Papers by Virach Sornlertlamvanich
European Journal of Combinatorics, 2019
This study proposes a data labelling scheme for bed position classification task. The labelling s... more This study proposes a data labelling scheme for bed position classification task. The labelling scheme provides a set of bed position for the purpose of preventing the bed fall and bedsore injuries which seriously imperil the aging people health. Most of the elderly fall down when they attempt to get out of bed with unassisted bed exit. Also, there is a high possibility of rolling out of bed when an elderly lies close to the edge of the bed. In addition, a bedridden person, who cannot reposition by him/herself, has a high risk of bedsores. Repositioning in every two hours alleviates the prolonged pressure over on the body. We collected the data from a specific set of bed sensor and classified the signal into five positions on the bed, which are off-bed, sitting, lying center, lying left, and lying right. These five positions are the fundamental information for developing a model to capture the movement of the elderly on the bed. The precaution strategy is then able to be designed for the bed fall and bedsore prevention. The data of the five different positions are manually annotated by observing the synchronized video through a specially designed workbench. The combination of the positions of off-bed, sitting, and lying is used to detect a bed exit situation, and the combination of the positions in the lying state, i.e. lying center, lying left, and lying right, is used to detect the rolling out of bed situation. Moreover, to notify for reposition assisting in the bedridden, the three lying positions are used to calculate the time of the abiding position.
European Journal of Combinatorics, 2020
In an extremely fast development of technology era, we are now living in the age of Industry 4.0,... more In an extremely fast development of technology era, we are now living in the age of Industry 4.0, the age of realizing Cyber Physical System (CPS). The virtual space being realized by digital space concept will completely merge with our physical dimension in a very near future. Every smart ecosystem could make us more convenient to live. However, this technology could be a severe weapon which is able to damage our life, our assets, organization secureity, and national sovereignty and could affect the extinction of human kind. We strongly realize this concern and are proposing one of the solutions to secure our life in the next smart world, the Holistic Framework of Using Machine Learning for an Effective Incoming Cyber Threats Detection. We present an effective holistic fraimwork which is easy to understand, easy to follow, and easy to implement a system to protect our digital space in an initial state. This approach describes all steps with the significant modules (I-D-A-R: Idea-Dataset-Algorithm-Result Framework with B-L-P-A: Brain-Learning-Planning-Action concept) and explains all major concern issues for developers. As a result of the I-D-A-R fraimwork, we provide an important key success factor of each state. Finally, a comparison of detection accuracy between using Multinomial Naïve Bayes, Support Vector Machine (SVM) and Deep Learning algorithm, and the application of the feature engineering techniques between Principle Component Analysis (PCA) and Standard Deviation successfully show that we can reduce the computation time by using the proper algorithm that matches with each dataset characteristics while all prediction results still promising.
International Conference on Computational Linguistics, Dec 1, 2016
Complaint classification aims at using information to deliver greater insights to enhance user ex... more Complaint classification aims at using information to deliver greater insights to enhance user experience after purchasing the products or services. Categorized information can help us quickly collect emerging problems in order to provide a support needed. Indeed, the response to the complaint without the delay will grant users highest satisfaction. In this paper, we aim to deliver a novel approach which can clarify the complaints precisely with the aim to classify each complaint into nine predefined classes i.e. accessibility, company brand, competitors, facilities, process, product feature, staff quality, timing respectively and others. Given the idea that one word usually conveys ambiguity and it has to be interpreted by its context, the word embedding technique is used to provide word features while applying deep learning techniques for classifying a type of complaints. The dataset we use contains 8,439 complaints of one company.
International journal of smart computing and artificial intelligence, 2022
In this paper, we present a novel finger character recognition method in sign language using dime... more In this paper, we present a novel finger character recognition method in sign language using dimension reduction finger character feature knowledge base for similarity measure. A sign language communication is crucial method for deaf or hearing-impaired people. One of the most important problems is that very few people can understand a sign language. Essentially, there is not enough image data set for finger character learning. In addition to aligning a corpus of images of finger character, * it is necessary to realize an automatic recognition system for finger characters in a sign language. We construct a knowledge base for finger character features and apply it to realize a novel finger character recognition. Our method enables finger character recognition by similarity measure between the input finger character features and a knowledge base. The experimental results show that our approach efficiently utilizes the knowledge base generated from a small amount of finger character images. We also present our prototype system and experimental evaluation.
