Language Independent Features Research Articles

Language-Independent Acoustic Biomarkers for Quantifying Speech Impairment in Huntington's Disease.

Changes in voice and speech are characteristic symptoms of Huntington's disease (HD). Objective methods for quantifying speech impairment that can be used across languages could facilitate assessment of disease progression and intervention strategies. The aim of this study was to analyze acoustic features to identify language-independent features that could be used to quantify speech dysfunction in English-, Spanish-, and Polish-speaking participants with HD. Ninety participants with HD and 83 control participants performed sustained vowel, syllable repetition, and reading passage tasks recorded with previously validated methods using mobile devices. Language-independent features that differed between HD and controls were identified. Principal component analysis (PCA) and unsupervised clustering were applied to the language-independent features of the HD data set to identify subgroups within the HD data. Forty-six language-independent acoustic features that were significantly different between control participants and participants with HD were identified. Following dimensionality reduction using PCA, four speech clusters were identified in the HD data set. Unified Huntington's Disease Rating Scale (UHDRS) total motor score, total functional capacity, and composite UHDRS were significantly different for pairwise comparisons of subgroups. The percentage of HD participants with higher dysarthria score and disease stage also increased across clusters. The results support the application of acoustic features to objectively quantify speech impairment and disease severity in HD in multilanguage studies. https://doi.org/10.23641/asha.25447171.

Spider Plus: A Text Classifier for Research Article Components

Uploading research articles to a database can be a complex process with multiple HTML fields. The complexity of this process makes users less efficient and productive. This study explored the use of the Multilayer Perceptron (MLP) algorithm using language-independent features (count and Boolean features) to create a multi-class text classifier that can classify 10 research article components (e.g., title, authors, abstract, etc.). The text classifier was developed to pave the way for a streamlined manual data entry process for uploading research articles to a database by implementing a single textarea in favor of using multiple HTML fields. The text samples were obtained from multiple sources using web scraping technology, consolidated, cleaned, and standardized. The 12 language-independent features were generated based on the textual formats of sample texts (e.g., count of capitalized letters, digits, punctuation, checking the existence of a URL pattern, etc.). Recursive feature elimination with cross validation (RFECV) was used to determine the optimal number of input features. The hyperparameter values of the model were determined through the grid search technique. A trial-and-error process was conducted to determine the number of hidden layers. The model achieved 95%, 94%, and 95% scores for micro, macro, and weighted average f1-scores, respectively. The model performed well in classifying research article components. However, it is sensitive to textual format (e.g., lower or upper case, punctuation used, etc.). For subsequent research in this area, this study recommends investigating the use of both language-dependent and language-independent features to address the limitations of the current model.

Enhanced Emotion Recognition from Spoken Assamese Dialect: A Machine Learning Approach with Language-Independent Features

Enhanced Emotion Recognition from Spoken Assamese Dialect: A Machine Learning Approach with Language-Independent Features

A New Lightweight Script Independent Scene Text Style Transfer Network

Scene text style transfer without a language barrier is an open challenge for the video and scene text recognition community because this plays a vital role in poster, web design, augmenting character images, and editing characters to improve scene text recognition performance and usability. This work presents a new model, called Script Independent Scene Text Style Transfer Network (SISTSTNet), for extracting scene characters and transferring text style simultaneously. The SISTSTNet performs mapping in language-independent feature space for transferring style. It is designed based on a Style Parameter Network and Target Encoder Network through lightweight MobileNetv3 convolutional and residual blocks to capture the style and shape to generate target characters. Similarly, a generative model is explored through the Visual Geometry Group (VGG) network for character replacement. The SISTSTNet is flexible and works on different languages and arbitrary examples in a neat and unified fashion. The experimental results on images in various languages, namely, English, Chinese, Hindi, Russian, Japanese, Arabic, Greek, and Bengali and cross-language validation demonstrate the effectiveness of the proposed method. The performance of the method is superior compared to the state-of-the-art methods in terms of quality measures, language independence, shape-preserving, and efficiency. The code and dataset will be released to the public to support reproducibility.

