| { | |
| "paper_id": "O13-1005", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:04:01.518407Z" | |
| }, | |
| "title": "Chinese Spelling Checker Based on Statistical Machine Translation", | |
| "authors": [ | |
| { | |
| "first": "Hsun-Wen", | |
| "middle": [], | |
| "last": "\u90b1\u7d62\u7d0b", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Applications National Tsing Hua University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "\u5433\u9451\u57ce", | |
| "middle": [], | |
| "last": "Chiu", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Applications National Tsing Hua University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Wu", | |
| "middle": [], | |
| "last": "Jian-Cheng", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Applications National Tsing Hua University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Jason", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "\u5f35\u4fca\u76db", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Applications National Tsing Hua University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Chang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Applications National Tsing Hua University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Chinese spell check is an important component for many NLP applications, including word processors, search engines, and automatic essay rating. However, compared to spell checkers for alphabetical languages (e.g., English or French), Chinese spell checkers are more difficult to develop, because there are no word boundaries in Chinese writing system, and errors may be caused by various Chinese input methods. Chinese spell check involves automatically detecting and correcting typos, roughly corresponding to misspelled words in English. Liu et al. (2011) show that people tend to unintentionally generate typos that sound similar (e.g., * \u63aa\u6298 [cuo zhe] and \u632b\u6298 [cuo zhe]), or look alike (e.g., *\u56fa\u96e3 [gu nan] and \u56f0\u96e3 [kun nan]). The methods for spell check can be broadly classified into two types: rule-based methods (", | |
| "pdf_parse": { | |
| "paper_id": "O13-1005", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Chinese spell check is an important component for many NLP applications, including word processors, search engines, and automatic essay rating. However, compared to spell checkers for alphabetical languages (e.g., English or French), Chinese spell checkers are more difficult to develop, because there are no word boundaries in Chinese writing system, and errors may be caused by various Chinese input methods. Chinese spell check involves automatically detecting and correcting typos, roughly corresponding to misspelled words in English. Liu et al. (2011) show that people tend to unintentionally generate typos that sound similar (e.g., * \u63aa\u6298 [cuo zhe] and \u632b\u6298 [cuo zhe]), or look alike (e.g., *\u56fa\u96e3 [gu nan] and \u56f0\u96e3 [kun nan]). The methods for spell check can be broadly classified into two types: rule-based methods (", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "In addition, we use n-grams consist of single-character words to distinguish between correct token sequences and typos: when a sequence of singleton words is not found in the reference list of dictionary entries plus the web-based character ngrams, we regard the ngram as containing a typo. For example, \"\u68ee\u6797 \u7684 \u82b3 \u591a \u7cbe\" [sen lin de fang duo jing]: bigrams such as \"\u7684 \u82b3\", and \"\u82b3 \u591a\" and trigrams such as \"\u7684 \u82b3 \u591a\" and \"\u82b3 \u591a \u7cbe\" are all considered as candidates for typos since those ngrams are not found in the reference list.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The third and final module is the error corrector. we use a SMT model to translate the sentences containing typos into correct ones. Once we generate a list of candidates of typos, we attempt to correct typos, using a statistical machine translation model to translate typos into correct word. The translation probability tp is a probability indicating how likely a typo is translated into a correct word. tp of each correction translation is calculated using the following formula:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "= log !\" ( ( ) ( ) \u2212 ( ) * ) !\"!!\"#$%! ! ! !\" !\"#$% !\" !\"#$%&", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "where freq(trans) is the frequency of translation, freq(candi) is the frequency of the candidate, and \u03b3 is the weight of different error types: visual or phonological. We use a simple, publicly available decoder written in Python which translates monotonically without reordering the Chinese words and phrases using translation probability and language models. To train our model, we used several corpora including Sinica Chinese Balanced Corpus, TWWaC (Taiwan Web as Corpus), a Chinese dictionary (MOE, 1997), and a confusion set (Liu et al., 2011) . The decoder reads a translation model in GIZA++ format, and a language model in SRILM format. We used the official dataset from SIGHAN 7 Bake-off 2013: Chinese Spell Check to evaluate our systems.", | |
| "cite_spans": [ | |
| { | |
| "start": 531, | |
| "end": 549, | |
| "text": "(Liu et al., 2011)", | |
| "ref_id": "BIBREF3" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The results produced by the proposed system were evaluated using precision rate, recall rate and F-score. We evaluated the results of detection as well as correction for many systems with different language resources and settings. The results show that using Web corpus achieves higher precision than dictionary or compiled corpus in detection systems. Using dictionary leads to the highest recall but slightly lower precision. By combining dictionary and Web corpus, we achieve the best precision, recall, and F-score. By restricting the sound confusion to identical sounds and the shape confusion to strongly similar shapes, we can improve precision dramatically. We can further improve the precision and recall, by using different weights in modeling probability of sound and shape based hypotheses which obtain precision rate of .95, recall rate of .56, and F-score of .70 in correction. Because typos are more often related to sound confusion than shape confusion, so giving higher weight to sound confusion indeed leads to further improvement in both precision and recall. In order to test whether we can reduce false alarms further, we tested our systems on a dataset with additional 350 sentences without typos. The best performing system obtain precision rate of .91, recall rate of .56, and F-score of .69 in correction. The results show that this system is very robust, maintaining high precision rate in different situations.