Search

In recent years, many scholars have chosen to use word lexicons to incorporate word information into a model based on character input to improve the performance of Chinese relation extraction (RE). For example, Li et al. proposed the MG-Lattice model in 2019 and achieved state-of-the-art (SOTA) results. However, MG-Lattice still has the problem of information loss due to its model structure, which affects the performance of Chinese RE. This paper proposes an adaptive method to include word information at the embedding layer using a word lexicon to merge all words that match each character into a character input-based model to solve the information loss problem of MG-Lattice. The method can be combined with other general neural system networks and has transferability. Experimental studies on two benchmark Chinese RE datasets show that our method achieves an inference speed up to 12.9 times faster than the SOTA model, along with a better performance. The experimental results also show that this method combined with the BERT pretrained model can effectively supplement the information obtained from the pretrained model, further improving the performance of Chinese RE.

Deep learning text detection is generally divided into two steps: prediction candidate box of depth model and post-processing, and post-processing usually uses NMS or prediction box to merge and connect. At present, many methods of text detection can detect the character position of the document image, but they are often not accurate enough, which affects the effect of subsequent recognition. To solve the problem of inaccurate detection, this article proposes a post-processing method of document detection based on geometric features from the perspective of post-processing. The post-processing method is mainly divided into four modules. Firstly, the background removal (BR) module separates the background information and character information through pixel threshold. Secondly, the candidate box expansion (CBE) module expands the prediction box in all directions by judging whether the boundary of the prediction box is in the character pixel. Then is the non-standard box removal (NBR) module, using the consistency principle of characters and surrounding characters to filter out the error detection of some prediction boxes. Finally, the module of repeating box removal (RBR) is used to remove the repeated prediction box. In order to verify the effectiveness of this method, a large number of experiments have been conducted on Standard yi, Chinese2k, English2k, ICDAR 2015 and ICDAR 2017(CTW-12k) datasets. The experimental results show that the method proposed in this article can improve the effect of text detection.

In this article, we present a new multi-lingual Optical Character Recognition (OCR) system for scanned documents. In the case of Latin characters, current open source systems such as Tesseract provide very high accuracy. However, the accuracy of the multi-lingual documents, including Asian characters, is usually lower than that for Latin-only documents. For example, when the document is the mix of English, Chinese and/or Korean characters, the OCR accuracy is lowered than English-only because the character/text properties of Chinese and Korean are quite different from Latin-type characters. To tackle these problems, we propose a new framework using three neural blocks (a segmenter, a switcher, and multiple recognizers) and the reinforcement learning of the segmenter: The segmenter partitions a given word image into multiple character images, the switcher assigns a recognizer for each sub-image, and the recognizers perform the recognition of assigned sub-images. The training of recognizers and switcher can be considered traditional image classification tasks and we train them with a supervised learning method. However, the supervised learning of the segmenter has two critical drawbacks: Its objective function is sub-optimal and its training requires a large amount of annotation efforts. Thus, by adopting the REINFORCE algorithm, we train the segmenter so as to optimize the overall performance, i.e., we minimize the edit distance of final recognition results. Experimental results have shown that the proposed method significantly improves the performance for multi-lingual scripts and large character set languages without using character boundary labels.

Segmenting characters in an image is a classic yet challenging task in computer vision. Correctly determining boundaries of adhesive characters with various scales and shapes is essential for character segmentation, especially for separating handwritten characters. Nevertheless, there is seldom work in the literature which can achieve satisfactory performance. In this article, by leveraging the ability of deep neural networks, we proposed a two-stage character segmentation network with two-stream attention and edge refinement (TSER) to tackle this problem. TSER firstly locates every character by object detection, then extracts their corresponding contours. In the process, a novel two-stream attention mechanism (TSAM) is proposed to make the network focus more on the discrepancy of character boundaries. Furthermore, a novel generating method is used to dynamically generate anchors on different feature levels to improve model’s sensitivity on the shapes and scales of characters. Eventually a cascaded edge refinement network is used to obtain contour of each character. To prove the efficiency and generalization ability of our model, we compared TSER with traditional algorithms and other deep learning models on two commonly used datasets in different segmentation tasks. The comparative result indicated that TSER reached state-of-the-art performance.

This study uses deep learning theory into the character recognition technology for Shui characters in ancient books, with the objectives of overcoming the instability of the high-pixel ancient Shui characters generative model and the need for large scale handwritten text data annotation among other issues. By constructing a multilayer adversarial neural network with a Laplacian structure, a clear generative model is established for original image data of Shui characters and a stable adversarial network model with multiple mapping relationships from coarse to fine is formed. Based on the analysis of the feature distance of Shui character image samples, the minimum inter-class spacing value and the optimal number of clusters are calculated. Combined with feedback from the classifier model, the optimal number of clusters in the clustering model is adjusted, an evaluation function with information entropy adjustment and clustering threshold convergence is constructed for the unsupervised labelling of Shui character image samples. In this paper, the feedback from the convolutional neural network is used to determine the algorithmic model of the hyperparameters for clustering annotation, and this structure is also designed to improve the recognition rate of handwritten Shui characters in ancient books.

