Accepted Papers

Single-channel Speech Separation Focusing on Attentionde

Xinshu Li, Zhenhua Tan, Zhenche Xia, Danke Wu, Bin Zhang, Department of Software Engineering, Northeastern University, Shenyang, China

ABSTRACT

In recent multi-speaker speech separation researches, the overall deep-learning based architecture consists of three parts: encoder, separator and decoder. But improvement strategies generally only focus on the separator in the middle, regardless of its input. The most common encoder structure at present is asingle 1D convolution layer followed by a nonlinear activation function, ReLU. In this paper, we firstlypropose a new encoder named Attention DE, trying to improve the input ef ectiveness of separator. Thenew encoder adds an extra 1D convolutional layers and incorporates the multi-head attention mechanismto enhance the feature aggregation ability of input speech. Secondly, based on Gated DPRNN, our newseparator use SepFormer Blocks to improve the training ef iciency and learn the speech sequence patterns better. Experiments show that our proposed method has achieved an advanced SI-SNRi of 20.3 dBonWSJ0-2MIX, which belongs to the advanced level of single-channel speech separation at present.

KEYWORDS

Single-channel, Deep-learning, Encoder, Attention, Transformer.

Pf-Net: Personalized Fliter for Speaker Recognition from Raw Wavefoem

Wencheng Li, Zhenhua Tan*, Jing Yu Ning, ZhenChe Xia, Danke Wu, Bin Zhang, Department of Information Safety, Northeastern University, Shen Yang, China

ABSTRACT

Speaker recognition using i-vector has been replaced by speaker recognition using deep learning. Speaker recognition based on Convolutional Neural Networks (CNNs) has been widely used in recent years, which learn low-level speech representations from raw waveforms. On this basis, a CNN architecture called SincNet proposes a kind of unique convolutional layer, which has achieved band-pass filters. Compared with standard CNNs, SincNet learns the low and high cut-off frequencies of each filter. This paper proposes an improved CNNs architecture called PF-Net, which encourages the first convolutional layer to implement more personalized filters than SincNet. PF-Net parameterizes the frequency domain shape and can realize band-pass filters by learning some deformation points in frequency domain. Compared with standard CNN, PF-Net can learn the characteristics of each filter. Compared with SincNet, PF-Net can learn more characteristic parameters, instead of only low and high cut-off frequencies. This provides a personalized filter bank for different tasks. As a result, our experiments show that the PF-Net converges faster than standard CNN and performs better than SincNet. Our codes will be publicity on github.com/tanopenlab.

KEYWORDS

Speaker Recognition, Raw Waveform, Personalized Filters, Deep learning.

Applied Monocular Reconstruction of Parametric Faces with Domain Engineering

Igor Borovikov, Karine Levonyan, Jon Rein, Pawel Wrotek and Nitish Victor, Electronic Arts, Redwood City, CA, USA

ABSTRACT

Many modern online 3D applications and videogames rely on parametric models of human faces for creating believable avatars. However, manual reproduction of someones facial likeness with a parametric model is difficult and time-consuming. Machine Learning solution for that task is highly desirable but is also challenging. The paper proposes a novel approach to the so-called Face-to-Parameters problem (F2P for short), aiming to reconstruct a parametric face from a single image. The proposed method utilizes synthetic data, domain decomposition, and domain adaptation for addressing multifaceted challenges in solving the F2P. The open-sourced codebase illustrates our key observations and provides means for quantitative evaluation. The presented approach proves practical in an industrial application; it improves accuracy and allows for more efficient models training. The techniques have the potential to extend to other types of parametric models.

KEYWORDS

Face Reconstruction, Parametric Models, Domain Decomposition, Domain Adaptation.

A CNN-based Approach for Multi-Classification of Brain Tumors

Sahiti Nallamolu, Hritik Nandanwar, Anurag Singh, Subalalitha C.N., Department of Computer Science and Engineering, SRM Institute of Science & Technology, Chennai, India

ABSTRACT

Early diagnosis of brain tumor plays an important factor in extending the life expectancy of a patient. Therefore, an accurate and timely diagnosis of the type of brain tumor will allow adequate treatment planning and medical assistance. Radiologists commonly use magnetic resonance imaging (MRI) scans to detect and classify brain tumors. The current methods used in the medical field for diagnosis are timeconsuming and prone to human error. In recent years, researchers have developed automated techniques for the segmentation and classification of MRI images resulting in a faster diagnosis process. Recent advancements in deep learning have shown greater efficiency in image recognition and classification tasks. In this paper, a convolutional neural network (CNN) (a widely used deep learning architecture for image classification tasks) is developed to classify MRI images into four brain tumor categories. Data augmentation is applied to the training dataset to generalize the images and avoid overfitting problem. Additionally, this paper compares the performance of various pre-trained models such as Vision Transformer (VIT), VGG19, ResNet50, Inception V3, and AlexNet50 with that of the proposed model. Each experiment then explores transfer learning techniques like fine-tuning and freezing layers. In the study, the proposed model yields the most efficient results with a classification accuracy of 94.72%.

KEYWORDS

Brain Tumor, Deep Learning, Convolutional Neural Network, Transfer Learning, Magnetic Resonance Imaging.

Cognitive graphical password based on recognition with improved user functionality

Mozhdeh Sarkhoshi and Qianmu Li, Nanjing University of Science and Technology, Nanjing, Jiangsu, China

ABSTRACT

The fact that photos and graphics are more easily recalled by humans than text led to the proposal that visual passwords may be a viable alternative to text passwords in certain situations. User-friendliness characteristics of existing models are based on graphical password recognition, and the introduction of a new model that is related to the specifications and features of ISO standard usability and to the specifications and features of general usability specifications and features is being considered. Once these criteria and characteristics and sub-components of usability had been compared, additional usability features that could be included into the new graphical password model provided were discovered. There was a presentation of the graphical password model, which was separated into two sections, which included new users and current users. A questionnaire was used to evaluate the usability features and applicability of the prototype system after it had been implemented as a prototype system. After this step, the system was implemented as a prototype system and its evaluation and evaluation through a questionnaire was used to evaluate the usability features and applicability of the prototype system. Then there will be user input on the whole system, as well as the outcomes. The characteristics and specifications of the usability of the visual password prototype will be gathered and examined in this study. All of the percentages collected in this publication in connection to the findings and results from the point of view of usability are such that it is possible to conclude that the new visual password system is acceptable in its current form.

