| { | |
| "paper_id": "O03-3002", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T08:01:55.417866Z" | |
| }, | |
| "title": "Interleaving Text and Punctuations for Bilingual Sub-sentential Alignment", | |
| "authors": [ | |
| { | |
| "first": "Wen-Chi", | |
| "middle": [], | |
| "last": "Hsie", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Tsing Hua University", | |
| "location": { | |
| "addrLine": "101, Kuangfu Road", | |
| "postCode": "300", | |
| "settlement": "Hsinchu", | |
| "country": "Taiwan, ROC" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Kevin", | |
| "middle": [], | |
| "last": "Yeh", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Tsing Hua University", | |
| "location": { | |
| "addrLine": "101, Kuangfu Road", | |
| "postCode": "300", | |
| "settlement": "Hsinchu", | |
| "country": "Taiwan, ROC" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Jason", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Chang", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "National Tsing Hua University", | |
| "location": { | |
| "addrLine": "101, Kuangfu Road", | |
| "postCode": "300", | |
| "settlement": "Hsinchu", | |
| "country": "Taiwan, ROC" | |
| } | |
| }, | |
| "email": "jschang@cs.nthu.edu.tw" | |
| }, | |
| { | |
| "first": "Thomas", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Chuang", | |
| "suffix": "", | |
| "affiliation": {}, | |
| "email": "tomchuang@cc.vit.edu.tw" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "We present a new approach to aligning bilingual English and Chinese text at sub-sentential level by interleaving alphabetic texts and punctuations matches. With sub-sentential alignment, we expect to improve the effectiveness of alignment at word, chunk and phrase levels and provide finer grained and more reusable translation memory.", | |
| "pdf_parse": { | |
| "paper_id": "O03-3002", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "We present a new approach to aligning bilingual English and Chinese text at sub-sentential level by interleaving alphabetic texts and punctuations matches. With sub-sentential alignment, we expect to improve the effectiveness of alignment at word, chunk and phrase levels and provide finer grained and more reusable translation memory.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Recently, there are renewed interests in using bilingual corpus for building systems for statistical machine translation (Brown et al. 1988 (Brown et al. , 1991 , including data-driven machine translation (Richardson et al. 2002) , computer-assisted revision of translation (Jutras 2000) and cross-language information retrieval (Kwok 2001 ). It is therefore useful for the bilingual corpus to be aligned at the sentence level and even subsentence level with very high precision (Moore 2002; Chuang, You and Chang 2002, Kueng and Su 2002) . Especially, for further analyses such as phrase alignment, word alignment (Ker and Chang 1997; Melamed 2000) and translation memory, high-precision alignment at sub-sentential levels would be very useful. Alignment at sub-sentential level has the potential of improving the effectiveness of alignment at word and phrase levels and providing finer grained and more reusable translation memory.", | |
| "cite_spans": [ | |
| { | |
| "start": 121, | |
| "end": 139, | |
| "text": "(Brown et al. 1988", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 140, | |
| "end": 160, | |
| "text": "(Brown et al. , 1991", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 205, | |
| "end": 229, | |
| "text": "(Richardson et al. 2002)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 274, | |
| "end": 287, | |
| "text": "(Jutras 2000)", | |
| "ref_id": "BIBREF7" | |
| }, | |
| { | |
| "start": 329, | |
| "end": 339, | |
| "text": "(Kwok 2001", | |
| "ref_id": "BIBREF10" | |
| }, | |
| { | |
| "start": 479, | |
| "end": 491, | |
| "text": "(Moore 2002;", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 492, | |
| "end": 507, | |
| "text": "Chuang, You and", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 508, | |
| "end": 529, | |
| "text": "Chang 2002, Kueng and", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 530, | |
| "end": 538, | |
| "text": "Su 2002)", | |
| "ref_id": "BIBREF9" | |
| }, | |
| { | |
| "start": 615, | |
| "end": 635, | |
| "text": "(Ker and Chang 1997;", | |