We developed an impulse radio ultra-wideband (IR-UWB) radar-based system that can recognize alphanumeric characters in midair without the need for any handheld device. The hardware consists of four IR-UWB radar sensors set up with a rectangular geometry. Writing a single character in midair results in artifacts that make some characters look similar on a position trajectory-based (x, y) plane, which makes them difficult to classify. Thus, we developed an algorithm that transforms 2D coordinate image data into trigonometric ratios (i.e., tangents) and plots them against the time axis to obtain unique images for training a convolutional neural network. An extended Kalman filter is used to obtain the 2D trajectories of hand motions. To evaluate our proposed method, we first applied it to characters that may be written in midair very simply without creating artifacts and compared its performance with that of a state-of-the-art digit classification algorithm. Then, we considered combining characters written midair with and without artifacts. After the individual character recognition, we combined the characters into words. We defined a specific marker based on an energy threshold to detect the start and end of a character for midair writing. The energy level was found to change drastically when the hand is pulled in and out of the radar plane. The proposed method was found to outperform the current state of the art at character classification when artifacts are present in the images.

Aiming at the problem of insufficient sentiment word extraction ability in existing text sentiment analysis methods and OOV (out-of-vocabulary) problem of pre-training word vectors, a neural network model combining multi-head self-attention and character-level embedding is proposed. An encoder-decoder (Seq2Seq) model is used for sentiment analysis of text. By adding character-level embedding to the word embedding layer, the OOV problem of pre-trained word vectors is solved. At the same time, the self-attention mechanism is used to perform attention calculations within the input sequence to find intra-sequence connections. The model uses bidirectional LSTM (Long Short-Term Memory) independently encodes the context semantic information of the input sequence to obtain deeper emotional features, to more effectively identify the emotional polarity of short text. The effectiveness of our method was verified on the public Twitter dataset and the movie review dataset published by ACL. Experimental results show that the accuracy of the model on the three categories of Twitter, movie reviews and IMDB datasets reaches 66.45%, 79.48% and 90.34%, respectively, which verifies that the model has excellent performance in datasets in different fields.

This paper investigates the stability analysis of sampled-data control systems with nonuniform sampling and communication delays. First, by taking full advantage of effective information of the sampling intervals [tk, tk+1) and [tk-τ, tk+1-τ), we propose a new looped Lyapunov functional called two-sided sampling-interval-dependent looped Lyapunov functional, which is tailored for the use of generalized free-matrix-based inequality and canonical Bessel-Legendre inequality. Second, by using the proposed looped Lyapunov functional and the integral inequality technique, some novel N-dependent stability criteria with less conservatism are derived for the considered systems. The obtained stability criteria are shown to have a hierarchical character: the higher level of hierarchy, the less conservatism of the derived result. Finally, two examples are provided to illustrate the effectiveness of the proposed approaches.

The recommender systems predict and calculate user preferences for recommendations. However, such predictions and calculations are neither accurate nor viable in the context of smart TV due to the reasons that it is a lean-back, non-personalized device, and normally enjoyed in groups. Hence, group recommendations have utmost importance, specifically from the perspectives of watching smart TV. The existing group recommendation techniques predict the individual's preferences and then create a virtual group profile for recommendations. However, identifying and satisfying every group member is challenging. Numerous techniques have been proposed, such as face detection and recognition systems, but these systems lead to security and privacy issues. This paper proposes a smart TV-based recommender system that aims to identify the individuals and group members from their “age,” “gender,” and “number” information to overcome the biases that occur due to predictions and estimations of user's preferences. The study proposes a novel formula and age-gender matrix for generating anonymous, consolidated, and secure profiles, including group profiles on a smart TV. This study further proposes a novel method for finding a dominant character in a group by utilizing the user's ratings. Results show that the group decision has a significant impact on supplying social metadata, such as ratings, comments, etc., which in turn improve recommendation results. For materializing the proposed work, smart TV's processing, storage, and camera are utilized. The prototypical implementation has been tested and analyzed with improved recommendation results and viewer(s) satisfaction.