Robust and language-independent acoustic features in Parkinson's disease.

The analysis of vocal samples from patients with Parkinson's disease (PDP) can be relevant in supporting early diagnosis and disease monitoring. Intriguingly, speech analysis embeds several complexities influenced by speaker characteristics (e.g., gender and language) and recording conditions (e.g., professional microphones or smartphones, supervised, or non-supervised data collection). Moreover, the set of vocal tasks performed, such as sustained phonation, reading text, or monologue, strongly affects the speech dimension investigated, the feature extracted, and, as a consequence, the performance of the overall algorithm. We employed six datasets, including a cohort of 176 Healthy Control (HC) participants and 178 PDP from different nationalities (i.e., Italian, Spanish, Czech), recorded in variable scenarios through various devices (i.e., professional microphones and smartphones), and performing several speech exercises (i.e., vowel phonation, sentence repetition). Aiming to identify the effectiveness of different vocal tasks and the trustworthiness of features independent of external co-factors such as language, gender, and data collection modality, we performed several intra- and inter-corpora statistical analyses. In addition, we compared the performance of different feature selection and classification models to evaluate the most robust and performing pipeline. According to our results, the combined use of sustained phonation and sentence repetition should be preferred over a single exercise. As for the set of features, the Mel Frequency Cepstral Coefficients demonstrated to be among the most effective parameters in discriminating between HC and PDP, also in the presence of heterogeneous languages and acquisition techniques. Even though preliminary, the results of this work can be exploited to define a speech protocol that can effectively capture vocal alterations while minimizing the effort required to the patient. Moreover, the statistical analysis identified a set of features minimally dependent on gender, language, and recording modalities. This discloses the feasibility of extensive cross-corpora tests to develop robust and reliable tools for disease monitoring and staging and PDP follow-up.

A deep learning approach to building a framework for Urdu POS and NER

The study examines various studies on Named Entity Recognition (NER) and Part of Speech (POS) tagging for the Urdu language conducted by academics and researchers. POS and NER tagging for Urdu still faces obstacles in terms of increasing accuracy while lowering false-positive rates and labelling unknown terms, despite the efforts of numerous researchers. In addition, ambiguity exists when tagging terms with different contextual meanings within a sentence. Due to the fact that Urdu is an inflectional, derivational, morphologically rich, and context-sensitive language, the existing models, such as Linguistic rule application, N-gram Markov model, Tree Tagger, random forest (RF) tagger, etc., were unable to produce accurate experimental results on Urdu language data. The significance of this study is that it fills a gap in the literature concerning the lack of POS and NER tagging for the Urdu language. For Urdu POS and NER tagging, we propose a deep learning model with a well-balanced set of language-independent features as well as a survey of important Urdu POS/NER techniques. In addition, this is the first study to use residual biDirectional residual Long short-term memory (residual biLSTM) architecture trained on the Urmono dataset in conjunction with the randomly initialised word2vec, fastText and mBERT embeddings are utilised to generate word or character vectors.For each experiment, the paper also employs the evaluation methods of Macro-F1, precision, precision, and recall. The proposed method with mbert embedding as word vectors provides best results of F1 score for POS and NER at 91.11% and 99.11% respectively. Also, the accuracy, precision and recall for POS are reported at 94.85%, 91.79% and 90.77%. Similarly, the accuracy, precision and recall for NER of the proposed model are reported at 99.77%, 98.78% and 99.45% respectively, which are higher than baseline models.