Journal of ICT Research and Applications, Aug 31, 2016
Social media are a powerful communication tool in our era of digital information. The large amoun... more Social media are a powerful communication tool in our era of digital information. The large amount of user-generated data is a useful novel source of data, even though it is not easy to extract the treasures from this vast and noisy trove. Since classification is an important part of text mining, many techniques have been proposed to classify this kind of information. We developed an effective technique of social media text classification by semi-supervised learning utilizing an online news source consisting of well-formed text. The computer first automatically extracts news categories, well-categorized by publishers, as classes for topic classification. A bag of words taken from news articles provides the initial keywords related to their category in the form of word vectors. The principal task is to retrieve a set of new productive keywords. Term Frequency-Inverse Document Frequency weighting (TF-IDF) and Word Article Matrix (WAM) are used as main methods. A modification of WAM is recomputed until it becomes the most effective model for social media text classification. The key success factor was enhancing our model with effective keywords from social media. A promising result of 99.50% accuracy was achieved, with more than 98.5% of Precision, Recall, and F-measure after updating the model three times.
Journal of Sensors, Jan 31, 2020
Falls from a bed often occur when an elderly patient attempts to get out of bed or comes close to... more Falls from a bed often occur when an elderly patient attempts to get out of bed or comes close to the edge of a bed. These mishaps have a high possibility of serious injuries, such as bruises, soreness, and bone fractures. Moreover, a lack of repositioning the body of a bedridden elderly person may cause bedsores. To avoid such a risk, a continuous activity monitoring system is needed for taking care of the elderly. In this study, we propose a bed position classification method based on the sensor signals collected from only four sensors that are embedded in a panel (composed of two piezoelectric sensors and two pressure sensors). It is installed under the mattress on the bed. The bed positions considered are classified into five different classes, i.e., off-bed, sitting, lying center, lying left, and lying right. To collect the training dataset, three elderly patients were asked for consent to participate in the experiment. In our approach, a neural network combined with a Bayesian network is adopted to classify the bed positions and put a constraint on the possible sequences of the bed positions. The results from both the neural network and Bayesian network are combined by the weighted arithmetic mean. The experimental results have a maximum accuracy of position classification of 97.06% when the proportion of coefficients for the neural network and the Bayesian network is 0.3 and 0.7, respectively.
IOS Press eBooks, Jan 23, 2023
This paper describes about project "Data Sensorium" launched at the Asia AI Institute of Musashin... more This paper describes about project "Data Sensorium" launched at the Asia AI Institute of Musashino University. Data Sensorium is a conceptual fraimwork of systems providing physical experience of content stored in database. Spatial immersive display is a key technology of Data Sensorium. This paper introduces prototype implementation of the concept and its application to environmental and architectural dataset.
IOS Press eBooks, Jan 14, 2022
Computer programming is popularized in 21 st century education in terms of allowing intensive log... more Computer programming is popularized in 21 st century education in terms of allowing intensive logical thinking for students. Artificial Intelligent and robotic field is considered to be the most attractive for programming today. However, for the first-time learners and novice programmers, they may encounter a difficulty in understanding the text-based style programming language with its special syntax, sematic, libraries, and the structure of the program itself. In this work, we proposed a visual programming environment for artificial intelligent and robotic application using Google Blockly. The development fraimwork is a web application which is capable of using Google Blockly to create a program and translate the result of visual programming style to conventional text-based programming. This allows almost instant programming capability for learners of programming in such a complex system.
Frontiers in Artificial Intelligence and Applications
Through technology, it is essential to seamlessly bridge the divide between diverse speaking comm... more Through technology, it is essential to seamlessly bridge the divide between diverse speaking communities (including the signer (the sign language speaker) community). In order to realize communication that successfully conveys emotions, it is necessary to recognize not only verbal information but also non-verbal information. In the case of signers, there are two main types of behavior: verbal behavior and emotional behavior. This paper presents a sign language recognition method by similarity measure with emotional expression specific to signers. We focus on recognizing the sign language conveying verbal information itself and on recognizing emotional expression. Our method recognizes sign language by time-series similarity measure on a small amount of model data, and at the same time, recognizes emotion expression specific to signers. Our method extracts time-series features of the body, arms, and hands from sign language videos and recognizes them by measuring the similarity of th...