KEYWORDS

graphic password, authentication, usability.

Phase Difference Based Doppler Disambiguation Method for TDM-MIMO FMCW Radars

Qingshan Shen and Qingbo Wang, College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China

ABSTRACT

State-of-the-art automotive radar sensors use a Mutliple-Input Mutiple-Output (MIMO) approach to obtain a better angular resolution. Time-Division Multiplexing (TDM) scheme is commonly applied to realize the orthogonality in time at the transmitter. Apart from its simplicity in implementation, TDM scheme has the drawback of a reduced maximum unambiguous doppler proportional to the number of transmitters. In this paper, a phase difference based doppler disambiguation method is proposed to regain the maximum unambiguous doppler which is equivalent to only one transmitter. This method works well when the number of transmitters is large. The proposed method is demonstrated with simulation and measurement data.

KEYWORDS

Doppler disambiguation, TDM, MIMO, FMCW, Phase difference.

Expert Systems Generating Machine for Image Processing Applications

Maan Ammar¹, Khuzama Ammar², Kinan Mansour³, Waad Ammar⁴, ¹Al Andalus University for medical sciences. Al Kadmous, Syria, ²Damascus University Hospital, Damascus, Syria, ³Al Andalus University Hospital, Al Kadmous, Syria, ⁴Zain Al Abedeen Hospital, Karbalaa, Iraq

ABSTRACT

We introduce in this paper what can be considered a new trend in expert systems field. It is generating different expert systems using the same software platform developed for this purpose, and called “Expert Systems Generating Machine for Image Processing Applications ESGMIPA”. The machine is used to generate different expert systems in completely different application fields which indicates the feasibility of the proposal. Using what we called Domain Expert Guided Heuristic Search (DEGHS) and the machine, we generated an expert system that succeeded in cases where no algorithmic approach can be applied. Generating different expert systems using the same machine depends on the well-known fact that the function of an expert system is determined mainly by its knowledge base. The machine developed expedite very much the development of the expert system¬¬ to reach best performance. The role of domain expert and the positive effect of the interaction between different domain experts in different fields is highlighted.

KEYWORDS

expert systems generating machine, expert guided heuristic search, handwriting extraction, bacteria type automatic detection, bacteria colony image.

The acquisition of English Suprasegmental pronunciation: Computer-assisted Language Learning Approach*

Smirkou Mohamed, Department of English, Faculty of Languages and Arts, Ibn Tofail University, Kenitra, Morocco

ABSTRACT

The aim of this experimental study is to establish Computer-assisted Language Learning (CALL) in teaching pronunciation. English pronunciation is traditionally taught through drills and hinges on teachers’ feedback on learners’ errors. To demonstrate computer-aided teaching, this study proposes Praat(meaning “talk” in Dutch)as a pedagogical tool to teaching suprasegmentals , particularly word-stress. This paper addresses the problem of stress misassignment and exploring how such prosodic features are further enhanced through visual feedback.By visually presenting word-stress, the auditory information is converted into visual information, hence achieving successful decoding. Pronunciation is often taught through the oral/aural medium.Praat opens up analysis to the visual medium and visual measurement of acoustic properties (duration, pitch, and intensity of sounds).Forty Moroccan learners participated in the study. The subjects under investigation are semester-one students at Ibn Tofail University, Kenitra. They were randomly split into two groups: the experimental group received a treatment (instructed using Praat) while the control group did not.The participants first sit for the pre-test, then the experimental group was instructed using Praat, and finally they sit for a post-test. T-test was employed to analyze the data. The findings reveal that Sig. (2-tailed) score was 0.004, smaller than the significance value 0.05 (i.e., p<

KEYWORDS

Praat, CALL, speech visualization, suprasegmentals, word stress, teaching pronunciation, Moroccan EFL learners.

Towards an Adaptive Cybersecurity Training (ACST) Framework for Social Media

Fai Ben Salamah, Dr Marco Palomino, Dr Maria Papadaki and Professor Steve Furnell University of Plymouth , School of Engineering, computing and Mathematics (Faculty of Science and Engineering)

ABSTRACT

Social media has become embedded in our everyday lives, our personal activities, and the workplace. Thus, the need to educate users on emerging cybersecurity challenges for social media has become imperative. As such, we have set out to investigate the feasibility and effectiveness of an awareness-raising and adaptive cybersecurity training system. Our investigation follows a combined qualitative and quantitative approach aided by a questionnaire that was administered online using Google Forms. We collected answers from 640 Kuwaiti employees in a variety of sectors: education, healthcare, leadership and management, arts, entertainment, and the police and military. We found that a one-fits-all training approach is highly ineffective, as peoples understanding and knowledge can vary greatly, and these are factors that influence the success of the approach. In other words, backgrounds, preferences, and perceptions of the trainees are essential considerations for developing a robust training system.

KEYWORDS

Cybersecurity, Education, Training, Social media.

Adult Learners’ Perceptions of Integrating Workplace Learning and Technical Higher Education Learning Using Technology for Their Professional Development

James Elikana Mmari, Eötvös Loránd University, Hungary

ABSTRACT

Educational technology has so far supported teaching and learning in many ways one cannot deny. This study focuses on exploring the perceptions of educators, employers and learners in technical higher education, on how they perceive the use of technology for attaining and achieving required course outcomes, working skills as well as individual’s professional development. A semi-structured interview was administered to these stakeholders. It was revealed that technology itself poses several challenges in its adaptation, especially in developing countries. While learners were more open and welcoming to the idea of integrating workplace learning and their ongoing studies, educators revealed their worries on the readiness in adapting technology from the decision making bodies, educators to the learners. Apart from giving insightful results for the further large-scale study, this pilot study has also helped the researchers to measure and check the validity of the data collection tool.

KEYWORDS

educational technology, workplace learning, higher technical education, professional development.

