Pub Date : 2019-11-01DOI: 10.1109/IALP48816.2019.9037683
Jiali Chen, Yu Hong, Jingli Zhang, Jianmin Yao
Sentence-level event detection (ED) is a task of detecting words that describe specific types of events, including the subtasks of trigger word identification and event type classification. Previous work straightforwardly inputs a sentence into neural classification models and analyzes deep semantics of words in the sentence one by one. Relying on the semantics, probabilities of event classes can be predicted for each word, including the carefully defined ACE event classes and a “N/A” class(i.e., non-trigger word). The models achieve remarkable successes nowadays. However, our findings show that a natural sentence may posses more than one trigger word and thus entail different types of events. In particular, the closely related information of each event only lies in a unique sentence segment but has nothing to do with other segments. In order to reduce negative influences from noises in other segments, we propose to perform semantics learning for event detection only in the scope of segment instead of the whole sentence. Accordingly, we develop a novel ED method which integrates sentence segmentation into the neural event classification architecture. Bidirectional Long Short-Term Memory (Bi-LSTM) with multi-head attention is used as the classification model. Sentence segmentation is boiled down to a sequence labeling problem, where BERT is used. We combine embeddings, and use them as the input of the neural classification model. The experimental results show that the performance of our method reaches 76.8% and 74.2% $F_{1}-$scores for trigger identification and event type classification, which outperforms the state-of-the-art.
{"title":"Using Mention Segmentation to Improve Event Detection with Multi-head Attention","authors":"Jiali Chen, Yu Hong, Jingli Zhang, Jianmin Yao","doi":"10.1109/IALP48816.2019.9037683","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037683","url":null,"abstract":"Sentence-level event detection (ED) is a task of detecting words that describe specific types of events, including the subtasks of trigger word identification and event type classification. Previous work straightforwardly inputs a sentence into neural classification models and analyzes deep semantics of words in the sentence one by one. Relying on the semantics, probabilities of event classes can be predicted for each word, including the carefully defined ACE event classes and a “N/A” class(i.e., non-trigger word). The models achieve remarkable successes nowadays. However, our findings show that a natural sentence may posses more than one trigger word and thus entail different types of events. In particular, the closely related information of each event only lies in a unique sentence segment but has nothing to do with other segments. In order to reduce negative influences from noises in other segments, we propose to perform semantics learning for event detection only in the scope of segment instead of the whole sentence. Accordingly, we develop a novel ED method which integrates sentence segmentation into the neural event classification architecture. Bidirectional Long Short-Term Memory (Bi-LSTM) with multi-head attention is used as the classification model. Sentence segmentation is boiled down to a sequence labeling problem, where BERT is used. We combine embeddings, and use them as the input of the neural classification model. The experimental results show that the performance of our method reaches 76.8% and 74.2% $F_{1}-$scores for trigger identification and event type classification, which outperforms the state-of-the-art.","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":"129631229","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/IALP48816.2019.9037669
Kun Ma, Lijiao Yang
To automate the graded reading task, we urgently need to extract and calculate the important index of the complexity of the relationship between the characters affecting the plot complexity of narrative literature. In order to realize this purpose, this paper describes a computational method for automatic analysis of the virtual social network from children’s literature works. We selected the required bibliography for primary school students recommended by the Ministry of Education, then automatically extract the characters of the novel by CRF, and constructs the character network based on the co-occurrence relationship. The statistical analysis method of complex network provides a quantitative basis for distinguishing the complexity of characters’ relationships in different texts. The results show that the structural characteristics of character interaction networks are similar to those of small world networks, and the selected network measurement indexes are significantly related to the complexity of text characters. Finally, we achieved effectively evaluating and predicting the complexity of the social networks from more extensive literature works some classical regression model based on machine learning.
