Integrating linguistic features has been widely utilized in statistical machine translation (SMT) systems, resulting improved quality. However, for low-resource languages such as Thai and Myanmar, the integration of neural (NMT) systems yet to be implemented. In this study, we propose transformer-based NMT models (transformer, multi-source transformer, shared-multi-source transformer models) using two-way Thai-to-Myanmar, Myanmar-to-English, Thai-to-English. Linguistic part-of-speech (POS)...

This study investigates the use of deep learning for classifying movie genres based on audio spectrograms. We construct a dataset trailers, transform them into spectrograms, and label by genre. Then, we utilize MATLAB's pre-trained convolutional neural networks (CNNs) clas- sication, comparing performance 9 different architectures, including MobileNet-v2, RestNet-18, DenseNet-201, Places365-GoogLeNet, VGG- 16, VGG-19, Inception-RestNet-v2, Inception-v3, NASANet-Mobile. evaluated all models...

Most of the current sentiment analysis techniques classifies emotions into two classes which are positive and negative. Some works classify them as positive, negative objective (neutral). In fact, there many kinds in human mind. Recently, psychological viewpoints have influenced most analysis. This psychology perspective was adopted to a wider range, more accurate manner. paper reviews computational representation so-called Hourglass Emotion. also proposes construction Thai resource based on...

Distributional semantics in the form of word embeddings are an essential ingredient to many modern natural language processing systems. The quantification semantic similarity between words can be used evaluate ability a system perform interpretation. To this end, number datasets have been created for English over last decades. For Thai few such resources available. In work, we create three by translating and re-rating popular WordSim-353, SimLex-999 SemEval-2017-Task-2 datasets. contain 1852...

This study investigates the use of an ARIMA model, coupled with Monte Carlo simulation, to forecast opening value a Volatility Index (VIX) time series. The data obtained from Chicago Board Options Exchange (CBOE) for years 1992–2019 have been transformed into stationary using detrend method and first-order difference. Augmented Dickey-Fuller (ADF) test is used ensure are adequately transformed. autocorrelation function (ACF) partial ACF (PACF) then identify series serial correlation...

As the manual creation of domain models and also linked data is very costly, extraction knowledge from structured unstructured has been one central research areas in Semantic Web field last two decades. Here, we look specifically at formalized natural language text, which most abundant source human available. There are many tools on hand for information English language, written Thai situation different. The goal this work to assess state-of-the-art formal then give suggestions practical...

Thai herbs have increasingly gained public attention. Recently, there are a number of herb websites. Each website has similar information but quite different details. For example, some webpages do not provide indicating which part can treat the specified symptom. In order to collect more complete information, we developed extraction process extract from multiple The employed HTML parser and file templates recognize useful in various webpage formats. Preliminary experiments gave satisfactory...

Research into semantic similarity has a long history in lexical semantics, and it applications many natural language processing (NLP) tasks like word sense disambiguation or machine translation. The task of calculating is usually presented the form datasets which contain pairs human-assigned score. Algorithms are then evaluated by their ability to approximate gold standard scores. Many such datasets, with different characteristics, have been created for English language. Recently, four those...

A software maintenance size can be used to predict effort or time. However, the traditional metric only relies on line of source code (LOC), which hardly is suitable for object-oriented software. This research proposed four new metrics based number classes, methods, average methods per class, and weighted class. An automated report generation system in HDD industry (ARGS_PMS) was as a case study measure performance each metric. We found that, enhanced tasks, an class (MS-MC) gave best...

Popular online trends detection from crowd becomes more and essential for both trend followers sellers. However, huge amount of posts, text images, has prevented to be manually processed. This article, focusing on mining, aims automatically extract popular trends. A case study is performed one the most discussion forum websites in Thailand - i.e., Pantip.com. The approach involves employing several unsupervised mining techniques, namely, TF-IDF HTML scores, supervised learning sentiment...

The massive volume of Twitter data has attracted much attention researchers to study their correlation with stock market. Tweets symbols can be identified by the prefix dollar sign or using some complex techniques. In this paper, we focus on discovering NASDAQ in a stream tweets. We propose simple but effective methodology recognize symbols. Stock from company list, WordNet, Wikipedia and sample tweets, as well classic method collocation discovery are employed filter stock-related...

This research proposes a method to dealing with multiple linear regression that integrates the seasonality as well effects of some special or unanticipated events for sales figures. The is then applied car figures in Thailand after having been through 2011 national big flood and 2011-2012 government’s initiative tax-incentive program boosting automobile industry. Besides Thailand’s Gross Domestic Products (GDP) 12-month Loan’s Interest Rate explanatory variables, seasonal dummy variables...

For the testing of container classes and algorithms or programs that operate on data in a container, these have property being homogeneous throughout container. We developed an approach for this situation called coverage testing, where automated test generation can systematically generate increasing size. Given program model, it be theoretically shown there exists sufficiently large set size N, such with larger than N does not detect more faults. A number experiments been conducted using C++...

Part-of-speech (POS) tagging is the process of assigning part-of-speech tag or other lexical class marker to each word in a sentence. It also one most important steps Natural Language Processing (NLP) task pipeline. There are several research works Myanmar POS implemented with different approaches. However, there only publicly available tagged corpus named myPOS corpus. The size this 11 thousand sentences. not enough train downstream NLP tasks, such as machine learning. For reason, we...

This research aims to apply artificial intelligence technology a manufacturing industry, specifically, forecast temperature and insulation values of motors from the CNC machine. Dataset motor sensors are collected forecasting models trained using four deep learning models, namely, multilayer perceptron (MLP), long-short term memory (LSTM), LSTM autoencoder, bidirectional (Bi-LSTM). Models evaluated by measuring deviation real values. Two measures, root mean square error (RMSE) absolute...

With the emergence of Semantic Web (or Linked Data), increased efforts have been made to automatically extract formalized semantic knowledge from natural language text. Most research work and tools for extraction are focusing on text in English language. In this work, wepresent our research-in-progress evaluating state-of-the-art Thai For purpose, we investigate existing literature group available into eight tasks. Our preliminary results survey show that there exist large gaps therefore...

Online shopping is increasingly popular for customers. However, there are also high concerns on online frauds. In order to alleviate this concern, a proposed Social Commerce pages' Credibility Analysis software (SCCA) developed. The web application can provide up-to-date information related the user's chosen social commerce pages Facebook. It provides figures and statistics that indicate page's credibility. addition, it sentiment scores, both positive negative, which analyzed based...

Query expansion techniques aim to improve a userpsilas search by adding new query terms an existing query. This can be done in two ways: having human user refines the first result set or using automatic extraction of from set. Both usually depend on frequency Web pages only. paper proposes upon technique link analysis called hypertext induce topic selection algorithm (HITS). The preliminary questions TREC is presented.