Just-In-Time Defect Prediction on JavaScript Projects: A Replication Study

Change-level defect prediction is widely referred to as just-in-time (JIT) defect prediction since it identifies a defect-inducing change at the check-in time, and researchers have proposed many approaches based on the language-independent change-level features. These approaches can be divided into two types: supervised approaches and unsupervised approaches, and their effectiveness has been verified on Java or C++ projects. However, whether the language-independent change-level features can effectively identify the defects of JavaScript projects is still unknown. Additionally, many researches have confirmed that supervised approaches outperform unsupervised approaches on Java or C++ projects when considering inspection effort. However, whether supervised JIT defect prediction approaches can still perform best on JavaScript projects is still unknown. Lastly, prior proposed change-level features are programming language–independent, whether programming language–specific change-level features can further improve the performance of JIT approaches on identifying defect-prone changes is also unknown. To address the aforementioned gap in knowledge, in this article, we collect and label the top-20 most starred JavaScript projects on GitHub. JavaScript is an extremely popular and widely used programming language in the industry. We propose five JavaScript-specific change-level features and conduct a large-scale empirical study (i.e., involving a total of 176,902 changes) and find that (1) supervised JIT defect prediction approaches (i.e., CBS+) still statistically significantly outperform unsupervised approaches on JavaScript projects when considering inspection effort; (2) JavaScript-specific change-level features can further improve the performance of approach built with language-independent features on identifying defect-prone changes; (3) the change-level features in the dimension of size (i.e., LT), diffusion (i.e., NF), and JavaScript-specific (i.e., SO and TC) are the most important features for indicating the defect-proneness of a change on JavaScript projects; and (4) project-related features (i.e., Stars, Branches, Def Ratio, Changes, Files, Defective, and Forks) have a high association with the probability of a change to be a defect-prone one on JavaScript projects.

Named Entity Recognition Using Conditional Random Fields

Named entity recognition (NER) is an important task in natural language processing, as it is widely featured as a key information extraction sub-task with numerous application areas. A plethora of attempts was made for NER detection in Western and Asian languages. However, little effort has been made to develop techniques for the Urdu language, which is a prominent South Asian language with hundreds of millions of speakers across the globe. NER in Urdu is considered a hard problem owing to several reasons, including the paucity of large, annotated datasets; an inaccurate tokenizer; and the absence of capitalization in the Urdu language. To this end, this study proposed a conditional-random-field-based technique with both language-dependent and language-independent features, such as part-of-speech tags and context windows of words, respectively. As a second contribution, we developed an Urdu NER dataset (UNER-I) in which a large number of NE types were manually annotated. To evaluate the effectiveness of the proposed approach, as well as the usefulness of the dataset, experiments were performed using the dataset we developed and an existing dataset. The results of the experiments showed that our proposed technique outperformed the baseline technique for both datasets by improving the F1 scores by 1.5% to 3%. Furthermore, the results demonstrated that the enhanced dataset was useful for learning and prediction in a supervised learning approach.

Push-to-Trend: A Novel Framework to Detect Trend Promoters in Trending Hashtags

Twitter trends have enabled the speedy dissemination of information with the ability to affect public opinion. Unfortunately, fake trends are also generated by malicious users to mislead the public. In general, Twitter users are studied in depth to identify humans, bots, spam, and fake accounts. However, artificial intelligence algorithms are not developed for the identification of ‘trend promoters’ generating fake trends. In this paper, we propose Push-To-Trend – a novel framework to detect ‘trend promoters’ in trending hashtags. For this purpose, first, we develop a dataset of TREP-21 containing 3,900 users labelled into two categories of ‘trend promoters’ and ‘normal users’. In addition, we design four discerning features of number of total tweets, duplicate tweets, overlapping ngram, and peak-to-mean ratio for trend promoters classification. Moreover, we thoroughly examine the features used for spam and bot accounts classification to filter three efficacious features for trend promoters identification. Leveraging these seven features, Push-To-Trend achieves the accuracy of 0.97 for TREP-21. Furthermore, we leverage our framework to identify and analyze trend promoters from the Urdu tweets repository “Anbar” which consists of 106.9 million tweets and 1.69 million users. The analysis of 602 most frequent hashtags in Anbar reveals that 15.7% of trend promoters generate 68.1% of total tweets related to hashtags. To the best of our knowledge, this is the first attempt to design machine learning models for the automatic classification of trend promoters. As such, our framework is generic and adaptable for tweets posted in different natural languages as it utilizes language-independent features.