{"title":"Automatic Extraction and Quantitative Evaluation of the Character Relationship Networks from Children’s Literature works","authors":"Kun Ma, Lijiao Yang","doi":"10.1109/IALP48816.2019.9037669","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037669","url":null,"abstract":"To automate the graded reading task, we urgently need to extract and calculate the important index of the complexity of the relationship between the characters affecting the plot complexity of narrative literature. In order to realize this purpose, this paper describes a computational method for automatic analysis of the virtual social network from children’s literature works. We selected the required bibliography for primary school students recommended by the Ministry of Education, then automatically extract the characters of the novel by CRF, and constructs the character network based on the co-occurrence relationship. The statistical analysis method of complex network provides a quantitative basis for distinguishing the complexity of characters’ relationships in different texts. The results show that the structural characteristics of character interaction networks are similar to those of small world networks, and the selected network measurement indexes are significantly related to the complexity of text characters. Finally, we achieved effectively evaluating and predicting the complexity of the social networks from more extensive literature works some classical regression model based on machine learning.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"68 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":"120907917","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/IALP48816.2019.9037687
Joseph Marvin Imperial, C. G. V. Ya-On, Jennifer C. Ureta
In this paper, we present an experimental development of a spell checker for the Tagalog language using a set of word list with 300 random root words and three inflected forms as training data and a two-layered architecture of combined Deterministic Finite Automaton (DFA) with Levenshtein edit-distance. A DFA is used to process strings to identify if it belongs to a certain language via the binary result of accept or reject. The Levenshtein edit-distance of two strings is the number (k) of deletions, alterations, insertions between two sequences of characters. From the sample trained wordlist, results show that a value of 1 for the edit-distance (k) can be effective in spelling Tagalog sentences. Any value greater than 1 can cause suggestion of words even if the spelling of words is correct due to selective and prominent usage of certain characters in the Tagalog language like a, n, g, t, s, l.
{"title":"An experimental Tagalog Finite State Automata spellchecker with Levenshtein edit-distance feature","authors":"Joseph Marvin Imperial, C. G. V. Ya-On, Jennifer C. Ureta","doi":"10.1109/IALP48816.2019.9037687","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037687","url":null,"abstract":"In this paper, we present an experimental development of a spell checker for the Tagalog language using a set of word list with 300 random root words and three inflected forms as training data and a two-layered architecture of combined Deterministic Finite Automaton (DFA) with Levenshtein edit-distance. A DFA is used to process strings to identify if it belongs to a certain language via the binary result of accept or reject. The Levenshtein edit-distance of two strings is the number (k) of deletions, alterations, insertions between two sequences of characters. From the sample trained wordlist, results show that a value of 1 for the edit-distance (k) can be effective in spelling Tagalog sentences. Any value greater than 1 can cause suggestion of words even if the spelling of words is correct due to selective and prominent usage of certain characters in the Tagalog language like a, n, g, t, s, l.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"2 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121004233","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/IALP48816.2019.9037676
Aibao Zhou, Yanbing Hu, Xiaoyong Lu, Tao Pan
Personal pronouns are of great pragmatic significance, and among their numerous functions, referential vagueness is the key to the exploration of the self. In Experiment 1, the relationship of first-, second-, and third personal pronouns with the self was discussed in acoustic condition, with gender consistency, pronoun type, and person type as independent variables. Experiment 2 records the frequency of the subjects on the SR button and the NSR button, and statistical analysis was performed on the SR button reaction. This study finds that the same pronouns show different self-cognitive processing under the different visual and acoustic stimuli, and the results support the dynamics of personal pronouns and possessive pronouns and self relationships.
{"title":"Acoustic Cues for Sensory Modality and Referential Vagueness of Personal Pronouns: Me or Not? You or Not?","authors":"Aibao Zhou, Yanbing Hu, Xiaoyong Lu, Tao Pan","doi":"10.1109/IALP48816.2019.9037676","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037676","url":null,"abstract":"Personal pronouns are of great pragmatic significance, and among their numerous functions, referential vagueness is the key to the exploration of the self. In Experiment 1, the relationship of first-, second-, and third personal pronouns with the self was discussed in acoustic condition, with gender consistency, pronoun type, and person type as independent variables. Experiment 2 records the frequency of the subjects on the SR button and the NSR button, and statistical analysis was performed on the SR button reaction. This study finds that the same pronouns show different self-cognitive processing under the different visual and acoustic stimuli, and the results support the dynamics of personal pronouns and possessive pronouns and self relationships.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"12 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":"122703299","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/IALP48816.2019.9037665
Huiping Wang, B. Wang
As traditional Chinese culture education grows fast, adaptive testing for ancient poetry seems promising. The prerequisite of adaptive learning is question bank, while the quality of question bank depends on the rationality of question difficulty. The paper proposes a method that measures ancient poetry difficulty for objective questions and builds a measurement system. The method includes five steps: (1) Identify the verses corresponding to the question. (2) Get four indexes of the corresponding verses: search volume, correctly answered frequency, sentence length and grade of the textbook which includes the verses.(3) Use analytic hierarchy process to index system for weight assignment. (4) Compute the weighted sum of the four indexes as the measurement of difficulty. (5) Classify the question bank according to the calculated difficulty. Experimental results showed the effectiveness of this measurement method, which thereby can be used in various adaptive testing of ancient poetry.