The accuracy of current natural scene text recognition algorithms is limited by the poor performance of character recognition methods for these images. The complex backgrounds, variations in the writing, text size, orientations, low resolution and multi-language text make recognition of text in natural images a complex and challenging task. Conventional machine learning and deep learning-based methods have been developed that have achieved satisfactory results, but character recognition for cursive text such as Arabic and Urdu scripts in natural images is still an open research problem. The characters in the cursive text are connected and are difficult to segment for recognition. Variations in the shape of a character due to its different positions within a word make the recognition task more challenging than non-cursive text. Optical character recognition (OCR) techniques proposed for Arabic and Urdu scanned documents perform very poorly when applied to character recognition in natural images. In this paper, we propose a multi-scale feature aggregation (MSFA) and a multi-level feature fusion (MLFF) network architecture to recognize isolated Urdu characters in natural images. The network first aggregates multi-scale features of the convolutional layers by up-sampling and addition operations and then combines them with the high-level features. Finally, the outputs of the MSFA and MLFF networks are fused together to create more robust and powerful features. A comprehensive dataset of segmented Urdu characters is developed for the evaluation of the proposed network models. Synthetic text on the patches of images with real natural scene backgrounds is generated to increase the samples of infrequently used characters. The proposed model is evaluated on the Chars74K and ICDAR03 datasets. To validate the proposed model on the new Urdu character image dataset, we compare its performance with the histogram of oriented gradients (HoG) method. The experimental results show that the aggregation of multi-scale and multilevel features and their fusion is more effective, and outperforms other methods on the Urdu character image and Chars74K datasets.

Recently, the Lattice structure has been proved to be effective for Chinese entity relation extraction by incorporating the word information. However, this method cannot utilize phrases that contain sentence structure information. In order to solve the problem, we introduced a Lattice model with phrase component representation, which can merge word, phrase and character input. We also apply the multi-head self-attention mechanism in the classification layer to effectively learn the dependencies between distant words. After comparative experiments, our model achieves better results.

Residual convolutional networks proved superiority in the image recognition field in addition to many other pattern recognitions related problems. A special benefit of residual convolutional networks has been proven in the field of handwriting recognition, especially in handprinted characters and online signature recognition. Two published research papers addressed the two problems with highly accurate results that compete with the state-of-the-art techniques. Both problems have two well-known challenges in object classification: intra-class variability and inter-class similarity. In the character recognition problem, different users produce different shapes for the same character in addition to the user himself, who produces different shapes for the same character resulting in high intra-class variability. Inter-class similarity exists due to the similarity between different characters. In handwritten signature recognition, intra- class variability is introduced by the same person where no two signatures of the same person may coincide. On the other hand, persons may have similar signatures resulting in inter-class similarity. Solving the two pattern recognition problems using deep learning techniques requires the selection of the best technique that can model the handwriting process, but at the same time, it requires a large training data. In the character recognition case, large training samples are available from many datasets resources. A recent data set representing this challenge is the EMINST dataset. An optimized residual architecture has been introduced to give an excellent solution for this problem in one of the comparison papers. In the case of signature recognition, a different solution based on a residual network has been introduced in a second paper because practically training samples that can be collected from new users are very rare. In this paper, the two techniques are compared, and conclusions that may generalize well in other problem domains are stated.

On the one hand, virtual characters may be similar to people in terms of appearance and behavior, but on the other hand, they cannot be considered fully "human". The main goal of this project is to answer the question about the boundaries of the analogy between interactions with real people and interactions with virtual characters. To investigate this, I plan three lines of research covering performance-based as well as economic and moral decision-making tasks. The tasks will differ in terms of the explicitness of the correct answer (from an easy to define the appropriate solution in performance tasks to the inability to choose the only correct option in moral dilemmas) and complexity of the social situation. The main outcome of the experiments will be the answer to the question of how "humanly" are virtual characters treated. I expect that if virtual human are treated like real people, I will observe similar behavioral patterns as when people have real social interactions. Additionally, I plan to systematically manipulate the external characteristics of a virtual character to find the determinants of its influence on participant’s behavior. Specifically, I will attempt to achieve the level of realism appropriate to induce a sense of copresence and consequent influence of virtual characters on people’s behavior. The entire project will therefore answer a broader question about the possibility, scope and nature of the influence of virtual characters on human behavior.

Nowadays, gender classification using voice has been implemented, for a case in the security sector. The voice of each individual has a unique character, because of the diversity in the sound spectrum, frequency, and amplitude between individual. So, the option of technical features to determine specific characteristics to identifying gender came the major issue in classifying. In this study proposed two parameters to determine gender classification based speaker’s voice. There was a fundamental frequency of the YIN algorithm and the cepstrum coefficient of the Mel Frequency Cepstral Coefficient (MFCC) with the classification of Euclidean Distance and Mahalanobis Distance. Based on the data analyzed, it has classified the data test of females on the YIN algorithm with 100% accuracy. Meanwhile, in the male data test, there were several steps to process data that was still misclassified. The first step with the YIN algorithm got an accuracy of 47.5%, then with a combination of YIN and MFCC-Euclidean Distance, the accuracy had raised to 98%. Then continued the Mahalanobis Distance classification to the combination of YIN and MFCC-Euclidean Distance, the accuracy had come 100%. So it can be concluded the parameters applied to determining gender classification based speaker’s voice affect the process and results of the classification.