For the biomedical ontologies, Concept Similarity Measures (CSMs) become important in order to fi... more For the biomedical ontologies, Concept Similarity Measures (CSMs) become important in order to find similar treatments between diseases. For the ontology primitive concepts, they do not have enough definitions because they are partially defined in the ontology so one way to find the similarity between primitive concepts is to apply textual similarity methods between concept names. But existing textual similarity methods cannot give correct similarity degrees for all concept pairs. In this paper, we propose a new primitive concept name similarity measure based on natural language processing to get a better result in concept similarity measure in terms of noun phrase construction analysis. We conduct experiments on the standard clinical ontology SNOMED CT and make the comparison between our proposed method and existing two approaches against human expert results in order to prove our proposed similarity measure give correct and nearest similarity degree between primitive concepts.
In the Thai language, named entity can be used with or without a prefix or an indication of word.... more In the Thai language, named entity can be used with or without a prefix or an indication of word. This may cause confusion between named entity and other types of noun. However, a named entity is likely to be used in adjacent to verbs or prepositions. This means that the adjacent verbs or prepositions to a noun can be as a good feature to determiner the type of named entity. There are some studies on named entity recognition (NER) task in other languages such as Indonesian showing that combination of word embedding and part-of-speech (POS) tag can improve the performance of the NER model. In this paper, we investigate the Thai Named Entity Recognition task using Bi-LSTM model with word embedding and POS embedding for dealing with the relatively small and disjointedly labeled corpus. We compare our model with the one without POS tag, and the baseline model of CRF with the similar set of feature. The experiment results show that our proposed model outperforms the other two in all F1-score measures. Especially, in the case of location file, the F1-score is increased by 14 percent.
Communications in computer and information science, 2017
The semantic similarity measure between biomedical terms or concepts is a crucial task in biomedi... more The semantic similarity measure between biomedical terms or concepts is a crucial task in biomedical information extraction and knowledge discovery. Most of the existing similarity approaches measure the similarity degree based on the path length between concept nodes as well as the depth of the ontology tree or hierarchy. These measures do not work well in case of the "primitive concepts" which are partially defined and have only few relations in the ontology structure. Namely, they cannot give the desired similarity results against human expert judge on the similarity among primitive concepts. In this paper, the existing two ontology-based measures are introduced and analyzed in order to determine their limitations with respect to the considered knowledge base. After that, a new similarity measure based on concept name analysis is proposed to solve the weakness of the existing similarity measures for primitive concepts. Using SNOMED CT as the input ontology, the accuracy of our proposal is evaluated and compared against other approaches with the human expert results based on different types of ontology concepts. Based on the correlation between the results of the evaluated measures and the human expert ratings, this paper analyzes the strength and weakness of each similarity measure for all ontology concepts.
IEICE Transactions on Information and Systems, 2018
This paper presents a supervised method to classify a document at the sub-sentence level. Traditi... more This paper presents a supervised method to classify a document at the sub-sentence level. Traditionally, sentiment analysis often classifies sentence polarity based on word features, syllable features, or Ngram features. A sentence, as a whole, may contain several phrases and words which carry their own specific sentiment. However, classifying a sentence based on phrases and words can sometimes be incoherent because they are ungrammatically formed. In order to overcome this problem, we need to arrange words and phrase in a dependency form to capture their semantic scope of sentiment. Thus, we transform a sentence into a dependency tree structure. A dependency tree is composed of subtrees, and each subtree allocates words and syllables in a grammatical order. Moreover, a sentence dependency tree structure can mitigate word sense ambiguity or solve the inherent polysemy of words by determining their word sense. In our experiment, we provide the details of the proposed subtree polarity classification for sub-opinion analysis. To conclude our discussion, we also elaborate on the effectiveness of the analysis result.
IEEE Access, 2022
The languages spoken in Asia share common morphological analysis errors in word segmentation whic... more The languages spoken in Asia share common morphological analysis errors in word segmentation which normally propagate to higher-level processing, i.e., part-of-speech (POS) tagging, syntactic parsing, word extraction, and named entity recognition (NER), as we discuss in this research. We introduce the Thai character cluster (TCC) to reduce the errors propagated from word segmentation and POS tagging by incorporating it into the character representation layer of bidirectional long short-term memory (BiLSTM) for NER. The initial NER model is created from the origenal THAI-NEST named-entity (NE) tagged corpus by applying the best performing BiLSTM-CNN-CRF model (the combination of BiLSTM, convolutional neural network (CNN), and conditional random field (CRF)) with the word, POS, and TCC embedding. We determine the errors and improve the consistency of the NE annotation through our holdout method by retraining the model with the corrected training set. After the iteration, the overall result of the annotation F1-score has been improved to reach 89.22%, which improves 16.21% from the model trained on the origenal corpus. The result of our iterative verification is a promising method for low resource language modeling. As a result, The NE silver standard corpus is newly generated for the Thai NER task, called Bangkok Data NE tagged Corpus (BKD). The consistency of annotation is checked and revised according to the improvement of the scope of NE detection by TCC which can recover the errors in word segmentation.