{"title":"An Measurement Method of Ancient Poetry Difficulty for Adaptive Testing","authors":"Huiping Wang, B. Wang","doi":"10.1109/IALP48816.2019.9037665","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037665","url":null,"abstract":"As traditional Chinese culture education grows fast, adaptive testing for ancient poetry seems promising. The prerequisite of adaptive learning is question bank, while the quality of question bank depends on the rationality of question difficulty. The paper proposes a method that measures ancient poetry difficulty for objective questions and builds a measurement system. The method includes five steps: (1) Identify the verses corresponding to the question. (2) Get four indexes of the corresponding verses: search volume, correctly answered frequency, sentence length and grade of the textbook which includes the verses.(3) Use analytic hierarchy process to index system for weight assignment. (4) Compute the weighted sum of the four indexes as the measurement of difficulty. (5) Classify the question bank according to the calculated difficulty. Experimental results showed the effectiveness of this measurement method, which thereby can be used in various adaptive testing of ancient poetry.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"38 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":"115462910","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/IALP48816.2019.9037679
Xuemei Tang, Shichen Liang, Jianyu Zheng, Renfen Hu, Zhiying Liu
In this paper, we propose an automated method for recognize allusions in Tang poetry. The representation of text is trained by BERT pre-trained by The SiKuQuanShu. The TOP-20 candidate allusions have the highest semantic similarity to the original sentence. Then update the ranking of candidate allusions by rule-based ranking algorithm. In the final experimental results, the precision of the correct allusion same as the final ranking TOP-I reached 63.74%, the precision of the correct allusion appears in the final ranking TOP-3 reached 70.66%, and the precision of the correct allusion appears in the final ranking TOP-5 reached 74.82%.
{"title":"Automatic Recognition of Allusions in Tang Poetry Based on BERT","authors":"Xuemei Tang, Shichen Liang, Jianyu Zheng, Renfen Hu, Zhiying Liu","doi":"10.1109/IALP48816.2019.9037679","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037679","url":null,"abstract":"In this paper, we propose an automated method for recognize allusions in Tang poetry. The representation of text is trained by BERT pre-trained by The SiKuQuanShu. The TOP-20 candidate allusions have the highest semantic similarity to the original sentence. Then update the ranking of candidate allusions by rule-based ranking algorithm. In the final experimental results, the precision of the correct allusion same as the final ranking TOP-I reached 63.74%, the precision of the correct allusion appears in the final ranking TOP-3 reached 70.66%, and the precision of the correct allusion appears in the final ranking TOP-5 reached 74.82%.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"55 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":"131524911","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/IALP48816.2019.9037697
Ju Lin, Zhuanzhuan Ji, Wenwei Dong, Yanlu Xie, Jinsong Zhang
Automatic prosodic boundary detection is useful for automatic speech processing, such as automatic speech recognition (ASR) and speech synthesis. In this paper, we propose two techniques to improve the boundary detection performance. First, in addition to prosody features (e.g, pitch, duration and energy), phonetic information (word/articulatory information) is integrated into the framework of prosodic boundary detection. We compared two forms of phonetic information: word form and articulatory form. Moreover, boundary detection can be regarded as a sequence labeling task. A deep Long Short-Term Memory (LSTM) is adopted for this task, which replaces the traditional Deep Neural Networks (DNN) model. The experimental results showed that the boundary detection performance can be improved by the additional phonetic information, with relative 5.9% (word form) and 9.8% (articulatory form) improvements respectively in contrast with the system that only used prosody features modeled. The articulatory information and prosody features with deep LSTM achieved the best result, with further performance enhancement from 76.35% to 77.85% (relative 6.3%) compared with that modeled by DNN.