| "ref_id": "BIBREF8" | |
| }, | |
| { | |
| "start": 636, | |
| "end": 649, | |
| "text": "Melamed 2000)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "Much work has been reported in the literature of computational linguistics on how to align sentences, while very little is touched on alignment just below the sentence level. The most effective approach for sentence alignment is the length-based approach proposed by Brown et al. (1991) and by Gale and Church (1991) . Both methods use normal distribution to model the ratio of lengths between the counterpart sentences measured in number of words or characters. Length-based approach for aligning parallel corpora has commonly been used and produces surprisingly good results for the language pair of French and English at success rates well over 96%. However, it does not perform as well for alignment of text in two distant languages such as Chinese and English. Yeh (2003) proposed a punctuation-based approach for sentence alignment which produces even high accuracy rates than the length based approach. It was pointed out that the ways different languages use punctuations are more or less similar and the correspondence of punctuations across different languages can be obtained using a small set of training data. By soft matching punctuations of the two languages in ordered comparison, the probabilities of mutual translation for a pair of bilingual sentences can be estimated more effectively than lengths. This is not surprising since the average sentence contains many punctuations which carry more information than lengths. Yeh also examined the results of punctuationbased sentence alignment and observed:", | |
| "cite_spans": [ | |
| { | |
| "start": 267, | |
| "end": 286, | |
| "text": "Brown et al. (1991)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 294, | |
| "end": 316, | |
| "text": "Gale and Church (1991)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 745, | |
| "end": 776, | |
| "text": "Chinese and English. Yeh (2003)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "\"Although word alignment links do cross one and other a lot, they general seem not to cross the links between punctuations. It appears that we can obtain sub-sentential alignment at clause and phrase levels from the alignment of punctuation.\"", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "This observation indicates that in bilingual corpus pieces of text delimited by punctuations behave much the same way as sentences with non-crossing alignment links. Therefore, it is reasonable to align pieces of text ending with a couple of punctuations, much the same way as sentence alignment. Building on their work, we develop a new approach to sub-sentential alignment by interleaving the matches of alphabetic texts and punctuations. In the following, we first give an example for bilingual sub-sentential alignment in Section 2. Then we introduce our probability model in Section 3. Next, we describe experimental setup and results in Section 4. We conclude in Section 5 with discussion and future work.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "Consider a pair of counterpart paragraphs in the official records of Hong Kong Legislative Council:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "\"My goal is simply this -to safeguard Hong Kong's way of life. This way of life not only produces impressive material and cultural benefits; it also incorporates values that we all cherish. Our prosperity and stability underpin our way of life. But, equally, Hong Kong's way of life is the foundation on which we must build our future stability and prosperity.\" \u6211\u7684\u76ee\u6a19\u5f88\u7c21\u55ae\uff0c\u5c31\u662f\u8981\u4fdd\u969c\u9999\u6e2f\u7684\u751f\u6d3b\u65b9\u5f0f\u3002\u9019\u500b\u751f\u6d3b\u65b9\u5f0f\uff0c\uf967\u55ae\u5728\u7269\u8cea\u548c\u6587\u5316\u65b9\u9762\u70ba\u6211\u5011\u5e36\uf92d\uf9ba \u91cd\u5927\u7684\uf9dd\u76ca\uff0c\u800c\u4e14\uf901\u878d\u5408\uf9ba\u5927\u5bb6\u90fd\u73cd\u60dc\u7684\u50f9\u503c\u89c0\u3002\u9999\u6e2f\u7684\u5b89\u5b9a\u7e41\u69ae\u662f\u6211\u5011\u751f\u6d3b\u65b9\u5f0f\u7684\u652f\u67f1\u3002\u540c\u6a23\u5730\uff0c \u6211\u5011\u672a\uf92d\u7684\u5b89\u5b9a\u7e41\u69ae\uff0c\u4ea6\u5fc5\u9808\u4ee5\u9999\u6e2f\u7684\u751f\u6d3b\u65b9\u5f0f\u70ba\u57fa\u790e\u3002(Source: Oct. 7, 