One important obstruction against Thai COVID-19 recovery is fake news shared on social media that is one of the “Artificial Intelligence Open Issues against COVID-19” reported by Montreal.AI. Misinformation spread is one of the main cyber-security threats that should be filtered out as the IDS for maintaining COVID-19 information quality. To detect fake news in Thai texts, Thai-NLP techniques are necessary. This paper proposes a state-of-the-art Thai COVID-19 fake news detection among word relations using transfer learning models. For pre-training from the global open COVID-19 datasets, the source dataset is constructed by English to Thai translating. The novel feature shifting is formulated to enlarge Thai text examples in target dataset. Machine translation can be used for constructing Thai source dataset to cope with the lack of local dataset for future Thai-NLP applications. To lead the knowledge in Thai text understanding forward, feature shifting is a promising accuracy improvement in fine-tuning stage.

<span>Thai textual memes have been popular in social media, as a form of image information summarization. Unfortunately, many memes contain some hateful content that easily causes the controversy in Thailand. </span><span>For global protection, t</span><span>he </span><em><span>Hateful Memes Challenge</span></em><span> is also provided by </span><em><span>Facebook AI</span></em><span> to enable researchers to compete their algorithms for combating the hate speech on memes as one of </span><em><span>NeurIPS’20</span></em><span> competitions. As well as in Thailand, this paper introduces the Thai textual meme detection as a new research problem in Thai natural language processing (Thai-NLP) that is the settlement of transmission linkage between scene text localization, Thai optical recognition (Thai-OCR) and language understanding. From the results, both regular and irregular text position can be localized by one-stage detection pipeline. More scene text can be augmented by different resolution and rotation. The accuracy of Thai-OCR using convolutional neural network (CNN) can be improved by recurrent neural network (RNN). Since misspelling Thai words are frequently used in social, this paper categorizes them as synonyms to train on multi-task pre-trained language model. </span>

No abstract

Rice is a staple food for around 3.5 billion people in eastern, southern and south-east Asia. Prior to being rice, the rice-grain (grain) is previously husked and/or milled by the milling machine. Relevantly, the grain quality depends on its pureness of particular grain specie (without the mixing between different grain species). For the demand of grain purity inspection by an image, many researchers have proposed the grain classification (sometimes with localization) methods based on convolutional neural network (CNN). However, those papers are necessary to have a large number of labeling that was too expensive to be manually collected. In this paper, the image augmentation (rotation, brightness adjustment and horizontal flipping) is appiled to generate more number of grain images from the less data. From the results, image augmentation improves the performance in CNN and bagof-words model. For the future moving forward, the grain recognition can be easily done by less number of images.

Thai is a non-tonal language usage for 70 million speakers in Thailand. A variety of Thai handwrit-ten styles has been a challenge in handwriting recognition. In this paper, we propose a novel "ThaiWrittenNet" based on Convolutional Neural Network (ConvNet or CNN) with a cutout to identify the handwritten recognitions. Deep Belief Network (DBN) is also combined with Con-vNet to reduce network complexity. From the results, ThaiWrittenNet outperforms the flat Con-vNet and other handcrafted features with traditional machine learning algorithms. It appears that DBN helps ConvNet to improve the accuracy of Thai-handwritten recognition.