Pub Date : 2019-11-01DOI: 10.1109/IALP48816.2019.9037657
Yiqing Zhang, Hai Zhao, Zhuosheng Zhang
In this paper, we focus on an examination-style reading comprehension task which requires a multiple choice question solving but without a pre-given document that is supposed to contain direct evidences for answering the question. Unlike the common machine reading comprehension tasks, the concerned task requires a deep understanding into the detail-rich and semantically complex question. Such a reading comprehension task can be considered as a variant of early deep question-answering. We propose a hybrid solution to solve the problem. First, an attentive neural network to obtain the keywords in question. Then a retrieval based model is used to retrieve relative evidence in knowledge sources with the importance score of each word. The final choice is made by considering both question and evidence. Our experimental results show that our system gives state-of-the-art performance on Chinese benchmarks and shows its effectiveness on English dataset only using unstructured knowledge source.
{"title":"Examination-Style Reading Comprehension with Neural augmented Retrieval","authors":"Yiqing Zhang, Hai Zhao, Zhuosheng Zhang","doi":"10.1109/IALP48816.2019.9037657","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037657","url":null,"abstract":"In this paper, we focus on an examination-style reading comprehension task which requires a multiple choice question solving but without a pre-given document that is supposed to contain direct evidences for answering the question. Unlike the common machine reading comprehension tasks, the concerned task requires a deep understanding into the detail-rich and semantically complex question. Such a reading comprehension task can be considered as a variant of early deep question-answering. We propose a hybrid solution to solve the problem. First, an attentive neural network to obtain the keywords in question. Then a retrieval based model is used to retrieve relative evidence in knowledge sources with the importance score of each word. The final choice is made by considering both question and evidence. Our experimental results show that our system gives state-of-the-art performance on Chinese benchmarks and shows its effectiveness on English dataset only using unstructured knowledge source.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134196211","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037673
Ying Li, Hongyu Li, Jing Liu
In this paper, we focus on addressing the over-sensitivity issue of neural machine reading comprehension (MRC) models. By oversensitivity, we mean that the neural MRC models give different answers to question paraphrases that are semantically equivalent. To address this issue, we first create a large-scale Chinese MRC dataset with high-quality question paraphrases generated by a toolkit used in Baidu Search. Then, we quantitively analyze the oversensitivity issue of the neural MRC models on the dataset. Intuitively, if two questions are paraphrases of each other, a robust model should give the same predictions. Based on this intuition, we propose a regularized BERT-based model to encourage the model give the same predictions to similar inputs by leveraging high-quality question paraphrases. The experimental results show that our approaches can significantly improve the robustness of a strong BERT-based MRC model and achieve improvements over the BERT-based model in terms of held-out accuracy. Specifically, the different prediction ratio (DPR) for question paraphrases of the proposed model decreases more than 10%.