Flood Prediction using Ensemble Machine Learning Model

Miah Mohammad Asif Syeed¹, Maisha Farzana¹, Ishadie Namir¹, Ipshita Ishrar¹, Meherin Hossain Nushra¹ and Tanvir Rahman², ¹Department of Computer Science and Engineering, BRAC University, Dhaka, Bangladesh, ²Department of Computer and Information Sciences, College of Engineering, University of Delaware, Delaware, United States of America

ABSTRACT

Floods are one of natures most catastrophic calamities which cause irreversible and immense damage to human life, agriculture, infrastructure and socio-economic system. Several studies on flood catastrophe management and flood forecasting systems have been conducted. The accurate prediction of the onset and progression of floods in real time is challenging. To estimate water levels and velocities across a large area, it is necessary to combine data with computationally demanding flood propagation models. This paper aims to reduce the extreme risks of this natural disaster and also contributes to policy suggestions by providing a prediction for floods using different machine learning models. This research will use Binary Logistic Regression, K-Nearest Neighbor (KNN), Support Vector Classifier (SVC), Decision tree Classifier and Stacked Generalization (Stacking) to provide an accurate prediction. With the outcome, a comparative analysis will be conducted to understand which model delivers a better accuracy.

KEYWORDS

Binary Logistic Regression, Stacked Generalization, Ensemble Machine Learning, Flood Prediction.

Bayesian Belief Network for Media Mixed Models

Cathy Zhao, Media Lab, Google, New York City, New York, USA

ABSTRACT

Econometric approaches such as time series cross-sectional models are widely adopted to estimate impacts of various marketing time series data on sales. Such multivariate linear regressions (or pooled regression) typically ignore the nonlinear relationship between the media and outcome. It is often excessively complex, and runs into overfitting problems. We are taking a machine learning approach to fit multidimensional data for media mix problems. Our core model is Bayesian belief network (BBN) which identifies the direct and indirect relationships between marketing, non-marketing and business performance variables, or the long-term effects of media, such as the long term impact of branding. Furthermore, we enhance the parameter estimates with another machine learning method, Kernel-based Regularized Least squares (KRLS) regression. The model works well to minimize overfitting, diminishing the influence of “bad leverage” points and efficiency gain even for a small sample size. Our approach does not rely on linearity or additivity assumptions, providing a holistic view about what we get from our media investment on sales, across channels, regions and campaigns.

KEYWORDS

Machine learning, Big data analytics, Bayesian networks, Graphic probabilistic model.

FunReading: A Game-based Reading Animation Generation Framework to Engage Kids Reading using AI and Computer Graphics Techniques (for Special Needs)

Jiayi Zhang¹, Jiayu Zhang², Justin Wang³, Yu Sun⁴, ¹Northwood High School, 4515 Portola Pkwy, Irvine, CA 92620, ²Northwood High School, 4515 Portola Pkwy, Irvine, CA 92620, ³The Peddie School, 201 S Main St. Hightstown, NJ 08520, ⁴California State Polytechnic University, Pomona, CA, 91768, Irvine, CA 92620

ABSTRACT

Children with ASD or ADHD are having a hard time learning and understanding, and there’s no perfect education system [1][2]. However, audios and animations can improve their reading effectiveness. This paper designs an application to have animated characters talking with audio based on the text using Optical Character Recognition, text to speech, and Natural Language Processing.

KEYWORDS

AI, Computer Graphics Techniques, Machine Learning.

AI in Telemedicine: An Appraisal on Deep Learning-based Approaches to Virtual Diagnostic Solutions (VDS)

Ozioma Collins Oguine and Kanyifeechukwu Jane Oguine, Department of Computer Science, University of Abuja, Nigeria

ABSTRACT

Telemedicine is an approach to healthcare delivery that has seen a wide range of adoptions and implementations. Its fast-paced development in our contemporary society is credence to the advancements in Artificial Intelligence and Information Technology. This paper carries out a descriptive study to broadly explore AIs implementations in healthcare delivery with a more holistic view of the usability of various Telemedical Innovations in enhancing Virtual Diagnostic Solutions (VDS). This research further explores notable developments in Deep Learning model optimizations for Virtual Diagnostic Solutions. A further research review on the prospects of Virtual Diagnostic Solutions (VDS) and foreseeable challenges was also highlighted. Conclusively, this research gives a general overview of Artificial Intelligence in Telemedicine with a central focus on Deep Learning-based approaches to Virtual Diagnostic Solutions.

KEYWORDS

Biomedical imaging, Telemedicine, Smart Healthcare, Medical Imaging, AI, Virtual Diagnostic Solutions.

An Approach of Information Extraction For Question Answering in Natural Language Processing

Priya K P and Harikrishnan T P, Datahub Technologies and R& D, Kochi, Ernakulam, Kerala, INDIA

ABSTRACT

Question Answering (QA) is a branch of the Natural Language Understanding (NLU) field (which falls under the NLP umbrella). It aims to implement systems that, when given a question in natural language, can extract relevant information from provided data and present it in the form of natural language answer. The problem of making a fully functional question answering system is a problem which has been quite popular among researchers. Information Extraction systems takes text in natural language as input and produces structured information specified by a certain criterion, which is relevant to that particular use case. This paper introduces Information Extraction echnology, its various sub-tasks focussing on question answering, highlights state-of-the-art research in various IE subtasks, current challenges, and future research directions.

KEYWORDS

Natural language processing(NLP), Information Retrieval(IR), Information Extraction(IE), Question Answering(QA).