1992, Governor Christopher Francis Patten's address to the HK LEGCO) By sub-sentential alignment, we mean identifying the shortest possible pair of counterpart texts ending with punctuations. From the example above, the following is the intended results of sub-sentential alignment:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "to safeguard Hong Kong's way of life. \u5c31\u662f\u8981\u4fdd\u969c\u9999\u6e2f\u7684\u751f\u6d3b\u65b9\u5f0f\u3002 This way of life not only produces impressive material and cultural benefits; \u9019\u500b\u751f\u6d3b\u65b9\u5f0f\uff0c\uf967\u55ae\u5728\u7269\u8cea\u548c\u6587\u5316\u65b9\u9762\u70ba\u6211\u5011\u5e36\uf92d\uf9ba\u91cd\u5927\u7684\uf9dd\u76ca\uff0c it also incorporates values that we all cherish. \u800c\u4e14\uf901\u878d\u5408\uf9ba\u5927\u5bb6\u90fd\u73cd\u60dc\u7684\u50f9\u503c\u89c0\u3002 Notice that longer pairs such as the following translation equivalent pair of sentences My goal is simply this -to safeguard Hong Kong's way of life. \u6211\u7684\u76ee\u6a19\u5f88\u7c21\u55ae\uff0c\u5c31\u662f\u8981\u4fdd\u969c\u9999\u6e2f\u7684\u751f\u6d3b\u65b9\u5f0f\u3002 does not fit the bill, since a finer grained subdivision into two 1-1 matches, (My goal is simply this -, \u6211\u7684 \u76ee\u6a19\u5f88\u7c21\u55ae\uff0c) and (to safeguard Hong Kong's way of life., \u5c31\u662f\u8981\u4fdd\u969c\u9999\u6e2f\u7684\u751f\u6d3b\u65b9\u5f0f\u3002) also preserve translation equivalence. Not unlike the situation in sentence alignment, there are many to one, one to many, and many to many matches. For instance, it is not possible to find a 1-1 match for \"This way of life not only produces impressive material and cultural benefits;\" since it only corresponds to \"\u9019\u500b\u751f\u6d3b \u65b9\u5f0f\uff0c\" in part. Therefore, we have to combine the subsequent clause \"\uf967\u55ae\u5728\u7269\u8cea\u548c\u6587\u5316\u65b9\u9762\u70ba\u6211\u5011 \u5e36\uf92d\uf9ba\u91cd\u5927\u7684\uf9dd\u76ca\uff0c\" for a 1-2 match.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "In this section we describe our probability model. To do so, we will first introduce some necessary notation. Let E be an English fragment e 1 , e 2 ,\u2026,e m and C be a Chinese paragraph c 1 , c 2 ,\u2026,c n , which e i and c j is a text-fragment as described in Section 2. We define a link l(e i , c j ) for e i and c j that are translation ( or part of a translation ) of one another. We define null link l(e i , c 0 ) for e i which does not correspond to a translation. The null link l(e 0 , c j ) is defined similarly. An alignment A for two paragraphs E and C is a set of links such that every text-fragment in E and C participates in at least one link, and a text-block linked to e 0 or c 0 participates in no other links.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "We define the alignment problem as finding the alignment A that maximizes P(A|E, C). An alignment A consists of t links {l 1 , l 2 ,\u2026, l t }, where each l k = (e ik , c jk ) for some i k and j k .We will refer to consecutive", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "subsets of A as } ,..., , { 1 j i i j i l l l l + =", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": ", Given this notation, P(A|E, C) can be decomposed as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "\u220f = \u2212 = = t k k k t l C E l P F E l P F E A P 1 1 1 1 ) , , | ( ) , | ( ) , | (", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "For each condition probability, given any pair e i and c j , the link probabilities can be determined directly from combining the probability of length-based model with punctuation-based model. From the paper of Gale and Church in 1993 for length-based model, we know the match probability is Prob( \u03b4 | match ) Prob(match) and Prob( \u03b4 | match ) can be estimated by", | |
| "cite_spans": [ | |
| { | |
| "start": 212, | |
| "end": 235, | |