{"title":"Towards Robust Neural Machine Reading Comprehension via Question Paraphrases","authors":"Ying Li, Hongyu Li, Jing Liu","doi":"10.1109/IALP48816.2019.9037673","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037673","url":null,"abstract":"In this paper, we focus on addressing the over-sensitivity issue of neural machine reading comprehension (MRC) models. By oversensitivity, we mean that the neural MRC models give different answers to question paraphrases that are semantically equivalent. To address this issue, we first create a large-scale Chinese MRC dataset with high-quality question paraphrases generated by a toolkit used in Baidu Search. Then, we quantitively analyze the oversensitivity issue of the neural MRC models on the dataset. Intuitively, if two questions are paraphrases of each other, a robust model should give the same predictions. Based on this intuition, we propose a regularized BERT-based model to encourage the model give the same predictions to similar inputs by leveraging high-quality question paraphrases. The experimental results show that our approaches can significantly improve the robustness of a strong BERT-based MRC model and achieve improvements over the BERT-based model in terms of held-out accuracy. Specifically, the different prediction ratio (DPR) for question paraphrases of the proposed model decreases more than 10%.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133214835","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037720
Rui Cai, Wei Wei, Jinsong Zhang
The scoring feedback of Computer Assisted Pronunciation Training (CAPT) systems facilitate learner’s instant awareness of their problems, easily lead to more practices. But whether it is enough to instruct the learners to understand how to correct their errors is still unknown. To see in depth, the impacts from CAPT technology on language learning, and to investigate learner’s correction strategy after receiving error warnings, this paper studies long term learning data of Chinese utterances by a number of CSL (Chinese as a Second Language) learners, with special efforts paid to the utterances of nasal Finals. The data resulted from a 3-week use of a CAPT APP, called “SAIT汉语” for Chinese learning, by 10 learners with different mother tongues. Major findings include: 1) Improvements were seen with almost all kinds of phonemes, except nasal Finals; 2) Data analyses showed that the learners had tried to lengthen the nasal codas after they received error warnings, while Chinese native data shows a significant nasalization period before a short coda. These results suggest that the scoring feedback can be beneficial to pronunciation training in most cases, except for some special ones. For the sounds such as Chinese nasal Finals, more appropriate feedback method is desired.
{"title":"Are Scoring Feedback of CAPT Systems Helpful for Pronunciation Correction? –An Exception of Mandarin Nasal Finals","authors":"Rui Cai, Wei Wei, Jinsong Zhang","doi":"10.1109/IALP48816.2019.9037720","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037720","url":null,"abstract":"The scoring feedback of Computer Assisted Pronunciation Training (CAPT) systems facilitate learner’s instant awareness of their problems, easily lead to more practices. But whether it is enough to instruct the learners to understand how to correct their errors is still unknown. To see in depth, the impacts from CAPT technology on language learning, and to investigate learner’s correction strategy after receiving error warnings, this paper studies long term learning data of Chinese utterances by a number of CSL (Chinese as a Second Language) learners, with special efforts paid to the utterances of nasal Finals. The data resulted from a 3-week use of a CAPT APP, called “SAIT汉语” for Chinese learning, by 10 learners with different mother tongues. Major findings include: 1) Improvements were seen with almost all kinds of phonemes, except nasal Finals; 2) Data analyses showed that the learners had tried to lengthen the nasal codas after they received error warnings, while Chinese native data shows a significant nasalization period before a short coda. These results suggest that the scoring feedback can be beneficial to pronunciation training in most cases, except for some special ones. For the sounds such as Chinese nasal Finals, more appropriate feedback method is desired.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131242648","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037656
A. Hamdulla, Guzalnur Dilmurat, Gulnur Arkin, Mijit Ablimit
Phonetics is both an ancient and a young subject. Syllables are important units of speech. Based on the data requirements of speech synthesis and speech recognition, this paper studies from the perspective of experimental phonetics. Firstly, different syllable words are counted from the large-scale “Speech Acoustic Parameters Database of Uyghur Language”, including monosyllable words, two-syllable words, three-syllable words and four-syllable words. Secondly, the prosodic parameters are extracted, and statistical analysis is made. Accordingly, the duration distribution of different length words for male and female speakers are studied, and the fixed CV type syllable duration of consonant, the duration of vowel, the whole syllable duration and the pitch of syllable are extracted and analyzed. The effect of different vowels on the duration of CV syllables is further studied, and provided the reliable parameter basis for Uyghur speech synthesis and speech recognition.