Distance-based motion adaptation leads to the formulation of a dynamical Distance Geometry Problem (dynDGP) where the involved distances simultaneously represent the morphology of the animated character, as well as a possible motion. The explicit use of inter-joint distances allows us to easily verify the presence of joint contacts, which one generally wishes to preserve when adapting a given motion to characters having a different morphology. In this work, we focus our attention on suitable representations of human-like animated characters, and study the advantages (and disadvantages) in using some of them. In the initial works on distance-based motion adaptation, a 3n-dimensional vector was employed for representing the positions of the n joints of the character at a given frame. Here, we investigate the use of another, very popular in computer graphics, representation that basically replaces every joint position in the three-dimensional space with a set of three sorted Euler angles. We show that the latter can in fact be useful for avoiding some of the artifacts that were observed in previous computational experiments, but we argue that this Euler-angle representation, from a motion adaptation point of view, does not seem to be the optimal one. By paying particular attention to the degrees of freedom of the studied representations, it turns out that a novel character representation, inspired by representations used in structural biology for molecules, may allow us to reduce the character degrees of freedom to their minimal value. As a result, statistical analysis on human motion databases, where the motions are given with this new representation, can potentially provide important insights on human motions. This study is an initial step towards the identification of a full set of constraints capable of ensuring that unnatural postures for humans cannot be created while tackling motion adaptation problems.

Along with the development of technology, especially in the field of games. Many programmers are competing to create games that can be liked by game fans in all walks of life. The educational game is a type of game that brings learning games while playing the idea about its purpose. This research makes educational games with Javanese letter learning materials. In this game, players will answer several questions using Latin letters. Then the player will answer using Javanese characters. If the player can complete the game well, then the player will get a score from the answer. In this study, the fuzzy logic method is applied to character games, this study aims to help manage game scores given to players who successfully solve questions. The use of the fuzzy logic method is taken to determine the game score. In research that has been carried out using fuzzy rules along with time variables and correct answer variables, this study shows 93 % of results were successfully applied to 20 users. This figure appears after the random distribution of the Questionnaire

This study produced a pop-up book based on QR code solid figure material with character strengthening curiosity in 5 elementary schools. The ADDIE model is used in this research and development. All grade 5th students of SD Negeri 2 Mojorembun, Nganjuk Regency were subjects of this study. The material experts, media experts and conduct validity tests by achieving very valid category results. At the same time, the practical test obtains a very high practical value and reaches the convenient category. The product's development results received a good response and enthusiasm from students in the learning activities. The results of the response questionnaire showed that the product developed practically used for learning activities and strengthened the character of curiosity based on the observation attitudes of students in the teaching and learning activities with excellent criteria. In addition, the use of product development can improve students' learning outcomes and provide an understanding of the material of solid figures. It is known from the value they get when working on exercises and getting an average score above the minimum completion criteria to learn the material of solid figures well and maximally.

In this paper, a deep metric learning method with combined loss of the triplet network and autoencoder is presented. Autoencoder is regarded as the regulation network to enable the embedding vector to have some latent features of the input image, and improve its performance. Compared with the pure triplet network, although it increases some complexity during training due to the addition of the decoder, but during testing, their complexities are exactly the same, because the decoder can be completely removed after training. The experiments of the proposed method, triplet network, and one-hot encoded network are performed on various character datasets to show that the proposed method not only achieve better classification performance, but also inherit the benefits of deep metric learning.

In this work, a generic framework is developed that simultaneously embeds 2D shapes’ registration, comparison and grouping, via the assumption that shapes of the same class are distorted level-sets of the same implicit function. The corresponding system is developed so as to deal with the automatic classification of documents according to their writers. The data that the system processes are the realizations of the individual characters, which are mutually aligned per document and per character, modulo affine transformations, and then reduced to a single representative shape. Stationarity conditions of these representatives are then used to statistically test the hypothesis that different documents bear a single representative. The considered documents are grouped according to their writer, by determining the maximal groups that also maximize the joint probability of the classification, computed over the characters, which occur in all documents. We have tested the system on 26 pages of Byzantine manuscripts that preserve Iliad. The computed classification of these pages in 4 writers has been independently verified by expert scientists.