| "text": "Gale and Church in 1993", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Prob( \u03b4 | match ) = 2 ( 1 -Prob( |\u03b4| ) )", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Where Prob( |\u03b4| ) is the probability that random variable, z, with a standardized ( mean zero, variance one) normal distribution, has magnitude at least as large as |\u03b4|. That is,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Where \u222b \u221e \u2212 \u2212 = \u03b4 \u03c0 \u03b4 dz e z 2 / 2 2 1 ) Prob(", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "We compute \u03b4 directly from the length of two portions of text, l 1 and l 2 , and the two parameters, c and s 2 . (Where c is the expected number of characters in L 2 per character in L 1 , and s 2 is the variance of the number of characters in L 2 per character in L 1 .) That is,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "2 1 1 2 / ) ( s l c l l \u00d7 \u2212 = \u03b4", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": ". Then, Prob( |\u03b4| ) is computed by integrating a standard normal distribution ( with mean zero and variance 1).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Then, from Yeh (2003), for punctuation-based model, we know:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "P pun (e i , c j ) = |) | |, (| ) , ( j i j i pc pe P pc pe P for some l k = (e i , c j )", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "where e i and c i is \u03bb, one, or two punctuations, e i , c j = English and Chinese text-block pe i = the ending English punctuations of e i , i = 1, m pc j = the ending Chinese punctuations c j , j = 1, n, P(pc i , pe i ) = probability of pc i translates into pe i , Thus, for each link l k given E, C and l, we can compute the probability as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "P( l k |E, C, l k-1 ) = P( \u03b4 | match )P(match) * P pun (e i , c j ) , So \u220f = = t k k k k l P m P P F E A P 1 pun ) ( ) ( ) ( ) , | ( \u03b4", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Probability Model", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "In order to assess the performance of our sub-sentential alignment model, we run the system on two test cases:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Experimental results", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "1. Official record of proceedings of Hong Kong Legislative Council at Oct. 7, 1992, 2 . Harry Potter Book I Chapter one.", | |
| "cite_spans": [ | |
| { | |
| "start": 70, | |
| "end": 85, | |
| "text": "Oct. 7, 1992, 2", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Experimental results", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "For probability of punctuation, we use a small set of hand aligned data which led to the following model parameters:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Experimental results", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "1. Punctuation translation probability (Table 1) , 2. Sentence match type probability (Table 2) . Preliminary results shown in Table 3 indicate precision rates of 85% and 93% for the two test cases.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 39, | |
| "end": 48, | |
| "text": "(Table 1)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 86, | |
| "end": 95, | |
| "text": "(Table 2)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 127, | |
| "end": 134, | |
| "text": "Table 3", | |
| "ref_id": "TABREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Experimental results", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "We propose a model interleaving length-based text alignment and punctuation alignment to carry out subsentential alignment. The method seems to work reasonably well with an average precision rate around 90% in the evaluation of a preliminary implementation. There is still a lot of room for improvement. We are currently working on identification of more punctuations useful for sub-sentential alignment, proper segmentation of text ending with punctuations, and better model for lengths of sub-sentential fragment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Discussion and future work", | |
| "sec_num": "5." | |
| }, | |
| { | |
| "text": "We are also looking into the issues of best weighting scheme of length and punctuation information. Finally, the cases where there is inverion of translated fragments are difficult to handle with length information alone. We are also preparing to work with additional lexical information to solve this kind of problem in the future.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Discussion and future work", | |
