KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 11

Some of the metrics are blocked by your
consent settings
Thai Generalized Alignment Task (TGAT): A Corpus and Comparative Study for Hallucination Detection in Thai
(2025-01-01)
Wangwon, Sivakorn
;
Pitijaroonpong, Rapepong
;
Chuangkrud, Piyawat
;
Damrongrat, Chaianun
;
Kongyoung, Sarawoot
Large Language Models (LLMs) face critical reliability challenges due to hallucination-the generation of factually inaccurate content. While hallucination detection has advanced for major languages, Thai remains underserved, lacking specialized alignment models and datasets. We introduce the Thai Generalized Alignment Task (TGAT), a seven-subtask corpus tailored for Thai hallucination detection, spanning Fact Verification, Natural Language Inference, Information Retrieval, Question Answering, Summarization, Semantic Textual Similarity, and Paraphrase Identification. We conduct a comparative study across transformer families-encoder-only, encoder-decoder, and decoder-only-evaluated on the test set of each sub-task to analyze architectural compatibility and generalization for Thai hallucination detection. We also perform an ablation study to quantify how the presence of each subtask dataset affects the overall average performance, clarifying the role of dataset composition. Our experiments show that decoder-only models consistently outperform encoder-decoder and encoder-only alternatives under a standardized zero-shot prompting setup, establishing strong baselines and offering guidance for model selection and dataset design in low-resource settings.
Some of the metrics are blocked by your
consent settings
Impact of Rhythm, Tempo, and Rest Variations on Pitch Detection in Deep Learning-Based Piano Transcription Models
(2024-01-01)
Pangwapee, Priyakorn
;
Mekkoktanphira, Juthakan
;
Dilokthanakul, Nat
;
Lochanachit, Sirasit
;
Kanungsukkasem, Nont
This paper investigates the impact of rhythm, tempo, and rest variations on pitch detection in deep learning-based models for piano transcription. We conducted a series of experiments using GRU and Transformer architectures, manipulating note lengths, rhythmic patterns, and rest intervals to assess their effect on pitch transcription accuracy. Our findings indicate that model performance is significantly influenced by these musical factors. The experiment with GRU shows notable sensitivity to rhythmic and rest changes. However, the Transformer model handles varied conditions more robustly. These findings help refine our approach to music transcription software, particularly in improving pitch recognition across varied rhythmic patterns, tempos and rests.
Some of the metrics are blocked by your
consent settings
When are Latent Topics Useful for Text Mining?: Enriching Bag-of-Words Representations with Information Extraction in Thai News Articles
(2023-01-01)
Kanungsukkasem, Nont
;
Chuangkrud, Piyawat
;
Pitichotchokphokhin, Pimpitcha
;
Damrongrat, Chaianun
;
Leelanupab, Teerapong
The Bag-of-Words (BOW) model is simple but one of the successful representations of text documents. This model, however, suffers from the sparse matrix, in which most of the elements are zero. Topic modeling is an unsupervised learning method that can represent text documents in a low-dimensional space. Latent Dirichlet Allocation (LDA) is a topic modeling technique used for topic extraction and data exploration, with interpretable output. This paper presents a thorough study of potential benefits of applying LDA, as a feature extraction, to topic discovery and document classification in Thai news articles, comparing with TF–IDF and Word2Vec. We also studied how much of the top Thai terms extracted from LDA with the different numbers of topics can be interpretable and meaningful, and can be a representative of the corpus. Besides, a set of Topic Coherence measures were included in our study to estimate the degree of semantic similarity of extracted topics. To compare the performance and optimization time of classification of features from the different feature extraction methods, various types of classifiers, e.g., Logistic Regression, Random Forest, XGBoosting, etc., were experimented.
