Pub Date : 2021-12-21DOI: 10.1109/iSAI-NLP54397.2021.9678167
S. Marukatat
During last decade, language modeling has been dominated by neural structures; RNN, LSTM or Transformer. These neural language models provide excellent performance to the detriment of very high computational cost. This work investigates the use of probabilistic language model that requires much less computational cost. In particular, we are interested in variable-order Markov model that can be efficiently implemented on a probabilistic suffix tree (PST) structure. The PST construction is cheap and can be easily scaled to very large dataset. Experimental results show that this model can be used to generated realistic sentences.
{"title":"Text generation by probabilistic suffix tree language model","authors":"S. Marukatat","doi":"10.1109/iSAI-NLP54397.2021.9678167","DOIUrl":"https://doi.org/10.1109/iSAI-NLP54397.2021.9678167","url":null,"abstract":"During last decade, language modeling has been dominated by neural structures; RNN, LSTM or Transformer. These neural language models provide excellent performance to the detriment of very high computational cost. This work investigates the use of probabilistic language model that requires much less computational cost. In particular, we are interested in variable-order Markov model that can be efficiently implemented on a probabilistic suffix tree (PST) structure. The PST construction is cheap and can be easily scaled to very large dataset. Experimental results show that this model can be used to generated realistic sentences.","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124060241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/iSAI-NLP54397.2021.9678156
We examined the Named Entity Recognition (NER) of organizations in the Thai government’s project documents using a simple data masking technique with the help of an external dictionary. Our framework demonstrated its potential in the case that the external dictionary was incomplete and might not be used to label the training data exhaustively. A data masking technique on the administrative area part of the organization names was employed in an attempt to discover more organization entities outside the dictionary. The experimental results showed that our model gained higher recall while sacrificing a relatively small amount of precision. The proposed approach was also capable of recognizing entities which were never seen in the dictionary.
{"title":"Named Entity Recognition of Thai Documents using CRF with a Simple Data Masking Technique","authors":"","doi":"10.1109/iSAI-NLP54397.2021.9678156","DOIUrl":"https://doi.org/10.1109/iSAI-NLP54397.2021.9678156","url":null,"abstract":"We examined the Named Entity Recognition (NER) of organizations in the Thai government’s project documents using a simple data masking technique with the help of an external dictionary. Our framework demonstrated its potential in the case that the external dictionary was incomplete and might not be used to label the training data exhaustively. A data masking technique on the administrative area part of the organization names was employed in an attempt to discover more organization entities outside the dictionary. The experimental results showed that our model gained higher recall while sacrificing a relatively small amount of precision. The proposed approach was also capable of recognizing entities which were never seen in the dictionary.","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129696791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/isai-nlp54397.2021.9678191
{"title":"iSAI-NLP Committee","authors":"","doi":"10.1109/isai-nlp54397.2021.9678191","DOIUrl":"https://doi.org/10.1109/isai-nlp54397.2021.9678191","url":null,"abstract":"","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129050097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/iSAI-NLP54397.2021.9678177
U. Puangragsa, Pitchayakorn Lomvisai, P. Phasukkit, Sarut Puangragsa, J. Setakornnukul, Nongluck Houngkamhang, Petchanon Thongserm, P. Dankulchai
4-Dimensional computed tomography (4DCT) is the most common technique to determine organ movement due to breathing motion. However, the ability of 4DCT to acquire CT images as a function of the respiratory phase increases higher radiation dose. To reduce the patient’s radiation dose, this study created lung motion prediction models used to estimate tumor target movement in ten respiratory phases by detecting only external organ movement during a complete respiration cycle without radiation with Kinect. The average overall amplitude difference between RPM and Kinect signals in the phantom experiment was 0.02 ± 0.1 mm. F1 score of 100% for all most all classifications except classification 2,3,6,7 and 8 of 85%,83%,90%, 84%,85% where irregular breathing pattern. Essentially, the proposed tumor movement scheme’s total accuracy (average of F1 scores) is 92.7 %. Deep learning model can predict tumor motion range and classification zone by used detection of the external respiratory signal
{"title":"Feasibility of Prediction Model for Internal Tumor Target Volume from 4-D Computed Tomography of Lung cancer","authors":"U. Puangragsa, Pitchayakorn Lomvisai, P. Phasukkit, Sarut Puangragsa, J. Setakornnukul, Nongluck Houngkamhang, Petchanon Thongserm, P. Dankulchai","doi":"10.1109/iSAI-NLP54397.2021.9678177","DOIUrl":"https://doi.org/10.1109/iSAI-NLP54397.2021.9678177","url":null,"abstract":"4-Dimensional computed tomography (4DCT) is the most common technique to determine organ movement due to breathing motion. However, the ability of 4DCT to acquire CT images as a function of the respiratory phase increases higher radiation dose. To reduce the patient’s radiation dose, this study created lung motion prediction models used to estimate tumor target movement in ten respiratory phases by detecting only external organ movement during a complete respiration cycle without radiation with Kinect. The average overall amplitude difference between RPM and Kinect signals in the phantom experiment was 0.02 ± 0.1 mm. F1 score of 100% for all most all classifications except classification 2,3,6,7 and 8 of 85%,83%,90%, 84%,85% where irregular breathing pattern. Essentially, the proposed tumor movement scheme’s total accuracy (average of F1 scores) is 92.7 %. Deep learning model can predict tumor motion range and classification zone by used detection of the external respiratory signal","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122694083","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-21DOI: 10.1109/iSAI-NLP54397.2021.9678164
Anuwat Chaiwongyen, Kanokkarn Pinkeaw, W. Kongprawechnon, Jessada Karnjana, M. Unoki
This paper proposes, effective method for replay attack detection used in an automatic speaker verification system. The replay attack is of interest because it is the most straightforward and effective attack and is challenging to detect. It is a playback of the recording of the voice of a target speaker. From the literature, no speech features work well with all classifiers, and there is no investigation of using ResNet-based model, called ResNeWt, with linear frequency cepstral coefficient (LFCC). Therefore, a replay attack detection model based on 18-layer ResNeWt that takes LFCCs as the input, was constructed in this paper. The proposes method was tested on a dataset provided by ASVspoof 2019 competition. In terms of the equal error rate (EER), the proposed method is the best in all existing methods, with an EER of 0.29%. The comparison in terms of replay attack detection was also made in detail. The performance of the proposed method in terms of the balanced accuracy, precision, recall, and F1-score was considerably better than existing methods.