{"title":"Statistical Analysis of Syllable Duration of Uyghur Language","authors":"A. Hamdulla, Guzalnur Dilmurat, Gulnur Arkin, Mijit Ablimit","doi":"10.1109/IALP48816.2019.9037656","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037656","url":null,"abstract":"Phonetics is both an ancient and a young subject. Syllables are important units of speech. Based on the data requirements of speech synthesis and speech recognition, this paper studies from the perspective of experimental phonetics. Firstly, different syllable words are counted from the large-scale “Speech Acoustic Parameters Database of Uyghur Language”, including monosyllable words, two-syllable words, three-syllable words and four-syllable words. Secondly, the prosodic parameters are extracted, and statistical analysis is made. Accordingly, the duration distribution of different length words for male and female speakers are studied, and the fixed CV type syllable duration of consonant, the duration of vowel, the whole syllable duration and the pitch of syllable are extracted and analyzed. The effect of different vowels on the duration of CV syllables is further studied, and provided the reliable parameter basis for Uyghur speech synthesis and speech recognition.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121591880","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037723
Jessica Naraiswari Arwidarasti, Ika Alfina, A. Krisnadhi
A constituency treebank is a key component for deep syntactic parsing of natural language sentences. For Indonesian, this task is unfortunately hindered by the fact that the only one constituency treebank publicly available is rather small with just over 1000 sentences, and not only that, it employs a format incompatible with readily available constituency treebank processing tools. In this work, we present a conversion of the existing Indonesian constituency treebank to the widely accepted Penn Treebank format. Specifically, the conversion adjusts the bracketing format for compound words as well as the POS tagset according to the Penn Treebank format. In addition, we revised the word segmentation and POS tagging of a number of tokens. Finally, we performed an evaluation on the treebank quality by employing the Shift-Reduce parser from Stanford CoreNLP to create a parser model. A 10-fold cross-validated experiment on the parser model yields an F1-score of 70.90%.
{"title":"Converting an Indonesian Constituency Treebank to the Penn Treebank Format","authors":"Jessica Naraiswari Arwidarasti, Ika Alfina, A. Krisnadhi","doi":"10.1109/IALP48816.2019.9037723","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037723","url":null,"abstract":"A constituency treebank is a key component for deep syntactic parsing of natural language sentences. For Indonesian, this task is unfortunately hindered by the fact that the only one constituency treebank publicly available is rather small with just over 1000 sentences, and not only that, it employs a format incompatible with readily available constituency treebank processing tools. In this work, we present a conversion of the existing Indonesian constituency treebank to the widely accepted Penn Treebank format. Specifically, the conversion adjusts the bracketing format for compound words as well as the POS tagset according to the Penn Treebank format. In addition, we revised the word segmentation and POS tagging of a number of tokens. Finally, we performed an evaluation on the treebank quality by employing the Shift-Reduce parser from Stanford CoreNLP to create a parser model. A 10-fold cross-validated experiment on the parser model yields an F1-score of 70.90%.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124191013","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037722
Jianyu Zheng, Kun Ma, Xuemei Tang, Shichen Liang
The translation activity involves both the source language and the target language. Compared to the standard texts in the two language, translated texts show unique language characteristics. In order to explore them from the perspective of integrality and complexity, we introduce complex network method into the study on translated Chinese. Firstly, selected the experimental texts from The ZJU Corpus of Translational Chinese (ZCTC) and its corresponding six sub-corpora, such as Press reportage and Popular lore. And then removed the punctuation and did word segmentation. Secondly, constructed a word co-occurrence network of translated Chinese. After analyzing and counting the parameters, such as their shortest path lengths, degree distributions and clustering coefficients in these networks, we verify that the word co-occurrence network of translated Chinese has small world effect and scale-free property. Finally, by constructing co-occurrence networks of standard Chinese and calculating their network parameters, we compare and verify the differences between translated Chinese and standard Chinese: “simplification” and the more usage of common words. Our work expands the application of complex network in translation studies, and provides a feasible approach for studying translated Chinese based on complex networks.