A Transformer-Based Approach to Multilingual Fake News Detection in Low-Resource Languages

Fake news classification is one of the most interesting problems that has attracted huge attention to the researchers of artificial intelligence, natural language processing, and machine learning (ML). Most of the current works on fake news detection are in the English language, and hence this has limited its widespread usability, especially outside the English literate population. Although there has been a growth in multilingual web content, fake news classification in low-resource languages is still a challenge due to the non-availability of an annotated corpus and tools. This article proposes an effective neural model based on the multilingual Bidirectional Encoder Representations from Transformer (BERT) for domain-agnostic multilingual fake news classification. Large varieties of experiments, including language-specific and domain-specific settings, are conducted. The proposed model achieves high accuracy in domain-specific and domain-agnostic experiments, and it also outperforms the current state-of-the-art models. We perform experiments on zero-shot settings to assess the effectiveness of language-agnostic feature transfer across different languages, showing encouraging results. Cross-domain transfer experiments are also performed to assess language-independent feature transfer of the model. We also offer a multilingual multidomain fake news detection dataset of five languages and seven different domains that could be useful for the research and development in resource-scarce scenarios.

Automatic Text Summarization of Konkani Folk Tales Using Supervised Machine Learning Algorithms and Language Independent Features

Automatic text summarization is an emerging field of research in Natural Language Processing. This work is a novel attempt to include a low-resource language to the domain of Automatic Text Summarization. We use supervised machine learning algorithms to perform single document extractive automatic text summarization on documents in a low-resource language, Konkani. In particular, we propose using language independent features to train supervised machine learning algorithms using a Konkani dataset, specifically devised for the experimentation using books on Konkani folktale literature. We approach the automatic text summarization task as a binary classification problem, and the algorithms, once trained, classify the sentences based on their relevance to generate a summary. Thereafter, the performance of popular linear and non-linear supervised machine learning algorithms is evaluated using K-fold cross-validation. The summary generated by the systems is compared with human-generated summaries to verify its effectiveness. The results show that the linear models exhibit better performance in comparison with the non-linear models; however, all the models could beat the baselines. The output produced by the proposed methodology generates promising summaries without the need for any language-specific domain knowledge.

Language-Independent Fake News Detection: English, Portuguese, and Spanish Mutual Features

Online Social Media (OSM) have been substantially transforming the process of spreading news, improving its speed, and reducing barriers toward reaching out to a broad audience. However, OSM are very limited in providing mechanisms to check the credibility of news propagated through their structure. The majority of studies on automatic fake news detection are restricted to English documents, with few works evaluating other languages, and none comparing language-independent characteristics. Moreover, the spreading of deceptive news tends to be a worldwide problem; therefore, this work evaluates textual features that are not tied to a specific language when describing textual data for detecting news. Corpora of news written in American English, Brazilian Portuguese, and Spanish were explored to study complexity, stylometric, and psychological text features. The extracted features support the detection of fake, legitimate, and satirical news. We compared four machine learning algorithms (k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGB)) to induce the detection model. Results show our proposed language-independent features are successful in describing fake, satirical, and legitimate news across three different languages, with an average detection accuracy of 85.3% with RF.

Real-Time Bot Detection from Twitter Using the Twitterbot+ Framework

Nowadays, bot detection from Twitter attracts the attention of several researchers around the world. Different bot detection approaches have been proposed as a result of these research efforts. Four of the main challenges faced in this context are the diversity of types of content propagated throughout Twitter, the problem inherent to the text, the lack of sufficient labeled datasets and the fact that the current bot detection approaches are not sufficient to detect bot activities accurately. We propose, Twitterbot+, a bot detection system that leveraged a minimal number of language-independent features extracted from one single tweet with temporal enrichment of a previously labeled datasets. We conducted experiments on three benchmark datasets with standard evaluation scenarios, and the achieved results demonstrate the efficiency of Twitterbot+ against the state-of-the-art. This yielded a promising accuracy results (>95%). Our proposition is suitable for accurate and real-time use in a Twitter data collection step as an initial filtering technique to improve the quality of research data.

