Browsing by Author "Jayasekara, N.E.C."

Now showing 1 - 3 of 3
  • No Thumbnail Available
    Item
    High Tech Vision to Detect Currency Denomination and Virtual Wallet to Retrieve the Monetary Position for Visually Debilitated People
    (Uva Wellassa University of Sri Lanka, 2020) Perera, N.B.L.K.; Franciscus, A.U.; Wickramasinghe, M.G.R.D.; Jayasekara, N.E.C.; Pathirana, K.P.P.S.; Jayakody, J.A.V.M.K.
    The transformation of currency notes and coins denomination recognition to an automated system as a solution for visually debilitated individuals to overcome the difficulties facing when handling monetary transactions. This research presents a model to detect currency notes and coins to visually debilitated individuals and to retrieve the current monetary position of them as per their obligation and provide audio output in the Sinhala language. The general procedure of the system includes digital image processing, convolutional neural network, voice identification algorithm, and monetary position calculation algorithm. Sri Lanka currency notes and coins images were captured in a wide variety of environments, in association with lighting conditions and background to make the data set, using the image preprocessing technique. The YOLOv2, R-CNN network model which is a high speed, real-time object detection algorithm to verify objects as currency notes and coins. Then by using Keras Xception model, predict images, feature extraction and fine-tuning have been done to train the data set. The Computer vision used to improve machine perception to retrieve real-time detection. The detected currency notes or coins denomination is provided as an audio output, then retrieves the obligation of the user, which is whether to debit, credit or to retrieve the current monetary position. The monetary position provides audio output in the virtual wallet as a substitute for a realworld wallet since impairments have a scarcity in memorizing their actual balance. The study revealed a system to detect and retrieve the currency denomination and monetary position of blind individuals with the overall accuracy rate of 100% in algorithm experiments. Keywords: Visually debilitated individuals, Currency recognition, Virtual wallet, monetary position
  • No Thumbnail Available
    Item
    Smart Reply Generation for SMS Using Natural Language Processing
    (Uva Wellassa University of Sri Lanka, 2020) Kaumada, M.W.S.; Sumanasekara, S.S.; Jayasekara, N.E.C.; Wimaladharma, S.T.C.I.
    The use of Short Message Service (SMS) is increasing due to the rapid increase in mobile phone usage and the simplicity in sending SMS messages. With the increasing complexity of human lives, people are seeking more efficient activities to save time. This research proposes an end-to-end method that automatically generates short responses known as Smart Replies by identifying the content of an SMS using natural language processing. There are a few pieces of research done on the topic of Smart Reply. Most of them are carried out for the emails. And the efficiency and the size of those existing models cannot be used in an offline mobile device. The application will use Natural Language Processing to process an incoming message and then uses a neural network to predict the most likely responses which will allow us to send it directly or edit it before sending it to the recipient. The Ubuntu Corpus dataset was used for training and testing the model by analysing its properties. It is identified that there are three main approaches: TF-IDF, Recurrent Neural networks (RNN), and Long Short-Term Memory (LSTM) that can be used in the model. After a performance test, identified the most suitable approach is LSTM. Accordingly built a Sequential Neural Network with a Dense with sigmoid activation using LSTM. Finally, extract the highest three responses from the trained model to show in the SMS application. This proposed model achieved around 92% percent of accurate results and it can be used offline and also it is a lightweight file that can be easily handled in a mobile device. Keywords: Smart reply, SMS, Natural language processing, Long short-term memory, Sequential neural network
  • No Thumbnail Available
    Item
    Smart SMS Classification for Android Operating System Using Natural Language Processing
    (Uva Wellassa University of Sri Lanka, 2020) Sumanasekara, S.S.; Kaumada, M.W.S.; Jayasekara, N.E.C.; Wimaladharma, S.T.C.I.
    The use of Short Message Service (SMS) is increasing as more people exchange SMS messages very frequently due to the rapid increase of mobile phone usage and the simplicity in sending SMS messages. However, this has led to an increase in mobile device attacks using SMS Spam. The two main categories of SMS Messages are spam messages and ham (legitimate) messages. Up to now, several kinds of research were done on SMS classification but all of them are on spam filtering techniques by using various algorithms and machine learning techniques. In this paper, we present a novel approach that can detect and filter both spam and ham messages into a better organization under six different predefined categories named as Primary for legitimate messages, Bank and Finance, Social and Web, Promotions, Service Provider Messages, and Spam Messages by using Natural Language Processing for Android Operating System. A smart messaging application that can properly organize SMS into categories will help to identify the SMS easily as they are classified under different tabs. Even though SMS can be identified and categorized manually with little or no effort by people, it remains difficult for mobile phones. A dataset is created according to the Sri Lankan context and various experiments are performed to evaluate the performance of the SMS Classification. Initially, the features were selected based on the behavior of messages and extracted the features from the dataset to get the feature vectors. Naive Bayes and Support Vector Machines algorithms were used to select the best classification algorithm. With the highest accuracy rate, the Support Vector Machines algorithm is selected to train the model while k-Fold cross-validation is used to perform the validation. Our proposed approach achieved a 93% accuracy rate and the model is deployed in the Android environment and its performance is confirmed using a proof of concept. Keywords: SMS classification, Natural language processing, Support vector machines, Naive bayes algorithm, Android