Recognizing printed characters is a little easier than when compared to handwritten characters by offline OCR machines. When the images of the characters are taken from the natural scene, they become hazy and in various angles. Then they are harder to recognize, even though they are printed. In this study, an attempt has been made to identify characters that are colored and in different orientations, using a deep convolutional neural network known as VGG16. The effect on classification accuracy with respect to epochs on the datasets is projected by conducting some experiments. Two Latin character datasets have been considered, such as Chars74kImg and Binary alphanumeric. The Chars74kImg dataset consists of RGB isolated character images taken from the real world, while the Binary alphanumeric dataset comprises binary versions of distinct characters in different orientations. With the Chars74kImg dataset, the suggested model obtains the highest accuracy of 78.04%, far better than the previous techniques found in the literature. Finally, along with quantitative analysis, a comparison of the suggested method to a few preexisting systems is presented.

In this paper, we present a neuromorphic character recognition application using a memristor crossbar architecture. The tested memristor crossbar architecture was composed of a single crossbar array with bipolar inputs for performing the Exclusive-NOR function to measure the similarity between the input character and the pre-stored characters. Ten characters, from "A" to "J", were used for testing the crossbar circuit. For being recognized, ten characters were converted to vectors and applied to the single memristor array with the size of 64×10, which pre-stored all character images, to find the similarity score between the input character and the characters stored in the array. The winner-take-all circuit finally selected the maximum current of the column which stored the pattern that was the best match with the input character. The memristor variation tolerance of the single crossbar array was verified in this work. The simulation results indicate that the single crossbar array with bipolar input performs precisely the Exclusive-NOR function between the input pattern and the patterns stored in the array as well as tolerates well the variation in memristance.

A Ji-shaped Chinese-character patch antenna is proposed for future communications. The proposed antenna is comprised of two dielectric layers. Ji-shaped patch is etched on the upper surface of the top layer, whereas a feeding network is designed in the other layer. Furthermore, the conformal Ji-shaped antenna is designed to be placed on the surface of the cylindrical objects. Simulated results demonstrate that a gain of 7.9±0.3 dBi over the whole operating frequency band (24.7527.5 GHz) can be obtained.

Social robot co-design requires aiding users as they imagine these novel devices within their everyday lives and enabling designers to understand and address users’ experiences. This paper presents the exploratory development and evaluation of a role-playing game aimed at identifying the desired features and uses of a social robot that can assist people diagnosed with depression. Participants (n = 16) played the game as a character with depression, designed a companion robot for that character, and chose reactions to daily challenges. Though participants initially selected robot capabilities based on their own needs, after the game they identified alternative designs that would better address daily challenges faced by individuals with depression. We discuss aspects of the game that allowed participants to understand how various robot characteristics can address the experience of depression and suggest how role-playing games can support users and designers in identifying beneficial features and uses of emerging robotic technologies.

Optical Character Recognition has been long studied over the past few years. The challenge remains for the specific purpose of extracting information from image documents. The aim of this study is to create an end-to-end pipeline for an Indonesian identity card. The final pipeline uses deep learning approach consist of Faster R-CNN for text detection, YOLOv5 for character detection, and Support Vector Machine for Character Recognition. The proposed pipeline showed a remarkable result for both the identity number and the full name. This provides a powerful tool for the auto-fill form and verification process effectively and efficiently.

Classical Chinese poetry is a kind of formatted text with phonological patterns. It has been a big challenge for the classical Chinese poetry generation. Because deep learning technology has made great progress in image and text processing, we present a classical Chinese poetry generation model based on Transformer-XL that can capture longer-term dependency than the original transformer. Our model creates the first character according to the title specified by a user and then concatenates the generated character as the input to predict the next character until reaches the predefined number of characters. Experimental results show that our model can generate poems that have good format, consistency and rhythm.

This paper leverages Norbert Weiner's cybernetic paradigm to understand and contextualize the nature and character of the emergent battlespace in the Age of Artificial Intelligence and Autonomous Weapon Systems. Given the emergence of the Internet of Things, which is co-constituting a technogeographical space, this paper contends that it is more productive to think of the emergent battlespace in cybernetic than in digital terms, which not only allows for the re-evaluation of the human, weapons, tactical mix, but which also necessitates a reassessment of the ethical concerns of warfare in the 21st Century.

Hate speech has become a critical problem in all social media, leading to many hate crimes alongside affecting the mental and emotional well-being of affected individuals. This calls out for methods to detect online hate speech more than ever. While numerous architectures exist for hate speech detection in unimodal setup (i.e., either textual or visual) we have targeted the problem in the context of both text and images inspired by the real-world raw data which involves several modalities. We propose a multimodal hate speech classifier, called as Character Text Image Classifier (CTIC), which builds upon Bidirectional Encoder Representations from Transformers (BERT), Capsule Network, and EfficientNet involving four modalities, namely word embeddings, character embeddings, sentence embeddings, and images. We report the experiments performed upon our proposed model and several other models which have been tested in the process. We train our model with different sampling techniques, and selective training, upon a Twitter dataset, called MMHS150K, consisting of both texts and associated images. Our proposed multimodal approach attains better performance than the previous models constructed upon the dataset.