Suicidal tendencies prediction in Greek poetry

Natural language processing (NLP) has been successfully used to predict a writer’s tendency of committing suicide, using various text types: suicide notes, micro-blog posts, lyrics and poems. This paper is an extended version of earlier work. We extend our previous work on text mining in Greek Poetry by employing more sophisticated approaches. More specifically we have applied (i) Deep Neural Networks (DNN), (ii) additional morphosyntactic and semantic features based on writers' emotions and Big Five personality traits and (iii) feature selection, for suicide prediction in Greek poetry. We extend previous research to Greek, i.e. a language that has not been tackled before in this setting, using both language-dependent (but easily portable across languages) and language-independent linguistic features in order to represent the poems of 13 Greek poets of the twentieth century. Our results differ significantly from previous literature. In general, our proposed DNN model offers promising results for suicide prediction, despite the fact that this task poses multiple difficulties, especially for a language with limited related research support.

A Novel Approach to Conditional Random Field-based Named Entity Recognition using Persian Specific Features

Named Entity Recognition is an information extraction technique that identifies name entities in a text. Three popular methods have been conventionally used namely: rule-based, machine-learning-based and hybrid of them to extract named entities from a text. Machine-learning-based methods have good performance in the Persian language if they are trained with good features. To get good performance in Conditional Random Field-based Persian Named Entity Recognition, a several syntactic features based on dependency grammar along with some morphological and language-independent features have been designed in order to extract suitable features for the learning phase. In this implementation, designed features have been applied to Conditional Random Field to build our model. To evaluate our system, the Persian syntactic dependency Treebank with about 30,000 sentences, prepared in NOOR Islamic science computer research center, has been implemented. This Treebank has Named-Entity tags, such as Person, Organization and location. The result of this study showed that our approach achieved 86.86% precision, 80.29% recall and 83.44% F-measure which are relatively higher than those values reported for other Persian NER methods.

MaldomDetector: A system for detecting algorithmically generated domain names with machine learning

One of the leading problems in cyber security at present is the unceasing emergence of sophisticated attacks, such as botnets and ransomware, that rely heavily on Command and Control (C&C) channels to conduct their malicious activities remotely. To avoid channel detection, attackers constantly try to create different covert communication techniques. One such technique is Domain Generation Algorithm (DGA), which allows malware to generate numerous domain names until it finds its corresponding C&C server. It is highly resilient to detection systems and reverse engineering, while allowing the C&C server to have several redundant domain names. This paper presents a malicious domain name detection system, MaldomDetector, which is based on machine learning. It is capable of detecting DGA-based communications and circumventing the attack before it makes any successful connection with the C&C server, using only domain name's characters. MaldomDetector uses a set of easy-to-compute and language-independent features in addition to a deterministic algorithm to detect malicious domains. The experimental results demonstrate that MaldomDetector can operate efficiently as a first alarm to detect DGA-based domains of malware families while maintaining high detection accuracy.