| "sec_num": "5." | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "We acknowledge the support for this study through grants from Ministry of Education, Taiwan (MOE EX-91-E-FA06-4-4). Thanks are also due to Jim Chang for preparing the training data and evaluating the experimental results.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Acknowledgements", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": " Table A . all incorrect alignments of this experiment. Shaded parts indicate imprecision in alignment results. We calculated the precision rates by dividing the number of unshaded sentences (counting both English and Chinese sentences) by total number of sentences proposed. Since we did not exclude aligned pair using a threshold, the recall rate should be the same as the precision rate.Sub-sentence alignment based on length and punctuation English text Chinese Text Now is the time to show how we mean to prepare for Hong Kong's future under that far-sighted concept, \u73fe\u5728\u4e5f\u662f\u6642\u5019\u8868\u660e\u6211\u5011\u6253\u7b97\u600e\u6a23\u6309\u7167\u300c\u4e00\u570b\uf978\u5236\u300d \u9019\u500b\u6975\u5177\u9060\ufa0a\u7684\u69cb\u601d\uff0c \"one country, two systems\". \u70ba\u9999\u6e2f\u7684\u672a\uf92d\u4f5c\u597d\u6e96\u5099\u3002 -we shall maintain an economy which continues to thrive and prosper, \u2500 \u6211\u5011\uf965\u53ef\uf9a8\u7d93\u6fdf\u6301\u7e8c\u7e41\u69ae\u84ec\u52c3\uff0c\u5275\u9020\u6240\u9700\u8ca1 \u5bcc\uff0c generating the wealth required to provide the standards of public service that people rightly demand; \u4f7f\u63d0\u4f9b\u7684\u516c\u5171\u670d\u52d9\uff0c\u80fd\u9054\u5230\u5e02\u6c11\u8981\u6c42\u7684\u5408\uf9e4\u6c34\u5e73\uff1b Our prescription for prosperity is straightforward. \u6211\u5011\u7de0\u9020\u7e41\u69ae\u7684\u914d\u65b9\u6e05\u695a\u7c21\u55ae\u3002\u6211\u5011\u76f8\u4fe1\uff0cWe believe that businessmen not politicians or officials make the best commercial decisions.\u6700\u4f73\u7684\u5546\u696d\u6c7a\u5b9a\u662f\u7531\u5546\u4eba\uff0c\u800c\uf967\u662f\u7531\u653f\u6cbb\u5bb6\u6216\u653f\u5e9c \u5b98\u54e1\u4f5c\u51fa\u7684\u3002 We believe that government spending must follow not outpace economic growth.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 1, | |
| "end": 8, | |
| "text": "Table A", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Appendix", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We believe in competition within a sound, \u800c\uf967\u61c9\u8d85\u903e\u7d93\u6fdf\u589e\u9577\u3002\u6211\u5011\uf901\u76f8\u4fe1\uff0c fair framework of regulation and law. \u61c9\u5728\u5065\u5168\u800c\u516c\u5e73\u7684\u6cd5\u898f\u4e0b\u9032\ufa08\u7af6\u722d\u3002 I am inviting distinguished members of the business community to join it.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u6211\u5011\u76f8\u4fe1\uff0c\u653f\u5e9c\u958b\u652f\u5fc5\u9808\u8ddf\u96a8\u7d93\u6fdf\u589e\u9577\uff0c", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Their mandate will be to advise me on: \u5c31\u4e0b\u958b\u4e8b\u9805\u5411\u6211\u63d0\u4f9b\u610f\ufa0a\uff1a", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "\u4e26\u6703\u9080\u8acb\u5546\u754c\u5091\u51fa\u4eba\u58eb\u52a0\u5165\u3002\u4ed6\u7684\u8077\u8cac\u662f\uff0c", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Word Association Norms, Mutual Information, and Lexicography", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Church", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Hank", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Computational Linguistics", | |
| "volume": "16", | |
| "issue": "1", | |
| "pages": "22--29", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Church, K and P. Hank, \"Word Association Norms, Mutual Information, and Lexicography,\" Computational Lin- guistics, 16:1, 1990, pp. 22-29.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "First International Conference on Language Resources & Evaluation: Proceedings", | |
| "authors": [ | |
| { | |
| "first": "Chinese", | |
| "middle": [], | |
| "last": "Sproat", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Word Segmentation", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1998, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "417--420", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sproat, Chinese Word Segmentation, First International Conference on Language Resources & Evaluation: Pro- ceedings, 1998, pp. 417-420.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A statistical method for finding word boundaries in Chinese text", | |
