Applied Article
H.3.8. Natural Language Processing
Davud Mohammadpur; Mehdi Nazari
Abstract
Text summarization has become one of the favorite subjects of researchers due to the rapid growth of contents. In title generation, a key aspect of text summarization, creating a concise and meaningful title is essential as it reflects the article's content, objectives, methodologies, and findings. Thus, ...
Read More
Text summarization has become one of the favorite subjects of researchers due to the rapid growth of contents. In title generation, a key aspect of text summarization, creating a concise and meaningful title is essential as it reflects the article's content, objectives, methodologies, and findings. Thus, generating an effective title requires a thorough understanding of the article. Various methods have been proposed in text summarization to automatically generate titles, utilizing machine learning and deep learning techniques to improve results. This study aims to develop a title generation system for scientific articles using transformer-based methods to create suitable titles from article abstracts. Pre-trained transformer-based models like BERT, T5, and PEGASUS are optimized for constructing complete sentences, but their ability to generate scientific titles is limited. We have attempted to improve this limitation by presenting a proposed method that combines different models along with a suitable dataset for training. To create our desired dataset, we collected abstracts and titles of articles published on the ScienceDirect.com website. After performing preprocessing on this data, we developed a suitable dataset consisting of 50,000 articles. The results from the evaluations of the proposed method indicate more than 20% improvement based on various ROUGE metrics in the generation of scientific titles. Additionally, an examination of the results by experts in each scientific field revealed that the generated titles are also acceptable to these specialists.
Technical Paper
H.6.5.13. Signal processing
Khadijeh Aghajani
Abstract
Voice Activity Detection (VAD) plays a vital role in various audio processing applications, such as speech recognition, speech enhancement, telecommunications, satellite phone, and noise reduction. The performance of these systems can be enhanced by utilizing an accurate VAD method. In this paper, multiresolution ...
Read More
Voice Activity Detection (VAD) plays a vital role in various audio processing applications, such as speech recognition, speech enhancement, telecommunications, satellite phone, and noise reduction. The performance of these systems can be enhanced by utilizing an accurate VAD method. In this paper, multiresolution Mel- Frequency Cepstral Coefficients (MRMFCCs), their first and secondorder derivatives (delta and delta2), are extracted from speech signal and fed into a deep model. The proposed model begins with convolutional layers, which are effective in capturing local features and patterns in the data. The captured features are fed into two consecutive multi-head self-attention layers. With the help of these two layers, the model can selectively focus on the most relevant features across the entire input sequence, thus reducing the influence of irrelevant noise. The combination of convolutional layers and self-attention enables the model to capture both local and global context within the speech signal. The model concludes with a dense layer for classification. To evaluate the proposed model, 15 different noise types from the NoiseX-92 corpus have been used to validate the proposed method in noisy condition. The experimental results show that the proposed framework achieves superior performance compared to traditional VAD techniques, even in noisy environments.
Original/Review Paper
H.3.8. Natural Language Processing
Nura Esfandiari; Kourosh Kiani; Razieh Rastgoo
Abstract
A chatbot is a computer program system designed to simulate human-like conversations and interact with users. It is a form of conversational agent that utilizes Natural Language Processing (NLP) and sequential models to understand user input, interpret their intent, and generate appropriate answer. This ...
Read More
A chatbot is a computer program system designed to simulate human-like conversations and interact with users. It is a form of conversational agent that utilizes Natural Language Processing (NLP) and sequential models to understand user input, interpret their intent, and generate appropriate answer. This approach aims to generate word sequences in the form of coherent phrases. A notable challenge associated with previous models lies in their sequential training process, which can result in less accurate outcomes. To address this limitation, a novel generative chatbot is proposed, integrating the power of Reinforcement Learning (RL) and transformer models. The proposed chatbot aims to overcome the challenges associated with sequential training by combining these two approaches. The proposed approach employs a Double Deep Q-Network (DDQN) architecture with utilizing a transformer model as the agent. This agent takes the human question as an input state and generates the bot answer as an action. To the best of our knowledge, this is the first time that a generative chatbot is proposed using a DDQN architecture with the embedded transformer as an agent. Results on two public datasets, Daily Dialog and Chit-Chat, validate the superiority of the proposed approach over state-of-the-art models involves employing various evaluation metrics.
Original/Review Paper
H.3. Artificial Intelligence
Pouria Rabiei; Nosratali Ashrafi-Payaman
Abstract
Today, the amount of data with graph structure has increased dramatically. Detecting structural anomalies in the graph, such as nodes and edges whose behavior deviates from the expected behavior of the network, is important in real-world applications. Thus, in our research work, we extract the structural ...
