An Approach for a Next-Word Prediction for Ukrainian Language

Text generation, in particular, next-word prediction, is convenient for users because it helps to type without errors and faster. Therefore, a personalized text prediction system is a vital analysis topic for all languages, primarily for Ukrainian, because of limited support for the Ukrainian language tools. LSTM and Markov chains and their hybrid were chosen for next-word prediction. Their sequential nature (current output depends on previous) helps to successfully cope with the next-word prediction task. The Markov chains presented the fastest and adequate results. The hybrid model presents adequate results but it works slowly. Using the model, user can generate not only one word but also a few or a sentence or several sentences, unlike T9.

Reply Using Past Replies—A Deep Learning-Based E-Mail Client

Email is the most common and effective source of communication for most enterprises and individuals. In the corporate sector the volume of email received daily is significant while timely reply of each email is important. This generates a huge amount of work for the organisation, in particular for the staff located in the help-desk role. In this paper we present a novel Smart E-mail Management System (SEMS) for handling the issue of E-mail overload. The Term Frequency-Inverse Document Frequency (TF-IDF) model was used for designing a Smart Email Client in previous research. Since TF-IDF does not consider semantics between words, the replies suggested by the model are not very accurate. In this paper we apply Document to Vector (Doc2Vec) and introduce a novel Gated Recurrent Unit Sentence to Vector (GRU-Sent2Vec), which is a hybrid model by combining GRU and Sent2Vec. Both models are more intelligent as compared to TF-IDF. We compare our results from both models with TF-IDF. The Doc2Vec model performs the best on predicting a response for a similar new incoming Email. In our case, since the dataset is too small to require a deep learning algorithm model, the GRU-Sent2Vec hybrid model cannot produce ideal results, whereas in our understanding it is a robust method for long-text prediction.

Improving Student’s Listening Skills through (TPS) Text Prediction Strategy by QR-Code Scanning Activity

This study is aimed to improve the student’s listening skills through (TPS) text prediction strategy by QR-code scanning activity. The method of analyzing data in this research is to use the classroom action research study conducted using 2 circles. The subjects of the study were in SMPN 3 Selat, to the ninth students of the first semester in academic year 2019/2020 which consist of 32 students with 15 male and 17 female. This study dealt with TPS (Text Prediction strategy by QR-Code scanning activity to improve students’ listening skills. The steps were planning, action, observation, reflection. After the series of action research in teaching and learning process in cycle I and cycle II by using the (TPS) text prediction strategy by QR-code scanning activity to increase student’s listening skills of narrative text, the researcher could take some result as follows: there are significant differences we can see from the result of the study before the action and after the action, even in cycle I and cycle II activity. In cycles, the student’s minimum completeness rises from 11 students to 25 students from before cycles. After cycle II, the student’s minimum completeness is raising also to 29 Students from 25 students who passed the minimum completeness in cycle I. The number of students who still failed in reaching the minimum completeness in cycle II is only 3 students. this may be observed deeply again in the next research to find out the factors and causes about the students who still having trouble mastering listening skills

Neural network dealing with Arabic language

Artificial Neural Networks have proved their efficiency in a large number of research domains. In this paper, we have applied Artificial Neural Networks on Arabic text to prove correct language modeling, text generation, and missing text prediction. In one hand, we have adapted Recurrent Neural Networks architectures to model Arabic language in order to generate correct Arabic sequences. In the other hand, Convolutional Neural Networks have been parameterized, basing on some specific features of Arabic, to predict missing text in Arabic documents. We have demonstrated the power of our adapted models in generating and predicting correct Arabic text comparing to the standard model. The model had been trained and tested on known free Arabic datasets. Results have been promising with sufficient accuracy.

Can learning from natural image denoising be used for seismic data interpolation?

We have developed an interpolation method based on the denoising convolutional neural network (CNN) for seismic data. It provides a simple and efficient way to break through the problem of the scarcity of geophysical training labels that are often required by deep learning methods. This new method consists of two steps: (1) training a set of CNN denoisers to learn denoising from natural image noisy-clean pairs and (2) integrating the trained CNN denoisers into the project onto convex set (POCS) framework to perform seismic data interpolation. We call it the CNN-POCS method. This method alleviates the demands of seismic data that require shared similar features in the applications of end-to-end deep learning for seismic data interpolation. Additionally, the adopted method is flexible and applicable for different types of missing traces because the missing or down-sampling locations are not involved in the training step; thus, it is of a plug-and-play nature. These indicate the high generalizability of the proposed method and a reduction in the necessity of problem-specific training. The primary results of synthetic and field data show promising interpolation performances of the adopted CNN-POCS method in terms of the signal-to-noise ratio, dealiasing, and weak-feature reconstruction, in comparison with the traditional [Formula: see text]-[Formula: see text] prediction filtering, curvelet transform, and block-matching 3D filtering methods.

Hand Gesture Recognition for Emoji and Text Prediction

Emoticons' are ideograms and smileys utilized in electronic messages and website pages. Emoticons exist in different classifications, including outward appearances, regular items, places and kinds of climate, and creatures. They are much similar to emojis, however emoticons are real pictures rather than typo graphics. This undertaking perceives the emoticons utilizing hand motions. We are detecting hand gestures and preparing a Convolutional Neural Network (CNN) model on a training dataset. We will make a database of hand gestures and train them. The system utilized here is a CNN. We are utilizing the SIFT filter to identify the hand and CNN for preparing the model. SIFT filter give a lot of highlights of an image that are not influenced by numerous factors, for example, object scaling and rotation. The SIFT filtering procedure comprises of two areas. The first is a procedure to identify intrigue focuses in the hand. Intrigue focuses are the points in the image in a 2D space that surpasses some limit measure and is better than straight forward edge recognition. The second segment is a procedure to make a vector like descriptor and this is the most special and prevalent part of the SIFT filter.