{"title":"Exploring Characteristics of Word Co-occurrence Network in Translated Chinese","authors":"Jianyu Zheng, Kun Ma, Xuemei Tang, Shichen Liang","doi":"10.1109/IALP48816.2019.9037722","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037722","url":null,"abstract":"The translation activity involves both the source language and the target language. Compared to the standard texts in the two language, translated texts show unique language characteristics. In order to explore them from the perspective of integrality and complexity, we introduce complex network method into the study on translated Chinese. Firstly, selected the experimental texts from The ZJU Corpus of Translational Chinese (ZCTC) and its corresponding six sub-corpora, such as Press reportage and Popular lore. And then removed the punctuation and did word segmentation. Secondly, constructed a word co-occurrence network of translated Chinese. After analyzing and counting the parameters, such as their shortest path lengths, degree distributions and clustering coefficients in these networks, we verify that the word co-occurrence network of translated Chinese has small world effect and scale-free property. Finally, by constructing co-occurrence networks of standard Chinese and calculating their network parameters, we compare and verify the differences between translated Chinese and standard Chinese: “simplification” and the more usage of common words. Our work expands the application of complex network in translation studies, and provides a feasible approach for studying translated Chinese based on complex networks.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122576708","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037672
Sufeng Duan, Hai Zhao, Junru Zhou, Rui Wang
Syntax has been shown a helpful clue in various natural language processing tasks including previous statistical machine translation and recurrent neural network based machine translation. However, since the state-of-the-art neural machine translation (NMT) has to be built on the Transformer based encoder, few attempts are found on such a syntax enhancement. Thus in this paper, we explore effective ways to introduce syntax into Transformer for better machine translation. We empirically compare two ways, positional encoding and input embedding, to exploit syntactic clues from dependency tree over source sentence. Our proposed methods have a merit keeping the architecture of Transformer unchanged, thus the efficiency of Transformer can be kept. The experimental results on IWSLT’ 14 German-to-English and WMT14 English-to-German show that our method can yield advanced results over strong Transformer baselines.
{"title":"Syntax-aware Transformer Encoder for Neural Machine Translation","authors":"Sufeng Duan, Hai Zhao, Junru Zhou, Rui Wang","doi":"10.1109/IALP48816.2019.9037672","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037672","url":null,"abstract":"Syntax has been shown a helpful clue in various natural language processing tasks including previous statistical machine translation and recurrent neural network based machine translation. However, since the state-of-the-art neural machine translation (NMT) has to be built on the Transformer based encoder, few attempts are found on such a syntax enhancement. Thus in this paper, we explore effective ways to introduce syntax into Transformer for better machine translation. We empirically compare two ways, positional encoding and input embedding, to exploit syntactic clues from dependency tree over source sentence. Our proposed methods have a merit keeping the architecture of Transformer unchanged, thus the efficiency of Transformer can be kept. The experimental results on IWSLT’ 14 German-to-English and WMT14 English-to-German show that our method can yield advanced results over strong Transformer baselines.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"46 6","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120922873","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037693
Huibin Ruan, Yu Hong, Yu Sun, Yang Xu, Min Zhang
Implicit causal relation recognition aims to identify the causal relation between a pair of arguments. It is a challenging task due to the lack of conjunctions and the shortage of labeled data. In order to improve the identification performance, we come up with an approach to expand the training dataset. On the basis of the hypothesis that there inherently exists causal relations in WHY-type Question-Answer (QA) pairs, we utilize WHY-type QA pairs for the training set expansion. In practice, we first collect WHY-type QA pairs from the Knowledge Bases (KBs) of the reading comprehension tasks, and then convert them into narrative argument pairs by Question-Statement Conversion (QSC). In order to alleviate redundancy, we use active learning (AL) to select informative samples from the synthetic argument pairs. The sampled synthetic argument pairs are added to the Penn Discourse Treebank (PDTB), and the expanded PDTB is used to retrain the neural network-based classifiers. Experiments show that our method yields a performance gain of 2.42% F 1-score when AL is used, and 1.61% without using.