Read More
Today, the amount of data with graph structure has increased dramatically. Detecting structural anomalies in the graph, such as nodes and edges whose behavior deviates from the expected behavior of the network, is important in real-world applications. Thus, in our research work, we extract the structural characteristics of the dynamic graph by using graph convolutional neural networks, then by using temporal neural network Like GRU, we extract the short-term temporalcharacteristics of the dynamic graph and by using the attention mechanism integrated with GRU, long-term temporal dependencies are considered. Finally, by using the neural network classifier, the abnormal edge is detected in each timestamp. Conducted experiments on the two datasets, UC Irvine messages and Digg with three baselines, including Goutlier, Netwalk and CMSketch illustrate our model outperform existing methods in a dynamic graph by 10 and 15% onaverage on the UCI and Digg datasets respectively. We also measured the model with AUC and confusion matrix for 1, 5, and 10 percent anomaly injection.
Other
J.10.5. Industrial
Arezoo Zamany; Abbas Khamseh; Sayedjavad Iranbanfard
Abstract
The international transfer of high technologies plays a pivotal role in the transformation of industries and the transition to Industry 5.0 - a paradigm emphasizing human-centric, sustainable, and resilient industrial development. However, this process faces numerous challenges and complexities, necessitating ...
Read More
The international transfer of high technologies plays a pivotal role in the transformation of industries and the transition to Industry 5.0 - a paradigm emphasizing human-centric, sustainable, and resilient industrial development. However, this process faces numerous challenges and complexities, necessitating a profound understanding of its key variables and concepts. The present research aimed to identify and analyze these variables in the realm of high technology transfer in Industry 5.0. Following a systematic literature review protocol, 84 relevant articles published between 2017 and 2024 were selected based on predefined criteria including relevance to the research topic, publication quality, and citation impact. These articles were analyzed using a comprehensive text mining approach incorporating keyword extraction, sentiment analysis, topic modeling, and concept clustering techniques implemented through Python libraries including NLTK, SpaCy, TextBlob, and Scikit-learn. The results categorize the key variables and concepts into five main clusters: high technologies (including AI, IoT, and robotics), technology transfer mechanisms, Industry 5.0 characteristics, implementation challenges (such as cybersecurity risks and high adoption costs) and opportunities (including increased productivity and innovation potential), and regulatory frameworks. These findings unveil various aspects of the technology transfer process, providing insights for stakeholders while highlighting the critical role of human-technology collaboration in Industry 5.0. The study's limitations include potential bias from focusing primarily on English-language literature and the inherent constraints of computational text analysis in capturing context-dependent nuances. This research contributes to a deeper understanding of technology transfer dynamics in Industry 5.0, offering practical implications for policymaking and implementation strategies.
Original/Review Paper
F.4.18. Time series analysis
Fatemeh Moodi; Amir Jahangard Rafsanjani; Sajjad Zarifzadeh; Mohammad Ali Zare Chahooki
Abstract
This article proposes a novel hybrid network integrating three distinct architectures -CNN, GRU, and LSTM- to predict stock price movements. Here with Combining Feature Extraction and Sequence Learning and Complementary Strengths can Improved Predictive Performance. CNNs can effectively identify short-term ...
Read More
This article proposes a novel hybrid network integrating three distinct architectures -CNN, GRU, and LSTM- to predict stock price movements. Here with Combining Feature Extraction and Sequence Learning and Complementary Strengths can Improved Predictive Performance. CNNs can effectively identify short-term dependencies and relevant features in time series, such as trends or spikes in stock prices. GRUs designed to handle sequential data. They are particularly useful for capturing dependencies over time while being computationally less expensive than LSTMs. In the hybrid model, GRUs help maintain relevant historical information in the sequence without suffering from vanishing gradient problems, making them more efficient for long sequences. LSTMs excel at learning long-term dependencies in sequential data, thanks to their memory cell structure. By retaining information over longer periods, LSTMs in the hybrid model ensure that important trends over time are not lost, providing a deeper understanding of the time series data. The novelty of the 1D-CNN-GRU-LSTM hybrid model lies in its ability to simultaneously capture short-term patterns and long-term dependencies in time series data, offering a more nuanced and accurate prediction of stock prices. The data set comprises technical indicators, sentiment analysis, and various aspects derived from pertinent tweets. Stock price movement is categorized into three categories: Rise, Fall, and Stable. Evaluation of this model on five years of transaction data demonstrates its capability to forecast stock price movements with an accuracy of 0.93717. The improvement of proposed hybrid model for stock movement prediction over existing models is 12% for accuracy and F1-score metrics.
Original/Review Paper
H.3. Artificial Intelligence
Fariba Taghinezhad; Mohammad Ghasemzadeh
Abstract
Artificial neural networks are among the most significant models in machine learning that use numeric inputs. This study presents a new single-layer perceptron model based on categorical inputs. In the proposed model, every quality value in the training dataset receives a trainable weight. Input data ...