A Comparison of Paper-Based Tests and Computer-Based Tests in the Emergence of COVID- 19 Era: Evidence by Event-Related Potentials

Reyhaneh Barani Toroghi, Department of Languages, Mashhad Branch, Islamic Azad University, Mashhad Iran

ABSTRACT

Scholars pay particular attention to reading comprehension skill and they believe that it is a criterion for determining whether a person is academically literate or not. Reading comprehension is one of the most important sections of the international tests including IELTS or TOFEL. Also, reading comprehension tests employed at schools or university are attempting to determine the comprehension level of the test takers specifically inCorona virus emergence era. Metacognitive strategies are used by test takers in order to gain their intended marks in tests rather than comprehending the text. Thus, it is a tool for readers to increase their decoding skills. This study aimed at exploring the comprehension level of the test takers without the effects of Metacognitive strategies. While there seems to be a deep linkage between the brain and reading comprehension, Event-related potentials (ERPs) as one of the methods of brain activity measuring allows researchers to observe reading-related brain processes and can document neural patterns at the millisecond level. This study aimed at high-lightening the role of computer-based testing through ERP. 10 Iranian IELTS candidates holding the band scores from 6 to 8 participated in this study. The reading comprehension items of a retired version of IELTS were given to the test takers as the paper-version test and as the stimuli in computer-version test format. In this research, mixed method was used. This study aimed at comparing the paper-version and computer-based tests. EEG signals of the participants were recorded during the computer-based version test and ERPs were extracted. Findings showed that computer-version testing can examine the readers’ comprehension level directly.

KEYWORDS

Event-related potentials (ERPs), reading comprehension, computer-version testing, paper-version testing, Metacognitive strategies.

Code-Mixed Hinglish to English Language Translation Framework

Ishali Jadhav, Aditi Kanade, Vishesh Waghmare, Sahej Singh Chandok and Prof.Ashwini Jarali, Department of Computer Engineering, Hope Foundation’s International Institute of Information Technology, Pune, India

ABSTRACT

Indians, like many other non-English speakers around the world, avoid using single code in their social media conversations. To show their linguistic proficiency, they utilize transliteration and randomly merge English words via mixing two or more languages (English-Hindi, English-Spanish, etc.). As a result, numerous social media applications generate significant amounts of unstructured text as a result of the surge in smart device usage. In the field of text mining, code-mixing (CM) is a rapidly evolving field of study. The current state of communications in social media, blogs, and reviews is a flurry of inventive, code-mixed messages. This is due to its modern but yet localized way of speaking. For different purposes, a language uses linguistic codes from other languages. Code-mixing of Hindi and English, is a common phenomenon in day-to-day language usage in India. This mixing is so common that people have started considering this a different language altogether and it is called by the name Hinglish. The use of multi-linguistic language in the new generation is widespread in the form of code-mixed data on social media and various other platforms, and therefore a translation system is required for helping the monolingual users, as well as for easier comprehension by language processing models. This mixed language poses a new challenge to the problem of machine translation. It is necessary to identify the foreign elements in the language and process them accordingly. In this study, we present a mechanism for machine translation of a bi-lingual language i.e. Hinglish to monolingual Hindi and English forms.

KEYWORDS

Code-Mixed Language, natural language processing, language identification, transliteration, machine translation, Hinglish.

Text Summarization of News in Tamil Epaper using Machine Learning

Dr.E.K.Vellingiriraj¹, Dr.M.Balamururgan², Dr.S.Nalini³, Dr.N.T.Renukdevi⁴, Dr.K.Sarawathi⁴, ¹Professor, Nandha Engineering College, Erode, India, ²Associate Professor, Christ University, Bengaluru, India, ³Assistant Professor, PSG College of Technology, Coimbatore, India, ⁴Assistant Professor, Kongu Engineering College, Erode, India

ABSTRACT

Automatic text summarization is the process of reducing the size of original content to reduced computational burden of handling original large volume of data. There are many research works has been introduced earlier for the automatic text summarization. In the existing work Machine Learning based Automatic Text Summarization (MLATS) is introduced for the automatic summarization outcome. However this research work doesn’t focus on semantic properties and interrelationship between different contents. This is resolved in the proposed work by introducing the method namely Ranking based Hierarchical Clustering Summarization Technique (RHCST). Initially, hierarchical clustering method is introduced to cluster the sentences which provide the similar meaning. And then optimal summarization is performed using Modified GA and Adaptive PSO method. Here accurate text summarization is ensured by adapting the ranking method along with modified GA and Adaptive PSO method. The overall implementation of the research work is done in the matlab simulation environment from which it is proved that the proposed method attains better outcome than the existing methodology.

KEYWORDS

Tamil News ePaper, Text Summarization, Rank based PSO, Modified GA, NLP.

Towards Modi Script Preservation: Tools for Digitization

Kishor Patil, Neha Gupta, Dr. Damodar M., Dr. Ajai Kumar, Centre for Development of Advanced Computing (C-DAC), Pune, India

ABSTRACT

Modi is a heritage script belong to Brahmi which is used mainly for writing Marathi, an Indo-Aryan language spoken in western and central India, mostly in the state of Maharashtra. “Modimanuscript "written from the past, reveals the history of the Maratha Empire from its inception under Chhattrapati Shivaji Maharaj to the creation of movable metal type when Modi was slowly relegated to an inferior position, unfolds perspectives and reflects the social, political and cultural sense of his time." There is a need among the historians, design professionals, researchers and students to understand the script better so that the rich heritage and cultural treasure available in this script is efficiently organized. Other regional languages such as Hindi, Gujarati, Kannada, Konkani and Telugu were also using Modi. This paper presents the first step of long-term development of various tools and technologies, which is collection, analysis, and digitization of Modi script.

KEYWORDS

Language preservation, language development, Modi, Language Analysis, Heritage script.

Task-Oriented Dialogue Systems: Performance vs Quality-Optima, A Review

Ryan Fellows^1*, Hisham Ihshaish¹, Steve Battle¹, Ciaran Haines¹, Peter Mayhew^1,2, J.Ignacio Deza^{1, 3}, ¹Computer Science Research Centre (CSRC), University of the West of England (UWE), Bristol, United Kingdom, ²GE Aviation, Cheltenham, United Kingdom, ³Universidad Atlántida Argentina, Mar del Plata, Argentina

ABSTRACT

Task-oriented dialogue systems (TODS) are continuing to rise in popularity as various industries find ways to effectively harness their capabilities, saving both time and money. However, even state-of-the-art TODS are not yet reaching their full potential. TODS typically have a primary design focus on completing the task at hand, so the metric of task-resolution should take priority. Other conversational quality attributes that may point to the success, or otherwise, of the dialogue, may be ignored. This can cause interactions between human and dialogue system that leave the user dissatisfied or frustrated. This paper explores the literature on evaluative frameworks of dialogue systems and the role of conversational quality attributes within dialogue systems, looking at if, how, and where they are utilised, and examining their correlation with the performance of the dialogue system.