| "authors": [ | |
| { | |
| "first": "Richard", | |
| "middle": [], | |
| "last": "Sproat", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chilin", | |
| "middle": [], | |
| "last": "Shih", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Computer Processing of Chinese & Oriental Languages", | |
| "volume": "4", | |
| "issue": "4", | |
| "pages": "336--351", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Richard Sproat, Chilin Shih, 1990, A statistical method for finding word boundaries in Chinese text. Computer Processing of Chinese & Oriental Languages, 4(4): 336-351.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "A stochastic finite-state word-segmentation algorithm for chinese", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Sproat", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chilin", | |
| "middle": [], | |
| "last": "Shih", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "William", | |
| "middle": [], | |
| "last": "Gale", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nancy", | |
| "middle": [], | |
| "last": "Chang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1996, | |
| "venue": "Computational Linguistics", | |
| "volume": "22", | |
| "issue": "3", | |
| "pages": "377--404", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "R. Sproat, Chilin Shih, William Gale, and Nancy Chang. 1996. A stochastic finite-state word-segmentation algo- rithm for chinese. Computational Linguistics, 22(3): 377-404.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Aligning sentences in parallel corpora", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [ | |
| "F" | |
| ], | |
| "last": "Brown", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Lai", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Mercer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1991, | |
| "venue": "29th Annual Meeting of the Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "169--176", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Brown, P. F., J. C. Lai and R. L. Mercer (1991), 'Aligning sentences in parallel corpora', in 29th Annual Meeting of the Associa- tion for Computational Linguistics, Berkeley, CA, USA. pp. 169-176.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Aligning Sentences in Bilingual Corpora Using Lexical Information", | |
| "authors": [ | |
| { | |
| "first": "Stanley", | |
| "middle": [ | |
| "F" | |
| ], | |
| "last": "Chen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "Lecture Notes in Artificial Intelligence", | |
| "volume": "2499", | |
| "issue": "", | |
| "pages": "21--30", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chen, Stanley F. (1993), Aligning Sentences in Bilingual Corpora Using Lexical Information. In Proceedings Chuang, T., G.N. You, J.S. Chang (2002) Adaptive Bilingual Sentence Alignment, Lecture Notes in Artificial Intelligence 2499, 21-30.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "A program for aligning sentences in bilingual corpus", | |
| "authors": [ | |
| { | |
| "first": "William", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Gale", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Kenneth", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Church", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "Computational Linguistics", | |
| "volume": "19", | |
| "issue": "", | |
| "pages": "75--102", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gale, William A. & Kenneth W. Church (1993), A program for aligning sentences in bilingual corpus. In Computational Linguistics, vol. 19, pp. 75-102.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "An Automatic Reviser: The TransCheck System", | |
| "authors": [ | |
| { | |
| "first": "J-M", | |
| "middle": [], | |
| "last": "Jutras", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Proc. of Applied Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "127--134", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Jutras, J-M 2000. An Automatic Reviser: The TransCheck System, In Proc. of Applied Natural Language Processing, 127-134.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "A class-based approach to word alignment", | |
| "authors": [ | |
| { | |