Read More
Artificial neural networks are among the most significant models in machine learning that use numeric inputs. This study presents a new single-layer perceptron model based on categorical inputs. In the proposed model, every quality value in the training dataset receives a trainable weight. Input data is classified by determining the weight vector that corresponds to the categorical values in it. To evaluate the performance of the proposed algorithm, we have used 10 datasets. We have compared the performance of the proposed method to that of other machine learning models, including neural networks, support vector machines, naïve Bayes classifiers, and random forests. According to the results, the proposed model resulted in a 36% reduction in memory usage when compared to baseline models across all datasets. Moreover, it demonstrated a training speed enhancement of 54.5% for datasets that contained more than 1000 samples. The accuracy of the proposed model is also comparable to other machine learning models.
Original/Review Paper
H.3. Artificial Intelligence
Seyed Alireza Bashiri Mosavi; Omid Khalaf Beigi; Arash Mahjoubifard
Abstract
Using intelligent approaches in diagnosing the COVID-19 disease based on machine learning algorithms (MLAs), as a joint work, has attracted the attention of pattern recognition and medicine experts. Before applying MLAs to the data extracted from infectious diseases, techniques such as RAT and RT-qPCR ...
Read More
Using intelligent approaches in diagnosing the COVID-19 disease based on machine learning algorithms (MLAs), as a joint work, has attracted the attention of pattern recognition and medicine experts. Before applying MLAs to the data extracted from infectious diseases, techniques such as RAT and RT-qPCR were used by data mining engineers to diagnose the contagious disease, whose weaknesses include the lack of test kits, the placement of the specialist and the patient pointed at a place and low accuracy. This study introduces a three-stage learning framework including a feature extractor by visual geometry group 16 (VGG16) model to solve the problems caused by the lack of samples, a three-channel convolution layer, and a classifier based on a three-layer neural network. The results showed that the Covid VGG16 (CoVGG16) has an accuracy of 96.37% and 100%, precision of 96.52% and 100%, and recall of 96.30% and 100% for COVID-19 prediction on the test sets of the two datasets (one type of CT-scan-based images and one type of X-ray-oriented ones gathered from Kaggle repositories).
Original/Review Paper
I.4. Life and Medical Sciences
Nasrin Aghaee-Maybodi; Amin Nezarat; Sima Emadi; Mohammad Reza Ghaffari
Abstract
Sequence alignment and genome mapping pose significant challenges, primarily focusing on speed and storage space requirements for mapped sequences. With the ever-increasing volume of DNA sequence data, it becomes imperative to develop efficient alignment methods that not only reduce storage demands but ...
Read More
Sequence alignment and genome mapping pose significant challenges, primarily focusing on speed and storage space requirements for mapped sequences. With the ever-increasing volume of DNA sequence data, it becomes imperative to develop efficient alignment methods that not only reduce storage demands but also offer rapid alignment. This study introduces the Parallel Sequence Alignment with a Hash-Based Model (PSALR) algorithm, specifically designed to enhance alignment speed and optimize storage space while maintaining utmost accuracy. In contrast to other algorithms like BLAST, PSALR efficiently indexes data using a hash table, resulting in reduced computational load and processing time. This algorithm utilizes data compression and packetization with conventional bandwidth sizes, distributing data among different nodes to reduce memory and transfer time. Upon receiving compressed data, nodes can seamlessly perform searching and mapping, eliminating the need for unpacking and decoding at the destination. As an additional innovation, PSALR not only divides sequences among processors but also breaks down large sequences into sub-sequences, forwarding them to nodes. This approach eliminates any restrictions on query length sent to nodes, and evaluation results are returned directly to the user without central node involvement. Another notable feature of PSALR is its utilization of overlapping sub-sequences within both query and reference sequences. This ensures that the search and mapping process includes all possible sub-sequences of the target sequence, rather than being limited to a subset. Performance tests indicate that the PSALR algorithm outperforms its counterparts, positioning it as a promising solution for efficient sequence alignment and genome mapping.
Original/Review Paper
Document and Text Processing
Zobeir Raisi; Vali Mohammad Nazarzehi
Abstract
The Persian language presents unique challenges for scene text recognition due to its distinctive script. Despite advancements in AI, recognition in non-Latin scripts like Persian still faces difficulties. In this paper, we extend the vanilla transformer architecture to recognize arbitrary shapes of ...
Read More
The Persian language presents unique challenges for scene text recognition due to its distinctive script. Despite advancements in AI, recognition in non-Latin scripts like Persian still faces difficulties. In this paper, we extend the vanilla transformer architecture to recognize arbitrary shapes of Persian text instances. We apply Contextual Position Encoding (CPE) to the baseline transformer architecture to improve the recognition of Persian scripts in wild images, especially for oriented and spaced characters. The CPE utilizes position information to generate contrastive data pairs that help better in capturing Persian characters written in a different direction. Moreover, we evaluate several state-of-the-art deep-learning models using our prepared challenging Persian scene text recognition dataset and develop a transformer-based architecture to enhance recognition accuracy. Our proposed scene text recognition architecture achieves superior word recognition accuracy compared to existing methods on a real-world Persian text dataset.