KEYWORDS

Dialogue Systems, Chatbot, Conversational Agents, AI, Natural Language Processing, Quality Attributes.

AMAR_ABSA: Arabic Mobile App Reviews Dataset for Aspect-based Sentiment Analysis Tasks

Daniel Voskergian¹ and Mahmoud Al-saheb², ¹Department of Computer Engineering, Al-Quds University, Jerusalem, Palestine, ²Department of Computer Engineering, Polytechnic University, Hebron, Palestine

ABSTRACT

Aspect-based sentiment analysis is a fine-grained analysis that aims to extract aspects and associated sentiments of an entity from a text. Until the current time of conducting this study, the number of publicly available Arabic datasets in the field of ABSA is very limited and supports the sentiment analysis tasks over very few domains (i.e., books, hotels). However, these datasets do not fulfill the need of researchers who want to experiment and perform their studies over other domains in the field. One domain not well-explored in the field of Arabic ABSA is the mobile app domain. To the best of our knowledge, we did not find any publicly available dataset in the mobile app domain that targets the Arabic language for the ABSA tasks. In this study, we constructed a domain-specific human-annotated dataset (AMAR_ABSA) that consists of musical mobile app reviews written in the Arabic language (MSA & DLA) and made it publicly available to enable the researchers to perform ABSA tasks in this domain. We designed AMAR_ABSA such that it can be used to serve various tasks. The tasks can be, but are not limited to 1) Sentiment Analysis, 2) Aspect Category Classification, 3) Aspect Sentiment Analysis 4) Rate prediction. This paper describes statistics about the dataset and provides a baseline evaluation for the subtask of aspect category extraction using state-of-the-art machine learning algorithms (i.e., SVM, LR, DT, RF, MNB classifiers).

KEYWORDS

Aspect-based sentiment analysis, mobile app domain, text classification, information extraction, natural language processing, machine learning, Arabic reviews.

Dispocen and the Project “Pulso Andaluz”: A Way to Study Lexical Centrality and Social Prototypes with Computational Linguistics

Belén Extremera-Pérez, Department of Greek Philology, Arab Studies, Linguistics and Documentation,Universidad de Málaga, Málaga, Spain

ABSTRACT

This paper presents the application of the program DispoCen to the project Pulso Andaluz. DispoCen— created at the University of Málaga, Málaga, Spain— consists of a library implemented in the language R for statistical analysis. DispoCen calculates the lexical centrality (and availability) index of a word within the results of a representative group in a community. In an attempt to improve the traditional formula to calculate lexical availabity by López and Strassburger, the researchers at the University of Málaga Antonio Ávila-Muñoz and José María Sánchez-Sáez have developed a new formula based on the Fuzzy Sets Theory. With this new formula, the index finds an objective foundation, and presents the prototype that a community has on certain idea or topic. This article explains how to use DispoCen for lexical research purposes, and shows the results obtained in the pilot survey for the project Pulso Andaluz.

KEYWORDS

cognitive linguistics, computational linguistics, lexical availability, lexical centrality.

Emoji-based Fine-grained Attention Network for Sentiment Analysis in the Microblog Comments

Deng Yang, Liu Kejian and Yang Cheng, Department of Computer Engineering, Xihua University, Chengdu, China

ABSTRACT

Microblogs Have Become A Social Platform For People To Express Their Emotions In Real-Time, And It Is A Trend To Analyze User Emotional Tendencies From The Information Of Micro blogs. The Dynamic Features Of Emojis Can Affect The Sentiment Polarity Of Micro blog Texts. Since existing Models Seldom Consider The Diversity Of Emoji Sentiment Polarity, The Paper Propose a Microblog Sentiment Classification Model Based On Albert-Faet. We Obtain Text Embedding Via Albert Pretraining Model And Learn The Inter-Emoji Embedding With an attention-Based Lstm Network. In Addition, A Fine-Grained Attention Mechanism Is Proposed to Capture The Word-Level Interactions Between Plain Text And Emoji. Finally, We concatenate These Features And Feed Them Into A Cnn Classifier To Predict The Sentiment Labels Of The Microblogs. To Verify The Effectiveness Of The Model And The Fine-Grained attention Network, We Conduct Comparison Experiments And Ablation Experiments. The comparison Experiments Show That The Model Outperforms Previous Methods In Three evaluation Indicators (Accuracy, Precision, And Recall) And The Model Can Significantly Improve Sentiment Classification. The Ablation Experiments Show That Compared With albert-Aet, The Proposed Model Albert-Faet Is Better In The Metrics, Indicating That The fine-Grained Attention Network Can Understand The Diversified Information Of Emoticons.

KEYWORDS

Sentiment Analysis, Pre-training Model, Emojis, Attention Mechanism.

Word Sense Disambiguation using Bert

V P Amrutha, S Dev Aravind, Navya, M Prathiba and K Usha, Department of Computer Science & Engineering, NSS College of Engineering Palakkad, Kerala, India

ABSTRACT

Word Sense Disambiguation(WSD) aims to find the exact sense of words which are polysemous in a particular context. Knowledge-based methods use lexical resources while supervised methods assume that the context can provide enough evidence on its own to disambiguate words. In this paper, we propose a hybrid approach that incorporates gloss into neural networks for WSD of a document. BERT has been empirically shown to provide very good contextualised representations of words in sentences. We fine-tune the pre-trained BERT models for WSD and provide them pairs of context and gloss as input.

KEYWORDS

Word Sense Disambiguation, Natural Language Processing, Polysemous, Transformers.