| "first": "Sue", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Ker", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [], | |
| "last": "Jason", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Chang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "Computational Linguistics", | |
| "volume": "23", | |
| "issue": "", | |
| "pages": "313--344", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ker, Sue J. & Jason S. Chang (1997), A class-based approach to word alignment. In Computational Linguistics, 23:2, pp. 313-344.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "A Robust Cross-Domain Bilingual Sentence Alignment Model", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Kueng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Keh-Yih", | |
| "middle": [], | |
| "last": "Su", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the 19th International Conference on Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kueng, T.L. and Keh-Yih Su, 2002. A Robust Cross-Domain Bilingual Sentence Alignment Model, In Proceedings of the 19th International Conference on Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "NTCIR-2 Chinese, Cross-Language Retrieval Experiments Using PIRCS", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Kwok", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "Proceedings of the Second NTCIR Workshop Meeting", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "14--20", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kwok, KL. 2001. NTCIR-2 Chinese, Cross-Language Retrieval Experiments Using PIRCS. In Proceedings of the Second NTCIR Workshop Meeting, pp. (5) 14-20, National Institute of Informatics, Japan.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "A portable algorithm for mapping bitext correspondence", | |
| "authors": [ | |
| { | |
| "first": "I", | |
| "middle": [], | |
| "last": "Melamed", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Dan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1997, | |
| "venue": "The 35th Conference of the Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Melamed, I. Dan (1997), A portable algorithm for mapping bitext correspondence. In The 35th Conference of the Association for Computational Linguistics (ACL 1997), Madrid, Spain.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Sentence and word alignment between Chinese and English", | |
| "authors": [ | |
| { | |
| "first": "Scott", | |
| "middle": [], | |
| "last": "Piao", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Songlin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Piao, Scott Songlin 2000 Sentence and word alignment between Chinese and English. Ph.D. thesis, Lancaster University.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF0": { | |
| "num": null, | |
| "content": "<table><tr><td>English</td><td>Chinese</td><td>Match</td><td/><td/><td>Match Type</td><td>Probability</td></tr><tr><td>Pun.</td><td>Pun.</td><td colspan=\"3\">Type Counts Probability</td><td>1-0</td><td>0.000197</td></tr><tr><td>,</td><td>\uff0c</td><td>1-1</td><td>541</td><td>0.809874</td><td>0-1</td><td>0.000197</td></tr><tr><td>,</td><td>\u3001</td><td>1-1</td><td>56</td><td>0.083832</td><td>1-1</td><td>0.6513</td></tr><tr><td>,</td><td>\u3002</td><td>1-1</td><td>41</td><td>0.061377</td><td>2-2</td><td>0.0066</td></tr><tr><td>,</td><td>\u300c</td><td>1-1</td><td>10</td><td>0.01497</td><td>1-2</td><td>0.0526</td></tr><tr><td>,</td><td>\uff1a</td><td>1-1</td><td>5</td><td>0.007485</td><td>2-1</td><td>0.1776</td></tr><tr><td>,</td><td>\uff1b</td><td>1-1</td><td>4</td><td>0.005988</td><td>Other</td><td>0.0066</td></tr></table>", | |
| "text": "Punctuation Translation probability Match probability of clauses", | |
| "html": null, | |
| "type_str": "table" | |
| }, | |
| "TABREF1": { | |
| "num": null, | |
| "content": "<table><tr><td>Test cases</td><td># of para-graphs</td><td colspan=\"2\"># of matches # of correct matches</td><td>Precision (%)</td></tr><tr><td>Official record of proceedings</td><td/><td/><td/><td/></tr><tr><td>of Hong Kong Legislative</td><td>10</td><td>188</td><td>174</td><td>93</td></tr><tr><td>Council</td><td/><td/><td/><td/></tr><tr><td>Harry Potter Book I Chapter 1</td><td>110</td><td>634</td><td>540</td><td>85</td></tr></table>", | |
| "text": "Performance evaluation for the two test cases", | |
| "html": null, | |
| "type_str": "table" | |
| } | |
| } | |
| } | |
| } |