is deemed to be hard task since most of the deceptive news has its roots in true news. With a minor fabrication in legitimate news, influential fake news can be created that can be used for political, entertainment, or business-related gains. This work provides a novel intuitive approach to exploit data from multiple sources to segregate news into real and fake. To efficiently capture the contextual information present in the data, Bidirectional Encoder Representations from Transformer (BERT) have been deployed. It attempts to further enhance the performance of the deceptive news detection model by incorporating information about the speaker profile and the credibility associated with him/her. A hybrid sequence encoding model has been proposed to harvest the speaker profile and speaker credibility data which makes it useful for prediction. On evaluation over benchmark fake news dataset LIAR, our model outperformed the previous state-of-the-art works. This attests to the fact that the speaker’s profile and credibility play a crucial role in predicting the validity of news.

Debunking Fake News by Leveraging Speaker Credibility and BERT Based Model

The command line has always been the most efficient method to interact with UNIX flavor based systems while offering a great deal of flexibility and efficiency as preferred by professionals. Such a system is based on manually inputting commands to instruct the computing machine to carry out tasks as desired. This human-computer interface is quite tedious especially for a beginner. And hence, the command line has not been able to garner an overwhelming reception from new users. Therefore, to improve user-friendliness and to mark a step towards a more intuitive command line system, we propose two predictive approaches that can benefit all kinds of users specially the novice ones by integrating into the command line interface. These methods are based on deep learning based predictions. The first approach is based on the sequence to sequence (Seq2seq) model with joint learning by leveraging continuous representations of a self-curated exhaustive knowledge base (KB) comprising an all-inclusive command description to enhance the embedding employed in the model. The other is based on the attention-based transformer architecture where a pretrained model is employed. This allows the model to dynamically evolve over time making it adaptable to different circumstances by learning as the system is being used. To reinforce our idea, we have experimented with our models on three major publicly available Unix command line datasets and have achieved benchmark results using GLoVe and Word2Vec embeddings. Our finding is that the transformer based framework performs better on two different datasets of the three in our experiment in a semantic deficit scenario like UNIX command line prediction. However, Seq2seq based model outperforms bidirectional encoder representations from transformers (BERT) based model on a larger dataset.

Predictive approaches for the UNIX command line: curating and exploiting domain knowledge in semantics deficit data

Despite being an open-source operating system pioneered in the early 90s, UNIX based platforms have not been able to garner an overwhelming reception from amateur end users One of the rationales for under popularity of UNIX based systems is the steep learning curve corresponding to them due to extensive use of command line interface instead of usual interactive graphical user interface In past years, the majority of insights used to explore the concern are eminently centered around the notion of utilizing chronic log history of the user to make the prediction of successive command The approaches directed at anatomization of this notion are predominantly in accordance with Probabilistic inference models The techniques employed in past, however, have not been competent enough to address the predicament as legitimately as anticipated Instead of deploying usual mechanism of recommendation systems, we have employed a simple yet novel approach of Seq2seq model by leveraging continuous representations of self-curated exhaustive Knowledge Base (KB) to enhance the embedding employed in the model This work describes an assistive, adaptive and dynamic way of enhancing UNIX command line prediction systems Experimental methods state that our model has achieved accuracy surpassing mixture of other techniques and adaptive command line interface mechanism as acclaimed in the past

Seq2Seq and Joint Learning Based Unix Command Line Prediction System

Text classification has become a key operation in various natural language processing tasks. The efficiency of most classification algorithms predominantly confide in the quality of input features. In this work, we propose a novel multi-class text classification technique that harvests features from two distinct feature extraction methods. Firstly, a structured heterogeneous text graph built based on document-word relations and word co-occurrences is leveraged using a Graph Convolution Network (GCN). Secondly, the documents are topic modeled to use the document-topic score as features into the classification model. The concerned graph is constructed using Point-Wise Mutual Information (PMI) between pair of word co-occurrences and Term Frequency-Inverse Document Frequency (TF-IDF) score for words in the documents for word co-occurrences. Experimentation reveals that our text classification model outperforms the existing techniques for five benchmark text classification data sets.

A Hybrid Classification Approach using Topic Modeling and Graph Convolution Networks

Drug repositioning offers an economical and efficient alternative to traditional drug discovery. It means that, a drug approved for effect against a particular disease is considered and its applications for novel pharmaceutical purposes are explored in shorter development timelines. Unlike conventional approaches, this work attempts to explore the network of existing drugs and its unmapped indications by treating drug repositioning as a classification problem. The proposed classification model attempts estimation of the relevance of a drug with an unmapped indication. An enhanced word representation model is used for this purpose by integrating knowledge obtained from a structured biological knowledge graph and medical literature. To harvest the structured biological data, we have leveraged multiple biological ontologies to achieve a formal framework in the form of a semantic knowledge graph. Our novelty lies in that we have exploited knowledge from biological knowledge graph and medical corpora to complement each other. This makes our method competent with well established drug repositioning techniques.

Apoorva Vikram Singh

Papers

Debunking Fake News by Leveraging Speaker Credibility and BERT Based Model

Predictive approaches for the UNIX command line: curating and exploiting domain knowledge in semantics deficit data

Seq2Seq and Joint Learning Based Unix Command Line Prediction System

A Hybrid Classification Approach using Topic Modeling and Graph Convolution Networks

Towards Better Drug Repositioning Using Joint Learning