Phrase based English to Manipuri SMT System for Tourism Domain

Maibam Indika Devi¹ and Bipul Syam Purkayastha², ¹Department of Computer Science, IGNTU-RCM, Kangpokpi, Manipur, ²Department of Computer Science, Assam University, Silchar, Assam

ABSTRACT

Statistical Machine Translation (SMT) is one of the ruling approach adopted for developingmajor translation systems today. Very little work in machine translation has been done for the Manipuri language. Here, a machine translation system from English to Manipuri is reported. The variance in the structure and morphology between English and Manipuri languages and the lack of resources for Manipuri languages pose a significant challenge in developing an MT system for the language pair. While English has poor morphology, SVO structure and belong to Indo-European family. Manipuri language has richer morphology, SOV structure and belongs to Sino-Tibetan family. Manipuri has two scripts- Bengali script and Meitei script. Here the Bengali script is used for developing the system. In this paper, the phrase based SMT technique is adopted using Moses toolkit and corpus from tourism domain is used for training the system.. The evaluation gives a score of 9.68 BLEU.

KEYWORDS

SMT, Phrase-based, English- Manipuri, Moses, BLEU.

Predictors of Personality Structure Analysis based on Personality Prediction Models Constructed by Open Data Source

Hua CHENG¹ and Dandan WANG², ¹Sport Science Institute, Lingnan Normal University, Zhanjiang, Guangdong, China, ²Department of Psychology, Guangdong Medical University, Zhanjiang, Guangdong, China

ABSTRACT

Objective: This study takes further step on understanding personality structure in order to cope with the mental health during the COVID-19 global pandemic situation.

Methods: Categorized the independent variables into biological, family and cultural predictors according to the datasets of the Big-5 personality survey online. And established multiple regression prediction models and exhaustive CHAID decision tree model of each personality trait.

Results: Females are different from males in personality. The personality changes when growing. One-handed dominants are less agreeable and open than those who use both hands. Different sexual orientation does have variety personality. Native language used and education attainment is significantly related to personality accordingly. Marriage did help shaping personality to be more extroverted, less neurotic or agreeable and more conscientious and open. People raised in urban are more agreeable and open. Neurotic and open people often come from small families. person participated in voting are more extroverted, conscientious and open but less neurotic and agreeable. Different religions and races have different characteristics in each dimension of personality and there is no clear pattern have been found.

Conclusion: Personality traits are indeed affected by multiple confounding factors. but the exploration on multiple cultures predictors still needed more details.

KEYWORDS

Big Five Personality traits, Multivariate Linear Regression Model, Exhaustive CHAID Decision Tree Model, Gender Differences, Cross-age Differences, Hand Preference, Sexual Orientation, Family and culture factors.

Computing Latin Squares Designs/ Split-Plot Designs Anova

Ishraga Mustafa Awad Allam, Information Technology & Network Administration, University of Khartoum, Khartoum City, Sudan

ABSTRACT

Computing experimental design in statistics: Latin Squares Designs/Split-Plot Designs ANOVA for n readings, stored or entered.

KEYWORDS

Statistics, Experimental in statistics: Latin Squares Designs/Split-Plot Designs ANOVA, ANOVA.

A Novel Approach for Neural Information Retrieval with Click-Attention Graph

Hsiao-Yu Tung, Chiun-Chieh Hsu and Ru-Min Hsiao, National Taiwan University of Science and Technology, Taipei, Taiwan

ABSTRACT

To deal with the lack of labeled training data, many researches of neural-network-based information retrieval use click logs to train models, which are easier to obtain and often seen as implicit relevance feedback left by users. However, most of the neural-network-based information retrieval methods take this kind of data as a substitute of labeled data. Therefore, they often ignore the rich information hidden in the users’ clicks and the problems of noise and sparsity hidden in the click logs. This paper proposes a new Neural IR method IRCGNN (Information Retrieval with Click Graph based on Neural Networks). It uses the graph neural network to aggregate relevant information from their neighbors in the click graph, which can attain richer and more complete representation via users’ clicks. Moreover, it utilizes the attention mechanism to make the aggregation process more refined, where the relevance between nodes can be measured from various perspectives. In addition, we make experiments to compare the performance of IRCGNN with those of other related methods. It reveals that IRCGNN outperforms the others due to the better ranking of documents with click graphs.

KEYWORDS

Information Retrieval、Click Graph、Neural Network、Attention Mechanism.

Outlier Detection and Reconstruction of Lost Land Surface Temperature Data in Remote Sensing

Muhammad Yasir Adnan, Prof Yong Xue and Richard Self, School of Computing & Engineering, University of Derby, United Kingdom

ABSTRACT

Missing values categorized as outliers are quite common in quantitative remote sensing. These are due to technical limitations and effect of weather on ability of instrument to capture data. To handle these missing values, we propose an Outlier-Search-and-Replace (OSR) algorithm by utilizing spatial as well as temporal information for detection and reconstruction of missing data. The algorithm searches for outlier in the data and reconstruct by finding the best possible match in spatial locations.

KEYWORDS

Remote Sensing, Missing Data Reconstruction, Outlier, MODIS, Land Surface Temperature.

A Comparison between Vgg16 and Xception Models Used as Encoders for Image Captioning

Asrar Almogbil, Amjad Alghamdi, Arwa Alsahli, Jawaher Alotaibi, Razan Alajlan and Fadiah Alghamdi, Department of Computer Science, college of Computer Science and Information Technology, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia

ABSTRACT

Image captioning is an intriguing topic of Natural Language Processing (NLP) and Computer Vision (CV). The present state of image captioning models allows it to be utilized for valuable tasks, but it demands a lot of computational power and storage memory space. Despite this problems importance, only a few studies have looked into models comparison in order to prepare them for use on mobile devices. Furthermore, most of these studies focus on the decoder part in an encoder-decoder architecture, usually the encoder takes up the majority of the space. This study provides a brief overview of image captioning advancements over the last five years and illustrate the prevalent techniques in image captioning and summarize the results. This paper also discussed the commonly used models, the VGG16 and Xception, while using the Long short-term memory (LSTM) for the text generation. Further, the study was conducted on the Flickr8k dataset.

KEYWORDS

Image Captioning, Encoder-Decoder Framework, VGG16, Xception, LSTM.