Some of the metrics are blocked by your
consent settings
Deep Neural Networks for the Qualitative Analysis of Myocardial Perfusion Emission Computed Tomography Images
(2023-01-01)
Pruthipanyasakul, Nareekarn
;
Kanungsukkasem, Nont
;
Urruty, Thierry
;
Leelanupab, Teerapong
Integrating AI into medical diagnosis can provide a more accurate diagnosis when medical staff make treatment decisions. This paper studied on several deep neural networks, re-used with further training for a specific task in classifying the stenosis of a patient's coronary artery. From a 4DM-SPECT application, we collected polar map images that report, for example, myocardial perfusion, function and defect severity from cardiac emission computed tomography examination. We conducted a comparative study to identify the optimal combination of various state-of-the-art pre-trained models (i.e., VGG19, ResNet50, DenseNet121, and EfficientNetB0-B3) and eight different modalities of the myocardial perfusion images for classifying the stenosis of the coronary artery.
Some of the metrics are blocked by your
consent settings
Discover Underlying Topics in Thai News Articles: A Comparative Study of Probabilistic and Matrix Factorization Approaches
(2020-06-01)
Pitichotchokphokhin, Pimpitcha
;
Chuangkrud, Piyawat
;
Kalakan, Kongkan
;
Suntisrivaraporn, Boontawee
;
Leelanupab, Teerapong
Topic modeling is an unsupervised learning approach, which can automatically discover the hidden thematic structure in text documents. For text mining, topic modeling is a language-independent technique that disregards grammar and word order. Apart from semantic and structural issues, Thai language is typically considered more complex than others. Due to the lack of word delimiter and a surfeit of composite words. Errors from word tokenization can create significant problems for any post processes of text, such as document retrieval, sentiment analysis, machine translation, etc., adversely decreasing the performance of text applications. Despite a strong correlation between word ordering and semantic meaning, topic modeling has been widely reported that it can extract latent information, aka. latent topic or latent semantic, encoded in documents. Although there were few previous research works on studying topic modeling in Thai language, they mostly focused on upstream processes of Natural Language Processing (NLP) in, for example, applying a refined stop-word list to, or adding N-gram on a single specific topic modeling method. To our knowledge, this paper is the first to explore different topic modeling approaches, i.e., Latent Dirichlet Allocation (LDA) and Nonnegative Metrix Factorization (NMF), in Thai Language to compare their coherence. We also employ and compare a set of state-of-the-art evaluation metrics based on Topic Coherence.
Some of the metrics are blocked by your
consent settings
A Scalable Service Architecture with Request Queuing for Resource-Intensive Tasks
(2020-06-01)
Chaowvasin, Kasidis
;
Sutanchaiyanonta, Pun
;
Kanungsukkasem, Nont
;
Leelanupab, Teerapong
Deploying Machine Learning (ML) prediction or Data Analytic (DA) process as a service in a Web API is not a trivial task. A number of settings and dependency requirements must be met to provide ML or DA successful solutions. In addition, an application that utilizes such an API needs to be always available to serve multiple users who can concurrently submit their requests. ML modeling or DA processing is a resource-intensive task, which can take a massive amount of time to process. Some tasks may take just a few minutes or hours while others may take several days to complete. In this paper, we design and develop a scalable architecture of API services for hosting ML models or DA functionalities in a production-grade deployment. The technologies of containerization and container orchestration, i.e., Docker and Kubernetes, have been employed to automate the deployment, scaling, and management of containerized ML or DA instances. To meet high-scale and high-availability requirements, the open-source message broker, i.e., RabbitMQ, is also used and containerized in Docker for scheduling multiple requests as task messages. These messages are then put into a task queue so that they will be processed later consecutively. Also, Nginx and Node.js with Express.js have been used and containerized as a web server and an API provider, respectively. We use a case-study of an intelligent system for processing documents about national research granting to validate our architecture.