In this paper, we consider the embedding problem of Chinese new word with respect to its conceptual definition or description, which is especially important for understanding specialty documents. We present a two-stage model to learn the Chinese new word embedding, where the first encodes the information of character components and context, and the second aggregates the semantics of multiple texts. We perform extensive experiments to verify the proposed method and the results outperform the state of art methods on both direct semantics verification and advanced NLP tasks. Comparing with previous methods that require a corpus or an elaborately designed dataset for learning a new word embedding, our method requires only a few pieces of text and supports the evolution of meanings. We also experimentally verify the effects of different parts of model, the number and types of conceptual texts. Finally, we present some biology texts to illustrate whether the specialty semantics are encoded in the word embedding.

Recently, the character-based model that incorporates potential word information has proven effective for Chinese named entity recognition (NER). However, due to the independence of the pre-trained character model and the lexicon, it will cause the embedding space to be misaligned and cannot be combined well. Chinese pre-trained encoders usually process text as characters. It ignores the information carried by the larger granular information, so the encoder cannot easily adapt to certain character combinations. Because large-grained information is ignored and Chinese does not have clear character boundaries, this will lead to the loss of important semantic information, which is an important problem for Chinese. In this paper, we propose PTWA: pre-training with word attention for Chinese named entity recognition. PTWA uses multi-head word attention to form a word vector from multiple word vectors, and proposes a word length prediction task to better integrate the word vector into pre-training. With the powerful capabilities of the transformer, PTWA can explicitly make full use of potential word information without adding an external lexicon, and can coexist with pre-trained models that implicitly use word information (such as BERT-WWM, and ERNIE). Experiments conducted on four Chinese NER datasets show that the performance of PTWA is better than other word-word models and Chinese pre-training models.

Research on the Effect of Using Character-Based Flipbook Teaching Media on Calculus Learning Outcomes. This research is based on problems in the world of education, especially the use of multimedia learning with the calculus of integral material in rotating objects. Research conducted at FKIP UHAMKA with the research method used is experimental design posttest only form and the research sample was 36 students divided into 2 class groups, the experimental class and the control class. Based on the results of the analysis calculation on the analysis prerequisite test that is using the post test sample the learning outcomes of calculus are normally distributed, namely 5.172

This study aims to determine the design stages of developing blended learning smartphone media learning products and to determine the feasibility of integral learning media using the developed flipbook. This study uses a small-scale trial on FHIP UHAMKA while for large-scale trials conducted by Malaysian UTP with 29 respondents including Physics Education FKIP UHAMKA, Mathematics Education FKIP UHAMKA and also Malaysian UTP. The method used is Research and Development (R&D) using the Brog and Gall development procedure. The results of the media interface test obtained a percentage of 77.18%. This shows that the products developed included in the category are very feasible to use while for the effectiveness of students as teaching material in the learning process, the average of the effectiveness of students is 83.26%, the value obtained can be said to be very well seen from the Likert scale index used, then from the assessment of the effectiveness of media students it is feasible to be used in integral learning. Blended learning media with character given to one class in FKIP UHAMKA, obtained an average post-test score of 48.04. The data obtained, the character development of the character questionnaire was 75.04.

Society 5.0 is associated not only with technological advances, but doing system updates in their respective fields. Respondents obtained were 37 in Universiti Teknologi Petronas (UTP). This study aims to analyze the effectiveness of character value in blended learning. The research model used is a quantitative method with a frequency histogram based on the Likert scale from (1 never to 5 always). This paper describes how to see the connection between one character value indicator and another. This result imply that the character analysis value in blended learning methods in teaching has received positive feedback from students, the average mean (responsibility = 81.08; independently = 85.63; discipline = 74.86; curiosity = 74.32; cooperation = 85.09).

This paper deals with Role Playing Games which known as a RPG in abbreviation, and it also review old and current RPG genre video games that found in the market from a well-known classic to its modern contemporary. RPG is a game genre that can be defined as a game in which its system focuses on Role-play aspect found in-game such as role-play as a hero or any specified class in-game, this also include player-built character experience or Meta build. The role-playing game can be seen as story-focused, there is always been a stake that the player has to conquer and it could be something about saving a world by defeating a boss, the player will be put in a shoe of the main character and it is up to the player to build the character to its fullest potential. On the other hand, RPG is also well known for its intricate storyline especially that of Japanese role-playing game also known as JRPG and over the year the story of the RPG genre has evolved becoming more serious and tackling what could be assumed as a realistic and fantasy depiction of a real-life world.