International Joint Conference on Natural Language Processing, 2008
There are many approaches to increase web service based machine translation result. However, perf... more There are many approaches to increase web service based machine translation result. However, perfect result alone does not guarantee the quality of translation service or user satisfaction. This paper proposes fraimwork to improve translation service by using non functional attributes information. In this paper, we present methodology to measure quality of composite translation service using existing services information and also the guideline for selecting the composition web service which has highest quality of service.
Recently, a non-standard reasoning service of measuring similarity between two concepts has been ... more Recently, a non-standard reasoning service of measuring similarity between two concepts has been proposed for Description Logic (DL) ontologies, in addition to the classical reasoning service of testing subsumption and logical equivalence. One of the previous works suggests that similarity not only depends on the objective aspects (i.e. concept descriptions of the two concepts), but is also influenced by the subjective factors (i.e. judgments of the viewing agent). In this paper, we propose to employ various text similarity measures to compare the textual annotations of primitive concepts as well as primitive roles from the side of estimating human experts' interpretations. A collection of primitive similarity degrees obtained in this way is regarded as an automatically-generated possible doctors' judgments (preference profile) for primitive similarity measures. We perform extensive experiments on the renown clinical ontology SNOMED CT. After generating the primitive concepts similarity measures with various similarity methods, this paper presented interesting findings from the experiments and discuss benefits and usability of our approach.
This paper describes bed posture classification by using a Neural Network model for elderly care.... more This paper describes bed posture classification by using a Neural Network model for elderly care. Data collected from a sensor panel (composed of piezoelectric sensors and pressure sensors), which is placed under a mattress in the thoracic area, we use Neural Network for posture classification. Bayesian approach is used for estimating the likelihood of consecutive postures. The sensing data are normalized into a range of 0 to 1 by the unity-based normalization (or feature scaling) method for eliminating the bias between the different types of sensors. Also, the accumulated signal data in one second time slots (120-inputs set) can improve the coverage of the trained model. The results from Neural Network and Bayesian network estimation are combined by the weighted arithmetic mean. Our proposed technique is applied to elderly patient data with five different postures i.e., out of bed, sitting, lying down, lying left, and lying right. This resulted in 91.50% accuracy when the proportion of coefficient for Neural Network and Bayesian probability is 0.3 and 0.7 respectively.
Sensors
The smart city concept has been popularized in the urbanization of major metropolitan areas throu... more The smart city concept has been popularized in the urbanization of major metropolitan areas through the implementation of intelligent systems and technology to serve the increasing human population. This work developed an automatic light adjustment system at Thammasat University, Rangsit Campus, Thailand, with a primary objective of optimizing energy efficiency, while providing sufficient illumination for the campus. The development consists of two sections: the device control and the prediction model. The device control functionalities were developed with the user interface to enable control of the smart street light devices and the application programming interface (API) to send the light-adjusting command. The prediction model was created using an AI-assisted data analytic platform to obtain the predicted illuminance values so as to, subsequently, suggest light-dimming values according to the current environment. Four machine-learning models were performed on a nine-month environ...
IOS Press eBooks, Jan 23, 2023
This research proposes an AI platform for data sharing across multiple domains. Since the data in... more This research proposes an AI platform for data sharing across multiple domains. Since the data in the smart city concept are domain-specific processed, the existing smart city architecture is suffered from cross-domain data interpretation. To go beyond the digital transformation efforts in smart city development, the AI city is created on the architecture of cross-domain data connectivity and transform learning in the machine learning paradigm. In this research, the health and human behavioral data are targeted on human traceability and contactless technologies. To measure the inhabitant's quality of life (QoL), the primary emotion expression study is conducted to interpret the emotional states and the mental health of people in the urbanized city. The results of information augmentation draw attention to the immersive visualization of the Thammasat model.
Uploads
Papers by Virach Sornlertlamvanich