Some of the metrics are blocked by your
consent settings
Financial Latent Dirichlet Allocation (FinLDA): Feature Extraction in Text and Data Mining for Financial Time Series Prediction
(2019-01-01)
Kanungsukkasem, Nont
;
Leelanupab, Teerapong
News has been an important source for many financial time series predictions based on fundamental analysis. However, digesting a massive amount of news and data published on the Internet to predict a market can be burdensome. This paper introduces a topic model based on latent Dirichlet allocation (LDA) to discover features from a combination of text, especially news articles and financial time series, denoted as Financial LDA (FinLDA). The features from FinLDA are served as additional input features for any machine learning algorithm to improve the prediction of the financial time series. We provide posterior distributions used in Gibbs sampling for two variants of the FinLDA and propose a framework for applying the FinLDA in a text and data mining for financial time series prediction. The experimental results show that the features from the FinLDA empirically add value to the prediction and give better results than the comparative features including topic distributions from the common LDA.
Some of the metrics are blocked by your
consent settings
Power of crowdsourcing in Twitter to find similar/related users
(2016-11-18)
Kanungsukkasem, Nont
;
Leelanupab, Teerapong
Identifying similar users in social media, e.g., Twitter, is very useful for a variety of applications, such as promoting social connections by user recommendation and defining the type of user accounts for target advertisement. Although many existing methods are available, especially that of Similar-To Framework of Twitter, they require a massive amount of data to process. This has become a major obstacle for non-commercial organizations and in particular academic researchers to study and analyze the similarity, due to the resource requirement to handle large datasets and the limitations of Twitter API to access Tweets per day. Accordingly, this paper proposes a new method that uses only a small amount of data and can be applied by not exceeding Twitter API limitations. Our method, called List Voting, uses only Lists feature that is provided by Twitter to determine accounts of similar users who are likely to produce similar contents. All the Lists are created by Twitter users. Thus, the List can be considered as a crowdsourcing. We also study the characteristics of the definition of crowdsourcing to confirm the consideration. Our experimental result shows that our method gets the benefit of the power of this crowdsourcing and can provide a list of users that are similar or related to a specified user.
Some of the metrics are blocked by your
consent settings
Snapboard: A shared space of visual snippets - A study in individual and asynchronous collaborative web search
(2015-01-01)
Leelanupab, Teerapong
;
Kruajirayu, Hannarin
;
Kanungsukkasem, Nont
People often engage in many search tasks that be collaborative, where two or more individuals work together with the joint information needs. We introduced and built CoZpace, a web-based application that enables a group of users to collaborate on searching the web. We also presented the main feature of CoZpace, named Snapboard, which is a shared board for a collection of group-created visual snippets. The visual snippet is a snapshot of focused and salient information captured by a user. It acts as a visual summarization of web pages, which allows any user to quickly recognize information and to revisit web pages. This paper describes example usage scenarios and initially investigates the ways Snapboard facilitates users in individual and asynchronous collaborative search. We then analyze users’ interactions and discuss how Snapboard supports search collaboration among study participants.
Some of the metrics are blocked by your
consent settings
Finding potential influences of a specific financial market in Twitter
(2015-01-01)
Kanungsukkasem, Nont
;
Leelanupab, Teerapong
This paper proposes a new framework to identify and rank Twitter accounts of which short messages or tweets may influence a specific financial stock price. In this paper, we start by mainly focusing on the first step of our framework, selecting potential influencers based on their association with a particular stock market. With numerous limitations of Twitter service to access and acquire Twitters data, a new methodology is also proposed to use a List feature to select potential financial influencers. It requires only Lists of accounts that contain an official Twitter account of the specific financial market, provided by Twitter's users, through Twitter API. Our methodology is designed to use only a small amount of data, by which it can be practically used under the limitation of Twitter API's crawling rate. Experimental results show that most of the potential influencers returned by our methodology are similar and related to the specific financial markets, which are companies listed on S&P 500 in this experiment. Most of the returned influencers are the official accounts of company or organization from the same sector and news media with a special emphasis on the same industry. Comparing to Twitter's user recommendation service (Who-To-Follow) and a crowdsourcing search for topic experts (Cognos), our methodology returns more related accounts in both percentage and the number.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results