The existing translation models have the problem that the task of translation thread is not clearly divided, which leads to the time-consuming of model translation. This paper designs a long character English aided translation model of foreign related contract text based on feature extraction algorithm. The feature vector of the characters to be recognized is extracted to recognize the long character structure of the foreign related contract text. Based on the feature extraction algorithm, the translation thread task is divided, the context of execution module and translation module is switched, the ambiguity of word segmentation is judged according to the syntactic and semantic information, the auxiliary word segmentation mode is set by gradient descent method, and the automatic alignment function of English auxiliary translation module is designed, to realize the construction of long character English auxiliary translation model of foreign related contract text. Experimental results: the average translation time of the designed model is 3.555s, which is 1.640s and 1.707s shorter than the two existing translation models, which proves that the performance of the model is better.

The human mind is an amazing piece of creation that can handle multiple modalities of input seamlessly and help to make sense about the surroundings. When it comes to making sense about speech, 2 main features of input are sound and vision (although there are many other components). Since not every mind is alike, some of them have trouble processing the sound aspect of input therefore vision becomes their primary source to process and understand speech. Lip reading is a skill that is used mainly by people suffering from hearing deformities and it involves large amount of language specific knowledge as well as contextual awareness i.e. using all possible visual clues that help to make sense of what the other person is saying and thus allow them to take part in the conversation. Recent breakthroughs in the field of Deep learning have clearly shown promise with models that have the ability to extract complex, intricate and generalizable patterns both in spatial as well as temporal dimension. In this paper we present FYEO (For Your Eyes Only) an end-to-end deep learning based solution that only uses vision as its single modality of input and generates a single word, character by character. The model is a modified version of the LipNet architecture from Deep Mind, to a subset of words curated from the Oxford-BBC Lip Reading in the Wild (LRW) dataset. Also, as a part of novel work FYEO is extended by adding attention mechanism for further improvement of the model’s contextual awareness and observe the model’s focus while making a prediction. The standard FYEO model achieves a length normalised test CER (character-error-rate) of 25.024%.

Traditional casting character recognition algorithms need to select appropriate position features for different scenes in the character location step, so it is difficult to realize the recognition task of casting embossed concave and convex characters in the different distribution in complex scenes. In this letter, a recognition method of casting embossed characters based on YOLO v5 is proposed. The fast and reliable depth learning algorithm YOLO v5 is used to automatically extract the image features and realize the recognition of casting embossed characters (including numbers and letters) Recognition. The experimental results show that the accuracy of the network model for steel seal character recognition is higher than traditional computer vision algorithms, the average processing time of the algorithm is quickly, and the weight file volume is small, which meets the accuracy and efficiency requirements of engineering application.

We address the character recognition challenge of similar-looking character classes. Human Vision System often misinterprets visually similar characters while they are present at singular instances. Often, we pay soft attention to a combination of characters to read and interpret a word. A graphical readout of a display device dashboard shows characters generated from a custom vectored font library. A poorly defined character library or dimension adjustment of fonts in the pre-defined layout can often distort the geometric shape of characters. It requires additional intellectual cues for a human to understand and causes poor accuracy by an automated system responsible for character recognition. Even after generalizing over a large-scale dataset, a well-trained character recognition engine misclassifies these custom character images and exhibits low model confidence. In this paper, we optimize a multi-scale Siamese network using multitask Learning to learn significant discriminative features of visually similar characters from a few labeled samples of a custom vectored English font dataset. Using classification and similarity learning, Multitask Learning improves recognition performance and introduces strong inductive biases. Experiments show that our method can effectively distinguish visually similar characters and improves overall classification accuracy.

Estimating the contribution of input towards the prediction via model explainability methods can be insufficient to bring explainability to neural network models if the input consists of character n-grams. As characters can be within multiple words, the section where the prediction has obtained the most information from can not be localised within the text. In this study, novel methods to estimate the importance of words on the prediction of neural network models with character ngram inputs have been proposed. Proposed methods have been tested on text obtained from an e-commerce platform and their performances have been quantitatively compared.

Recent years have witnessed increased interests in scene text recognition (STR). Current state-of-the-art (SOTA) approaches adopt sequence-to-sequence (Seq2Seq) structure to leverage the mutual interaction between images and textual information. However, these methods still struggle to recognize texts in arbitrary shapes. The leading cause is that it brings about information loss and negative noises when directly compressing two-dimension image features into one-dimension vectors. This paper proposes a novel framework named hierarchical refined attention network (HRAN) for STR. HRAN obtains refined representations with the hierarchical attention, which localizes the precise region of current character from two-dimension perspective. Two novel co-attention mechanisms, stacked and guided co-attention, explicitly leverage dependency between spatial-aware contextual features and region-aware visual features without extra character annotations. Experiments show that both on regular and irregular texts, HRAN achieves highly competitive performance compared to SOTA models.