Voice Chatbot for Hospitality

SAGINA ATHIKKAL and JOHN JENQ, Montclair State University, NJ USA

ABSTRACT

Chatbot is a machine with the ability to answer automatically through a conversational interface. A chatbot is considered as one of the most exceptional and promising expressions of human computer interaction. Voice-based chatbots or artificial intelligence (AI) devices transform human-computer bidirectional interactions that allow users to navigate an interactive voice response (IVR) system with their voice generally using natural language. In this paper, we focus on voice based chatbots for mediating interactions between hotels and guests from both the hospitality technology providers’ and guests’ perspectives. We developed a hotel web application with the capability to receive a voice input. The application was developed with Speech recognition and deep synthesis API for voice to text and text to voice conversion, a closed domain question answering (cdQA) NLP solution was used for query the answer.

KEYWORDS

Natural Language Processing, Chatbot, Voice Based Digital Assistants, Closed Domain Question Answering.

Necessity of Integrating Ethical, Legal and Social Aspects into the Common Process Models for it Projects

Sascha Alpers, FZI Forschungszentrum Informatik, Germany

ABSTRACT

There are many different process models for managing software development projects. These also consider the elicitation and management of requirements - with very different agile or classical ap-proaches. The framework provided in particular by ethical aspects, legal constraints and questions of social technology design (ELSA or ELSI) is not explicitly addressed in the process models. There are separate approaches for this, such as the IEEE Standard Model Process for Addressing Ethical Concerns during System Design (IEEE7000-2021). However, the lack of explicit integration of these issues in common process models such as SCRUM or V-Modell-XT does not ensure that the necessary space for reflection on ELSA is opened up in development projects. The article discusses the problem and shows possible solution options for further discourse.

KEYWORDS

procedure model, ethics, law, social technology design.

A Method to Compactly Store Scrambled Data Alongside Standard Unscrambled Disc Images of CD-ROMs

Jacob Hauenstein, Computer Science Department, The University of Alabama in Huntsville, Huntsville, Alabama, USA

ABSTRACT

When archiving and preserving CD-ROM discs, data sectors are often read in a so-called “scrambled mode” before being unscrambled and further processed into a standard disc image. Processing of scrambled data into a standard disc image is potentially lossy, but standard disc images exhibit greater software compatibility and usability compared to scrambled data. Consequently, for preservation purposes, it is often advantageous to store both the scrambled data and the corresponding standard disc image, resulting in high storage demands. Here, a method that enables compact storage of scrambled data alongside the corresponding (unscrambled) standard CD-ROM disc image is introduced. The method produces a compact representation of the scrambled data that is derived from the standard disc image. The method allows for (1) storage of the standard unscrambled disc image in unmodified form, (2) easy reconstruction of the scrambled data, and (3) a substantial space savings compared standard data compression techniques.

KEYWORDS

compact disc, compression, data archival, data preservation, scrambled.

An Overview of Phishing Victimization: Human Factors and The Emotions’ Role

Mousa Jari, School of Computing, Newcastle University, Newcastle, UK, College of Applied Computer Science , King Saud University, Riyadh, Saudi Arabia

ABSTRACT

Phishing is one of the most common types of online scam, where a phisher plays with the psyche and emotions of a victim and manages to cause damage to the victim, varying from retrieving passwords to stealing money. The most common pattern of phishing is to share some malicious link/URL with the victim, which, when clicked, results in information leakage through which a victim can lose his confidential information, passwords or even money [1], [2]. The increasing trend in phishing has made it something worth studying, due to which several pieces of research have been conducted in this subject area. In this research study, we explore and highlight emotional factors that have been identified in previous studies that are found to be significant in phishing victimization. In addition, we compare what security organizations and researchers have highlighted in terms of phishing types, training, and phishing categories. The identification has been done through the literature review, in which credible and published sources have been taken into account.

KEYWORDS

Phishing, emotions, information, victimization & training.

Performance Testing: A Structural Approach for Load Testing of Web-based Applications

Dr. Pankaj Moharil¹ and Dr. Sudarson Jena², 1Sant Gadge Baba Amravati University, Amravati, Maharashtra, India, ²SUIIT, Sambalpur, Odhisha, India

ABSTRACT

In today’s digital era the web applications generating business and promotes goodwill among peoples and prospects. Web application becomes strong marketing media for enterprises; software companies, all types of business organizations from small, large and medium scale by delivering solid messages among peoples and society where customers expecting fast and reliable services from such web applications and network applications. Dynamic change in users requests forces software organizations and companies to concentrate on the efficiency and quality of such web applications with heavy load and stress. In order to satisfy the customer and fulfilling various requirements and complexity of web applications organizations have to focus and concentrate on the performance improvement and factors affecting the performance of the web applications. Load testing helps in solving complex web application efficiency and quality problem. The question arises that how to do the load testing of such web applications in a systematic manner which will cover all the characteristics of the performance. This paper focuses on the performance testing issues and introduces a structure for load testing of web-based applications. This load testing structure aims to provide the tester team with an integrated process from requirement gathering, workload management, scalability analysis, and reliability assessment, to test execution. Based on the stakeholder’s requirements and load given to the application, the structural design helps in performing the load testing process, the proper software objects conserved by the design. The design helps in generating accurate workload models for load management, scalability measurement techniques for the proper scaling of web applications and reliability assessment helps in improving the quality and efficiency of software applications. This helps software testers to manage the heavy load on web applications with stability, sustainability and fast responsiveness of applications. In addition, it strengthens and helps the tester to understand the design and implementation of the application and use of automated tools in load testing. Thus, the efficiency of load testing can be easily simplified.

KEYWORDS

Requirement gathering, Performance, workload, Scalability, Reliability, Testing.

An Real-time Multiplayer FPS Game using 3D Modeling and AI Machine Learning

John Zhang¹ and Yu Sun², ¹Crean Lutheran High school , 12500 Sand Canyon Ave, Irvine, CA 92618, ²California State Polytechnic University, Pomona, CA, 91768, Irvine, CA 92

ABSTRACT

AIs have been a key component in the gaming industry throughout its history. Developers have had multiple ways of creating new AI models that best suit their game in order to enhance the playing experience. However, with the increase in the popularity of online multiplayer games, AIs now have to compete with the experience of playing with other people [1]. In an attempt to enhance AI behaviors to match that of a real player, the paper discusses the one solution for creating models that can be used for further AI research. Through utilizing some of the built in features of Unity as well as Photon Network services, the game Maze Escape combines the multiplayer aspect of FPS games and some simple game AI models to allow them to be compared against each other in order to more easily recreate multiplayer experience using AI bots [2][3]. Thus, this paper hopes to encourage developers to think about how AIs are not only used to enhance single player experiences but it can also be used in multiplayer.