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Many avenues exist for future research and improvement of our system. For example, new terms can be automatically discovered and added to the Chinese dictionary to improve both detection and correction performance. Part of speech tagging can be performed to provide more information for error detection. Named entities can be recognized in order to avoid false alarms. Supervised statistical classifier can be used to model translation probability more accurately. Additionally, an interesting direction to explore is using Web ngrams in addition to a Chinese dictionary for correcting typos. Yet another direction of research would be to consider errors related to missing or unnecessary characters.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In summary, we have introduced in this paper, we proposed a novel method for Chinese spell check. Our approach involves error detection and correction based on the phrasal statistical machine translation framework. The error detection module detects errors by segmenting words and checking word and phrase frequency based on a compiled dictionary and Web corpora. The phonological or morphological spelling errors found are then corrected by running a decoder based on statistical machine translation model (SMT). The results show that the proposed system achieves significantly better accuracy in error detecting and more satisfactory performance in error correcting than the state-of-the-art systems. The experiment results show that the method outperforms previous work.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing(ROCLING 2013)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Improve the detection of improperly used Chinese characters with noisy channel model and detection template", | |
| "authors": [ | |
| { | |
| "first": "Yong-Zhi", | |
| "middle": [], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yong-Zhi Chen (2010). Improve the detection of improperly used Chinese characters with noisy channel model and detection template. Master thesis, Chaoyang University of Technology.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Automatic Chinese character error detecting system based on n-gram language model and pragmatics knowledge base", | |
| "authors": [ | |
| { | |
| "first": "Ta-Hung Hung \uff06 Shih-Hung", | |
| "middle": [], | |
| "last": "Wu", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ta-Hung Hung \uff06 Shih-Hung Wu (2009). Automatic Chinese character error detecting system based on n-gram language model and pragmatics knowledge base. Master thesis, Chaoyang University of Technology.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A rule based Chinese spelling and grammar detection system utility", | |
| "authors": [ | |
| { | |
| "first": "Ying", | |
| "middle": [], | |
| "last": "Jiang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Tong", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Tao", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Fangjie", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Wenting", | |
| "middle": [], | |
| "last": "Cheng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xiaofei", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chenghui", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Weijian", | |
| "middle": [], | |
| "last": "Zhang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "International Conference on System Science and Engineering (ICSSE)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "437--440", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ying Jiang, Tong Wang, Tao Lin, Fangjie Wang, Wenting Cheng, Xiaofei Liu, Chenghui Wang, \uff06 Weijian Zhang (2012). A rule based Chinese spelling and grammar detection system utility. 2012 International Conference on System Science and Engineering (ICSSE), pages 437 -440, 30 June -2 July 2012.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Visually and phonologically similar characters in incorrect Chinese words: Analyses, identification, and applications", | |
| "authors": [ | |
| { | |
| "first": "Chao-Lin", | |
| "middle": [], | |
| "last": "Liu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Min-Hua", | |
| "middle": [], | |
| "last": "Lai", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kan-Wen", | |
| "middle": [], | |
| "last": "Tien", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yi-Hsuan", | |
| "middle": [], | |
| "last": "Chuang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shih-Hung", | |
| "middle": [], | |
| "last": "Wu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chia-Ying", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "ACM Trans. Asian Lang, Inform. Process", | |
| "volume": "10", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chao-Lin Liu, Min-Hua Lai, Kan-Wen Tien, Yi-Hsuan Chuang, Shih-Hung Wu, \uff06 Chia-Ying Lee (2011). Visually and phonologically similar characters in incorrect Chinese words: Analyses, identification, and applications. ACM Trans. Asian Lang, Inform. Process. 10, 2, Article 10, pages 39, June 2011.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "MOE Dictionary new edition. Taiwan: Ministry of Education", | |
| "authors": [ | |
| { | |
| "first": "Moe", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "MOE. (2007). MOE Dictionary new edition. Taiwan: Ministry of Education.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "A hybrid approach to automatic Chinese text checking and error correction", | |
| "authors": [ | |
| { | |
| "first": "Hongchi", | |
| "middle": [], | |
| "last": "Fuji Ren", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Qiang", | |
| "middle": [], | |
| "last": "Shi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Zhou", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "IEEE International Conference on Systems, Man, and Cybernetics", | |
| "volume": "3", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Fuji Ren, Hongchi Shi, \uff06 Qiang Zhou (2001). A hybrid approach to automatic Chinese text checking and error correction. 2001 IEEE International Conference on Systems, Man, and Cybernetics, Vol. 3, pages 1693 -1698, 07 -10 Oct. 2001.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": {} | |
| } | |
| } |