{"title":"Improving Mandarin Prosody Boundary Detection by Using Phonetic Information and Deep LSTM Model","authors":"Ju Lin, Zhuanzhuan Ji, Wenwei Dong, Yanlu Xie, Jinsong Zhang","doi":"10.1109/IALP48816.2019.9037697","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037697","url":null,"abstract":"Automatic prosodic boundary detection is useful for automatic speech processing, such as automatic speech recognition (ASR) and speech synthesis. In this paper, we propose two techniques to improve the boundary detection performance. First, in addition to prosody features (e.g, pitch, duration and energy), phonetic information (word/articulatory information) is integrated into the framework of prosodic boundary detection. We compared two forms of phonetic information: word form and articulatory form. Moreover, boundary detection can be regarded as a sequence labeling task. A deep Long Short-Term Memory (LSTM) is adopted for this task, which replaces the traditional Deep Neural Networks (DNN) model. The experimental results showed that the boundary detection performance can be improved by the additional phonetic information, with relative 5.9% (word form) and 9.8% (articulatory form) improvements respectively in contrast with the system that only used prosody features modeled. The articulatory information and prosody features with deep LSTM achieved the best result, with further performance enhancement from 76.35% to 77.85% (relative 6.3%) compared with that modeled by DNN.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"688 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":"132184755","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/IALP48816.2019.9037688
Chia-Yu Li, Ngoc Thang Vu
Code-Switching (CS) is a common linguistic phenomenon in multilingual communities that consists of switching between languages while speaking. This paper presents our investigations on end-to-end speech recognition for Mandarin-English CS speech. We analyze different CS specific issues such as the properties mismatches between languages in a CS language pair, the unpredictable nature of switching points, and the data scarcity problem. We exploit and improve the state-of-the-art end-to-end system by merging nonlinguistic symbols, by integrating language identification using hierarchical softmax, by modeling subword units, by artificially lowering the speaking rate, and by augmenting data using speed perturbed technique and several monolingual datasets to improve the final performance not only on CS speech but also on monolingual benchmarks in order to making the system more applicable on real life settings. Finally, we explore the effect of different language model integration methods on the performance of the proposed model. Our experimental results reveal that all the proposed techniques improve the recognition performance. The best combined system improves the baseline system by up to 35% relatively in terms of mixed error rate and delivers acceptable performance on monolingual benchmarks.
{"title":"Integrating Knowledge in End-to-End Automatic Speech Recognition for Mandarin-English Code-Switching","authors":"Chia-Yu Li, Ngoc Thang Vu","doi":"10.1109/IALP48816.2019.9037688","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037688","url":null,"abstract":"Code-Switching (CS) is a common linguistic phenomenon in multilingual communities that consists of switching between languages while speaking. This paper presents our investigations on end-to-end speech recognition for Mandarin-English CS speech. We analyze different CS specific issues such as the properties mismatches between languages in a CS language pair, the unpredictable nature of switching points, and the data scarcity problem. We exploit and improve the state-of-the-art end-to-end system by merging nonlinguistic symbols, by integrating language identification using hierarchical softmax, by modeling subword units, by artificially lowering the speaking rate, and by augmenting data using speed perturbed technique and several monolingual datasets to improve the final performance not only on CS speech but also on monolingual benchmarks in order to making the system more applicable on real life settings. Finally, we explore the effect of different language model integration methods on the performance of the proposed model. Our experimental results reveal that all the proposed techniques improve the recognition performance. The best combined system improves the baseline system by up to 35% relatively in terms of mixed error rate and delivers acceptable performance on monolingual benchmarks.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"47 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":"123306240","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}
Conversational Artificial Intelligence is revolutionizing the world with its power of converting the conventional computer to a human-like-computer. Exploiting the speaker’s intention is one of the major aspects in the field of conversational Artificial Intelligence. A significant challenge that hinders the effectiveness of identifying the speaker’s intention is the lack of language resources. To address this issue, we present a domain-specific speech command classification system for Sinhala, a low-resourced language. It accomplishes intent detection for the spoken Sinhala language using Automatic Speech Recognition and Natural Language Understanding. The proposed system can be effectively utilized in value-added applications such as Sinhala speech dialog systems. The system consists of an Automatic Speech Recognition engine to convert continuous natural human voice in Sinhala language to its textual representation and a text classifier to accurately understand the user intention. We also present a novel dataset for this task, 4.15 hours of Sinhala speech corpus in the banking domain. Our new Sinhala speech command classification system provides an accuracy of 89.7% in predicting the intent of an utterance. It outperforms the state-of-the-art direct speech-to-intent classification systems developed for the Sinhala language. Moreover, the Automatic Speech Recognition engine shows the Word Error Rate as 12.04% and the Sentence Error Rate as 21.56%. In addition, our experiments provide useful insights on speech-to-intent classification to researchers in low resource spoken language understanding.