KEYWORDS

Multiplayer FPS Game, 3D Modeling, AI Machine.

WaterPolo Pro: A 3D Water Polo Sports Game for WaterPolo Engagement and Practic

Ethan Pendroy¹, Ryan Yan² and Ang Li³, ¹San Clemente high schoolm, 700 Avenida Pico, San Clemente, CA 92673, ²Cal Poly Pomona, 3801 West Temple Avenue, Pomona, California 91768, ³California State University, Long Beach, 1250 Bellflower Blvd, Long Beach, USA

ABSTRACT

“What is an ef ective method of educating the public about water polo in an enjoyable way?” is the question that this research aims to answer. The approach chosen in this research was to create a water polo video game,which would involve the players with an interactive experience that would provide a more thorough understanding of the sport [7]. To test the ef ectiveness of this game in educating individuals, two samples of ten individuals were asked to play the game twice and take a post-game survey that asked whether this game helped educate about water polo. One sample was made up of individuals from the general public that have played water polo before, while everyone from the other sample had no previous experience in water polo. The scores from the games that each individual played were recorded, and the survey results were accumulated and divided based on which sample the individual was from. The results indicated that both samples generally improved their scores after a second play-through of the game and felt that they had learned more about water polo [5]. More individuals from the sample who had never played water polo before reported learning from this game than individuals from the sample who had played water polo before.

KEYWORDS

Water Polo, Simulation, Video Game, A* Pathfinding.

Comparison of Forecasting Methods of House Electricity Consumption for Honda Smart Home

Farshad Ahmadi Asl and Mehmet Bodur, Eastern Mediterranean University, Famagusta, via Mersin 10, Turkey

ABSTRACT

Electricity consumption of buildings composes a major part of the energy consumption of a city. Forecasting the electricity opens the possibility for home energy management systems, leading to design more sustainable houses in the future and reducing the total energy usage. Energy performance in buildings is influenced by many factors like outside temperature, humidity, and a variety of electric devices, therefore, multivariate prediction methods are preferred rather than univariate. Honda smart home US data set was selected to compare three methods for minimum forecasting MAE and RMSE: Artificial Neural Networks (ANN), Support Vector Regression (SVR), and Fuzzy Rule-Based Systems (FRBS) for Regression by constructing many models out of each method, on a multivariate data set in different time-terms. The comparison shows that SVR is a superior method over the rest.

KEYWORDS

Forecasting, Mathematical Models, Electricity, Prediction, Consumption, ANN, SVR, FRBS.

A Social-Driven Intelligent System to Assist the Classification of Pet Emotions using Deep Learning and Big Data Analysis

Hans Li¹ and Yu Sun², ¹Damien High School, 2280 Damien Ave, La Verne, CA 91750, ²California State Polytechnic University, Pomona, CA, 91768, Irvine, CA 92620

ABSTRACT

Pets have always been a big part of families, and people always imagine what their pet is thinking by their actions and face [1]. However, No one can tell what a pet might be thinking unless they are very familiar with them [2]. This paper develops/designs/proposes an application/software/tool to... [There is an AI that could try to understand the pet’s emotion, and people can share their photos to other pet lovers on the app, which will further make the AI more accurate [3]. We applied our application to people who have pets of all kind and conducted a qualitative evaluation of the approach. The results show that [the AI is decently accurate and the app is fairly easy to use from feed backs made by testers. And the AI have the potential to get more accurate in the future with the more data customers posts, and thus will give more accurate results back to the users].

KEYWORDS

Social-Driven, Machine Learning, Classification.

Cascaded Machine Learning-Based Emotion Detection Approach using EEG Signals

Nayana Vaity and Pankaj Kawadkar, Madhyanchal Professional University, India

ABSTRACT

Emotion is the basic information about the mental status of a human. The human brain stores the activity of human action and reaction as electroencephalogram (EEG) signals. The EEG-based emotion detection is a new area of research in affective computing for medical science and the human-computer interface. For EEG-based emotion identification, this research proposed a cascaded machine learning algorithm. A multi-stage support vector machine and a random forest classifier are combined in the cascaded machine learning algorithm. The multi-stage support vector expands the classification process learning space. The feature extraction strategy is also used in the detection process. The feature components of EEG data are easily extracted using stationary wavelet transforms. The retrieved features form a matrix, which is then transposed into feature vectors. The suggested approach uses MATLAB tools to simulate. Apply SEED datasets to the validation process and measure standard characteristics like accuracy, specificity, and sensitivity. The proposed algorithm compares CNN and ensemble classifier results. The proposed strategy improves performance by around 4% and 2%, respectively.

KEYWORDS

Emotion Detection, EEG signal, SWT, SVM, RF, Machine Learning.

Mass Surveillance, Behavioural Control, And Psychological Coercion the Moral Ethical Risks in Commercial Devices

Yang Pachankis, Universal Life Church, Modesto, California, USA

ABSTRACT

The research observed, in parallel and comparatively, a surveillance state’s use of communication & cyber networks with satellite applications for power political & realpolitik purposes, in contrast to the outer space security & legit scientific purpose driven cybernetics. The research adopted a psychoanalytic & psychosocial method of observation in the organizational behaviors of the surveillance state, and a theoretical physics, astrochemical, & cosmological feedback method in the contrast group of cybernetics. Military sociology and multilateral movements were adopted in the diagnostic studies & research on cybersecurity, and cross-channeling in communications were detected during the research. The paper addresses several key points of technicalities in security & privacy breach, from personal devices to ontological networks and satellite applications - notably telecommunication service providers & carriers with differentiated spectrum. The paper discusses key moral ethical risks posed in the mal-adaptations in commercial devices that can corrupt democracy in subtle ways but in a mass scale.

KEYWORDS

Cybersecurity, Risk Prevention, Psychosocial Cybernetics, Cyber Surveillance, Time & Entropy.