Current text spotting methods perform text detection and text recognition separately. However, in complex scenes where bounding boxes of texts with various shapes are often overlapped, text detection becomes error-prone. By contrast, character detection is more non-ambiguous and easier to learn. In this paper, we present a highly efficient one-stage method named PointerNet for arbitrary-shaped text spotting. Unlike previous methods, PointerNet does not rely on text detection and opens a novel spotting-by-character-detection paradigm. In particular, to connect characters to texts, we propose a simple yet highly effective strategy named pointer that learns the 2D offset from the center of the current character to the center of the subsequent character. Evaluations demonstrate that our PointerNet achieves state-of-the-art performance and is more efficient than current methods (75ms vs. 133ms compared with FOTS). Our code will be publicly available.

Recently, incorporating word information into a character-based model has been proved to be effective for Chinese NER task. However, most existing work ignore the radical information. A novel CRW-NER model is proposed to utilize multiple embeddings in this paper. Besides, the GRU-GatedConv in CRW-NER model is explored to utilize effectively long-distance contextual character information. Experimental results on three public datasets demonstrate the validity of CRW-NER model. The CRW-NER model achieves more excellent performance than the state-of-the-art comparison models.

To reinforce the adaptability of word segmentation model and make up for the inaccessibility of large-scale annotated corpus in military domain, we propose a self-adaptive word segmentation model in military domain based on conditional random field. The model adequately extracts the features of character, position and part-of-speech from texts to increase the accuracy of the general word segmentation model. Meanwhile, it introduces the boundary entropy and dictionary features, and realizes the self-adaptability of the word segmentation model in military domain, based on reverse maximum matching of word segmentation results using the military dictionary. The experiment shows when the training corpus is general and the testing corpus is military, the precision, recall and F1-measure value increased by 9.45%, 9.97% and 9.72%, respectively, compared with the CRF model in general domain, which effectively improves the self-adaptability of segmentation method in military domain.

We present the isometry between the 2-dimensional Funk model and the Finsler-Poincar\'e disk. Then, we introduce the Finslerian Poincar\'e upper half plane model, which turns out to be also isometrically equivalent to the previous models. As application, we state the gapless character of the first eigenvalue for the aforementioned three spaces., Comment: IEEE 15th International Symposium on Applied Computational Intelligence and Informatics, Timisoara, Romania

Using a standard suffix-tree algorithm, we study the repetition characteristic $v(t)$ , which has been introduced by F. Golcher, for different models of random symbolic sequences and compare it with the corresponding data obtained for natural-language texts and program codes. The character of $v(t)$ function, the saturated repetition parameter $v_{0}$ averaged at large enough times $t$ and the appropriate standard deviation $\Delta v_{0}$ are examined for 144 natural, random and randomized texts of different types. The main peculiarities of repetitions peculiar for the Simon, Markov and Miller's monkey text-generating models are analyzed. The results obtained for these analytically tractable models can be useful for developing mathematical fundamentals of the repetition characteristic.

Optical Character Recognition is widely used in scenarios from license plate recognition, chip detection to text translation. However, in some special scenarios, there are very few data for training or the target datasets are not marked. This paper mainly aims to establish a multi-scene text recognition model that can achieve accurate results even when the target test dataset is not marked. The model reduces the distance between domains at the feature level by using an adversarial loss and flipping the gradient. For the adversarial loss, each letter has a different domain discriminator to align the features at the character level. We also align the source domain features and target domain features in the background. Experimental results show that our model achieves good performance when the training dataset and test dataset have domain shift, and suggest possible scenarios for potential applications.

In the manufacturing process of piston, most of the piston cavities are printed with different character sequences to describe the specifications of piston. It is labor-consuming and inefficient to read the piston cavity character sequences manually. Although scholars have done a lot of researches in the field of industrial character recognition, there are few researches on piston cavity character recognition. A piston cavity character recognition method based on Faster R-CNN and priori knowledge library of character sequences is presented. First, we design a ring light source and an imaging device for the piston cavity based on the characters of the piston cavity protruding upward and the texture of the metal being easily reflective. Second, we use the character images in piston cavity collected by the imaging device to make character dataset. Third, according to the noisy background of the piston cavity image, Gaussian filtering and morphological operations were used to obtain a clean background image of the piston cavity. Fourth, use the Faster R-CNN training dataset to get the character recognition model, and then use the character type and position information detected by the recognition model to form a character sequence. Fifth, according to the highly similar characteristics of the character sequences of the same piston model, a character sequences prior knowledge library is constructed to correct the recognition results of Faster R-CNN. The experimental results show that the accuracy rate of the character sequences detected by the character recognition model is 95.5%. And when combined with the prior library of character sequences, the accuracy rate of the character sequences is 99%.