Spam profiles detection on social networks using computational intelligence methods: The effect of the lingual context

In online social networks, spam profiles represent one of the most serious security threats over the Internet; if they do not stop producing bad advertisements, they can be exploited by criminals for various purposes. This article addresses the nature and the characteristics of spam profiles in a social network like Twitter to improve spam detection, based on a number of publicly available language-independent features. In order to investigate the effectiveness of these features in spam detection, four datasets are extracted for four different language contexts (i.e. Arabic, English, Korean and Spanish), and a fifth is formed by combining them all. We conduct our experiments using a set of five well-known classification algorithms in spam detection field, k-Nearest Neighbours ( k-NN), Random Forest (RF), Naive Bayes (NB), Decision Tree (DT) (J48) and Multilayer Perceptron (MLP) classifiers, along with five filter-based feature selection methods, namely, Information Gain, Chi-square, ReliefF, Correlation and Significance. The results show oscillating performance of each classifier across all datasets, but improved classification results with feature selection. In addition, detailed analysis and comparisons are carried out on two different levels: in the first level, we compare the selected features’ importance among the feature selection methods, whereas in the second level, we observe the relations and the importance of the selected features across all datasets. The findings of this article lead to a better understanding of social spam and improving detection methods by considering the various important features resulting from the different lingual contexts.

Organization, Bot, or Human: Towards an Efficient Twitter User Classification

Today, through Twitter, researchers propose approaches for classifying user accounts. However, they have to face confidence challenges owing to the diversity of the types of data propagated throughout Twitter. In addition, the messages from Twitter are imprecise, very short and even written in many dialects and languages. Moreover, the majority of the related works focus on the overall user’s activity, which makes them not suitable at the post-level classification. This paper presents an alternative approach for classifying user accounts as being accounts of bots, humans or organizations. The suggested approach consists in accurately classifying user accounts from one single post by leveraging a minimal number of language-independent features. We performed several experiments over a Twitter datasets and supervised learn-based algorithms. Our results demonstrated that simply using a minimal number of language-independent features extracted from one single post is sufficient to classify user accounts accurately and quickly. Our proposed approach yielded high F1-measure (>95%) and high AUC (>99%) using Random Forest.

Funnelling

Cross-lingual Text Classification (CLC) consists of automatically classifying, according to a common set C of classes, documents each written in one of a set of languages L , and doing so more accurately than when “naïvely” classifying each document via its corresponding language-specific classifier. To obtain an increase in the classification accuracy for a given language, the system thus needs to also leverage the training examples written in the other languages. We tackle “multilabel” CLC via funnelling , a new ensemble learning method that we propose here. Funnelling consists of generating a two-tier classification system where all documents, irrespective of language, are classified by the same (second-tier) classifier. For this classifier, all documents are represented in a common, language-independent feature space consisting of the posterior probabilities generated by first-tier, language-dependent classifiers. This allows the classification of all test documents, of any language, to benefit from the information present in all training documents, of any language. We present substantial experiments, run on publicly available multilingual text collections, in which funnelling is shown to significantly outperform a number of state-of-the-art baselines. All code and datasets (in vector form) are made publicly available.

Part of Speech Tagging in Urdu: Comparison of Machine and Deep Learning Approaches

In Urdu, part of speech (POS) tagging is a challenging task as it is both inflectionally and derivationally rich morphological language. Verbs are generally conceived a highly inflected object in Urdu comparatively to nouns. POS tagging is used as a preliminary linguistic text analysis in diverse natural language processing domains such as speech processing, information extraction, machine translation, and others. It is a task that first identifies appropriate syntactic categories for each word in running text and second assigns the predicted syntactic tag to all concerned words. The current work is the extension of our previous work. Previously, we presented conditional random field (CRF)-based POS tagger with both language dependent and independent feature set. However, in the current study, we offer: 1) the implementation of both machine and deep learning models for Urdu POS tagging task with well-balanced language-independent feature set and 2) to highlight diverse challenges which cause Urdu POS task a challenging one. In this research, we demonstrated the effectiveness of machine learning and deep learning models for Urdu POS task. Empirically, we have evaluated the performance of all models on two benchmark datasets. The core models evaluated in this study are CRF, support vector machine (SVM), two variants of the deep recurrent neural network (DRNN), and a variant of n-gram Markov model the bigram hidden Markov model (HMM). The two variants of DRRN models evaluated include forward long short-term memory (LSTM)-RNN and LSTM-RNN with CRF output.