{"title":"Using WHY-type Question-Answer Pairs to Improve Implicit Causal Relation Recognition","authors":"Huibin Ruan, Yu Hong, Yu Sun, Yang Xu, Min Zhang","doi":"10.1109/IALP48816.2019.9037693","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037693","url":null,"abstract":"Implicit causal relation recognition aims to identify the causal relation between a pair of arguments. It is a challenging task due to the lack of conjunctions and the shortage of labeled data. In order to improve the identification performance, we come up with an approach to expand the training dataset. On the basis of the hypothesis that there inherently exists causal relations in WHY-type Question-Answer (QA) pairs, we utilize WHY-type QA pairs for the training set expansion. In practice, we first collect WHY-type QA pairs from the Knowledge Bases (KBs) of the reading comprehension tasks, and then convert them into narrative argument pairs by Question-Statement Conversion (QSC). In order to alleviate redundancy, we use active learning (AL) to select informative samples from the synthetic argument pairs. The sampled synthetic argument pairs are added to the Penn Discourse Treebank (PDTB), and the expanded PDTB is used to retrain the neural network-based classifiers. Experiments show that our method yields a performance gain of 2.42% F 1-score when AL is used, and 1.61% without using.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120807035","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037733
Angelica H. De La Cruz, Rodolfo C. Raga
Speaker diarization pertains to the process of determining speaker identity at a given time in an audio stream. It was first used for speech recognition and over time became useful in other applications such as video captioning and speech transcription. Recently, deep learning techniques have been applied to speaker diarization with considerable success, however, deep learning are conventionally data intensive and collecting large training samples can be difficult and expensive to collect especially for resource scarce languages. This study focused on investigating a speaker diarization approach for meeting room conversations in the Filipino language. To compensate for lack of resources, a one shot learning strategy was explored using Siamese neural network. Among the experiments conducted, the lowest diarization error rate yielded to 46%. There are, however, more parameters that can be tuned to improve the diarization results. To the best of our knowledge, no work in speaker diarization dedicated for Filipino language has yet been done.
{"title":"Development of a Filipino Speaker Diarization in Meeting Room Conversations","authors":"Angelica H. De La Cruz, Rodolfo C. Raga","doi":"10.1109/IALP48816.2019.9037733","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037733","url":null,"abstract":"Speaker diarization pertains to the process of determining speaker identity at a given time in an audio stream. It was first used for speech recognition and over time became useful in other applications such as video captioning and speech transcription. Recently, deep learning techniques have been applied to speaker diarization with considerable success, however, deep learning are conventionally data intensive and collecting large training samples can be difficult and expensive to collect especially for resource scarce languages. This study focused on investigating a speaker diarization approach for meeting room conversations in the Filipino language. To compensate for lack of resources, a one shot learning strategy was explored using Siamese neural network. Among the experiments conducted, the lowest diarization error rate yielded to 46%. There are, however, more parameters that can be tuned to improve the diarization results. To the best of our knowledge, no work in speaker diarization dedicated for Filipino language has yet been done.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"162 6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129175650","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 : 2019-11-01DOI: 10.1109/IALP48816.2019.9037706
Zhensong Li, Jie Zhu, Zhixiang Luo, Saihu Liu
Text categorization is an important task in natural language processing, and it has a wide range of applications in real life. In this paper, two N-Gram feature models (MLP, FastText) and two sequential models (sepCNN, Bi-LSTM) are used to study the automatic classification for Tibetan text based on syllables and vocabulary. The experiment on Tibetan language data collected by China Tibet News Network shows that the classification accuracy is about 85%.
{"title":"Research on Tibetan Text Classification Method Based on Neural Network","authors":"Zhensong Li, Jie Zhu, Zhixiang Luo, Saihu Liu","doi":"10.1109/IALP48816.2019.9037706","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037706","url":null,"abstract":"Text categorization is an important task in natural language processing, and it has a wide range of applications in real life. In this paper, two N-Gram feature models (MLP, FastText) and two sequential models (sepCNN, Bi-LSTM) are used to study the automatic classification for Tibetan text based on syllables and vocabulary. The experiment on Tibetan language data collected by China Tibet News Network shows that the classification accuracy is about 85%.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132634783","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：