{"title":"Speech Command Classification System for Sinhala Language based on Automatic Speech Recognition","authors":"Thilini Dinushika, Lakshika Kavmini, Pamoda Abeyawardhana, Uthayasanker Thayasivam, Sanath Jayasena","doi":"10.1109/IALP48816.2019.9037648","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037648","url":null,"abstract":"Conversational Artificial Intelligence is revolutionizing the world with its power of converting the conventional computer to a human-like-computer. Exploiting the speaker’s intention is one of the major aspects in the field of conversational Artificial Intelligence. A significant challenge that hinders the effectiveness of identifying the speaker’s intention is the lack of language resources. To address this issue, we present a domain-specific speech command classification system for Sinhala, a low-resourced language. It accomplishes intent detection for the spoken Sinhala language using Automatic Speech Recognition and Natural Language Understanding. The proposed system can be effectively utilized in value-added applications such as Sinhala speech dialog systems. The system consists of an Automatic Speech Recognition engine to convert continuous natural human voice in Sinhala language to its textual representation and a text classifier to accurately understand the user intention. We also present a novel dataset for this task, 4.15 hours of Sinhala speech corpus in the banking domain. Our new Sinhala speech command classification system provides an accuracy of 89.7% in predicting the intent of an utterance. It outperforms the state-of-the-art direct speech-to-intent classification systems developed for the Sinhala language. Moreover, the Automatic Speech Recognition engine shows the Word Error Rate as 12.04% and the Sentence Error Rate as 21.56%. In addition, our experiments provide useful insights on speech-to-intent classification to researchers in low resource spoken language understanding.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"45 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":"122103833","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/IALP48816.2019.9037652
Z. Han, Chengguo Lv, Qiansheng Wang, G. Fu
Chinese has been widely used by people from all over the world. Various Chinese spelling errors may occur inevitably from Chinese as Foreign Language (CFL) beginners. This paper presents a method for Chinese spelling check to detect and correct spelling errors in a sentence. Our approach is based on the sequence labeling model using the bi-direction LSTM network. We also utilize the Pinyin feature and Chinese strokes feature to improve performance. The evaluation on the SIGHAN-8 shows that our approach gets good performance on both detection and correction tasks.
{"title":"Chinese Spelling Check based on Sequence Labeling","authors":"Z. Han, Chengguo Lv, Qiansheng Wang, G. Fu","doi":"10.1109/IALP48816.2019.9037652","DOIUrl":"https://doi.org/10.1109/IALP48816.2019.9037652","url":null,"abstract":"Chinese has been widely used by people from all over the world. Various Chinese spelling errors may occur inevitably from Chinese as Foreign Language (CFL) beginners. This paper presents a method for Chinese spelling check to detect and correct spelling errors in a sentence. Our approach is based on the sequence labeling model using the bi-direction LSTM network. We also utilize the Pinyin feature and Chinese strokes feature to improve performance. The evaluation on the SIGHAN-8 shows that our approach gets good performance on both detection and correction tasks.","PeriodicalId":208066,"journal":{"name":"2019 International Conference on Asian Language Processing (IALP)","volume":"15 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":"117108049","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：