{"title":"Replay Attack Detection in Automatic Speaker Verification Based on ResNeWt18 with Linear Frequency Cepstral Coefficients","authors":"Anuwat Chaiwongyen, Kanokkarn Pinkeaw, W. Kongprawechnon, Jessada Karnjana, M. Unoki","doi":"10.1109/iSAI-NLP54397.2021.9678164","DOIUrl":"https://doi.org/10.1109/iSAI-NLP54397.2021.9678164","url":null,"abstract":"This paper proposes, effective method for replay attack detection used in an automatic speaker verification system. The replay attack is of interest because it is the most straightforward and effective attack and is challenging to detect. It is a playback of the recording of the voice of a target speaker. From the literature, no speech features work well with all classifiers, and there is no investigation of using ResNet-based model, called ResNeWt, with linear frequency cepstral coefficient (LFCC). Therefore, a replay attack detection model based on 18-layer ResNeWt that takes LFCCs as the input, was constructed in this paper. The proposes method was tested on a dataset provided by ASVspoof 2019 competition. In terms of the equal error rate (EER), the proposed method is the best in all existing methods, with an EER of 0.29%. The comparison in terms of replay attack detection was also made in detail. The performance of the proposed method in terms of the balanced accuracy, precision, recall, and F1-score was considerably better than existing methods.","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114887132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A data category recommendation framework for Thailand’s open government data portal (ThOGD) is proposed to assist data providers when publishing and registering a new dataset into the portal’s data catalog. However, existing approaches such as a multi-label classification problem, have not adopted the semantic features of data categories sufficiently. Deep learning model for Natural Language Processing has recently demonstrated to achieve high potential in learning the different degrees of semantic feature abstraction because all layers of multi-head attention blocks are provided with different fragments of metadata descriptions and corresponding tags. To obtain a robust recommendation result, this paper proposes DataCat: a Category Recommendation Framework using the attention-based framework through the ThOGD portal. Within this framework, the integrated multi-layers with particular semantic information are directly attached to the output layer of a network to enhance the effectiveness of information retrieval. The results point out that the attention-based framework has a weighted effect on loss of optimization. The performance when looking at the macro average of precision and F1-score improves by 0.664% and 0.557%, respectively. The micro average of those improves by 0.806%, and 0.698%, respectively.
提出了一个针对泰国开放政府数据门户(ThOGD)的数据类别推荐框架，以帮助数据提供者将新数据集发布并注册到门户的数据目录中。然而，现有的方法，如多标签分类问题，并没有充分利用数据类别的语义特征。自然语言处理的深度学习模型在学习不同程度的语义特征抽象方面具有很大的潜力，因为多层多头注意块的每一层都提供了不同的元数据描述片段和相应的标签。为了获得稳健的推荐结果，本文通过ThOGD门户提出了基于注意力的类别推荐框架DataCat: a Category recommendation Framework。在该框架中，将具有特定语义信息的集成多层直接附加到网络的输出层，以提高信息检索的有效性。结果表明，基于注意力的框架对优化损失具有加权效应。当观察精度和f1分数的宏观平均值时，性能分别提高了0.664%和0.557%。其微观平均值分别提高了0.806%和0.698%。
{"title":"DataCat: Attention-based Open Government Data (OGD) Category Recommendation Framework","authors":"Natnaree Sornkongdang, Nuttapong Sanglerdsinlapachai, Chutiporn Anutariya","doi":"10.1109/iSAI-NLP54397.2021.9678174","DOIUrl":"https://doi.org/10.1109/iSAI-NLP54397.2021.9678174","url":null,"abstract":"A data category recommendation framework for Thailand’s open government data portal (ThOGD) is proposed to assist data providers when publishing and registering a new dataset into the portal’s data catalog. However, existing approaches such as a multi-label classification problem, have not adopted the semantic features of data categories sufficiently. Deep learning model for Natural Language Processing has recently demonstrated to achieve high potential in learning the different degrees of semantic feature abstraction because all layers of multi-head attention blocks are provided with different fragments of metadata descriptions and corresponding tags. To obtain a robust recommendation result, this paper proposes DataCat: a Category Recommendation Framework using the attention-based framework through the ThOGD portal. Within this framework, the integrated multi-layers with particular semantic information are directly attached to the output layer of a network to enhance the effectiveness of information retrieval. The results point out that the attention-based framework has a weighted effect on loss of optimization. The performance when looking at the macro average of precision and F1-score improves by 0.664% and 0.557%, respectively. The micro average of those improves by 0.806%, and 0.698%, respectively.","PeriodicalId":339826,"journal":{"name":"2021 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP)","volume":"720 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133181170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：