| { |
| "paper_id": "O05-1028", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T07:58:52.572754Z" |
| }, |
| "title": "Improving Translation of Unknown Proper Names Using a Hybrid Web-based Translation Extraction Method", |
| "authors": [ |
| { |
| "first": "Min-Shiang", |
| "middle": [], |
| "last": "Shia", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Information Engineering National Cheng Kung University", |
| "location": { |
| "country": "Taiwan, R.O.C" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Jiun-Hung", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Information Engineering National Cheng Kung University", |
| "location": { |
| "country": "Taiwan, R.O.C" |
| } |
| }, |
| "email": "jhlin@csie.ncku.edu.tw" |
| }, |
| { |
| "first": "Scott", |
| "middle": [], |
| "last": "Yu", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Information Engineering National Cheng Kung University", |
| "location": { |
| "country": "Taiwan, R.O.C" |
| } |
| }, |
| "email": "scottyu@csie.ncku.edu.tw" |
| }, |
| { |
| "first": "Wen-Hsiang", |
| "middle": [], |
| "last": "Lu", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Information Engineering National Cheng Kung University", |
| "location": { |
| "country": "Taiwan, R.O.C" |
| } |
| }, |
| "email": "whlu@mail.ncku.edu.tw" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "Recently, we have proposed several effective Web-based term translation extraction methods exploring Web resources to deal with translation of Web query terms. However, many unknown proper names in Web queries are still difficult to be translated by using our previous Web-based term translation extraction methods. Therefore, in this paper we propose a new hybrid translation extraction method, which combines our pervious Web-based term translation extraction method and a new Web-based transliteration method in order to improve translation of unknown proper names. In addition, to efficiently construct a good quality transliteration model, we also present a mixed-syllable-mapping transliteration model and a Web-based semi-supervised learning algorithm to explore search-result pages further for collecting large amounts of English-Chinese transliteration pairs from the Web.", |
| "pdf_parse": { |
| "paper_id": "O05-1028", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "Recently, we have proposed several effective Web-based term translation extraction methods exploring Web resources to deal with translation of Web query terms. However, many unknown proper names in Web queries are still difficult to be translated by using our previous Web-based term translation extraction methods. Therefore, in this paper we propose a new hybrid translation extraction method, which combines our pervious Web-based term translation extraction method and a new Web-based transliteration method in order to improve translation of unknown proper names. In addition, to efficiently construct a good quality transliteration model, we also present a mixed-syllable-mapping transliteration model and a Web-based semi-supervised learning algorithm to explore search-result pages further for collecting large amounts of English-Chinese transliteration pairs from the Web.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "In machine translation (MT) (Brown et al. 1993) or cross-language information retrieval (CLIR) (Jaleel and Larkey 2003; Pirkola et al. 2003) , unknown term translation are still problematic and remain to be solved. Conventionally, most of the existing MT or CLIR systems rely mainly on general-purpose bilingual dictionaries, which usually lack translations of proper names or technical terms, and thus are unable to deal with such problems. We have proposed an effective Web-based approach to exploring abundant language-mixed texts on the Web like anchor texts and search-result pages for alleviating the difficulty of unknown query term translation (Lu et al. 2002 Cheng et al. 2004) . However, the approach employing statistical techniques still suffers from the problem of data sparseness and indirect association errors in finding translations of low-frequency unknown terms (Melamed, 2000) .", |
| "cite_spans": [ |
| { |
| "start": 28, |
| "end": 47, |
| "text": "(Brown et al. 1993)", |
| "ref_id": null |
| }, |
| { |
| "start": 95, |
| "end": 119, |
| "text": "(Jaleel and Larkey 2003;", |
| "ref_id": "BIBREF0" |
| }, |
| { |
| "start": 120, |
| "end": 140, |
| "text": "Pirkola et al. 2003)", |
| "ref_id": "BIBREF24" |
| }, |
| { |
| "start": 652, |
| "end": 667, |
| "text": "(Lu et al. 2002", |
| "ref_id": "BIBREF19" |
| }, |
| { |
| "start": 668, |
| "end": 686, |
| "text": "Cheng et al. 2004)", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 881, |
| "end": 896, |
| "text": "(Melamed, 2000)", |
| "ref_id": "BIBREF22" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "According to the report in previous research (Davis et al. 1998) , around 50% of unknown terms are proper names. To improve translation of unknown proper names, in this paper, we propose a hybrid translation extraction method, which is composed of our pervious search-result-based term translation extraction method (Section 3.2) and a new Web-based transliteration method (Section 3.3).", |
| "cite_spans": [ |
| { |
| "start": 45, |
| "end": 64, |
| "text": "(Davis et al. 1998)", |
| "ref_id": "BIBREF7" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Transliteration is the process that converting a sequence of substrings or characters in the source language (e.g., English) into a pronunciation-approximate substring/character sequence in the target language (e.g., Chinese). Many researchers have proposed phoneme-based mapping techniques for proper name transliteration (Jung et al. 2000; Knight & Graehl 1998; Lin & Chen 2002; Meng et al. 2001; Virga & Khudanpur 2003) , but converting an English word from phonemic representation to Chinese Pinyin and from Pinyin to Chinese characters may cause double errors. Taking this problem into consideration, we thus try to adopt direct orthographical mapping for proper name transliteration and propose a simple mixed-syllable-mapping transliteration model which can effectively increase the correct mapping between an English-Chinese transliteration pair with different number of transliteration unit (syllable), such as \" Ericsson\"(\u6613\u5229\u4fe1) with four English transliteration units \" e\" , \" ri\" , \" c\" , \" sson\"and three Chinese transliteration units \" \u6613 \" , \" \u5229 \" , \" \u4fe1 \"(Section 3.3).", |
| "cite_spans": [ |
| { |
| "start": 323, |
| "end": 341, |
| "text": "(Jung et al. 2000;", |
| "ref_id": "BIBREF12" |
| }, |
| { |
| "start": 342, |
| "end": 363, |
| "text": "Knight & Graehl 1998;", |
| "ref_id": null |
| }, |
| { |
| "start": 364, |
| "end": 380, |
| "text": "Lin & Chen 2002;", |
| "ref_id": "BIBREF18" |
| }, |
| { |
| "start": 381, |
| "end": 398, |
| "text": "Meng et al. 2001;", |
| "ref_id": "BIBREF23" |
| }, |
| { |
| "start": 399, |
| "end": 422, |
| "text": "Virga & Khudanpur 2003)", |
| "ref_id": "BIBREF29" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Additionally, to train a good quality transliteration model which is used to filter out impossible transliteration candidates in the process of extracting translation of unknown proper names, we also present a Web-based semi-supervised learning algorithm to collect large amounts of English-Chinese transliteration pairs from the Web. Experimental results show that our new approach can make improvements for translation of unknown proper names.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Term translation extraction is a significant research topic in the field of machine translation. A number of related researches (Gale and Church 1991; Kupiec 1993; Melamed 2000; Smadja et al. 1996) have used sentence-aligned parallel corpora to extract translations since the advent of statistical translation model (Brown et al. 1990 (Brown et al. , 1993 . For example, Melamed (2000) proposed statistical translation models to improve the techniques of word alignment by taking advantage of pre-existing knowledge and overcome the problems of indirect association errors, i.e., erroneous translational correspondence arose from highly co-occurred relevant terms. Although high accuracy of translation extraction can be easily achieved by these techniques, sufficiently large parallel corpora for various subject domains and language pairs currently are not always available.", |
| "cite_spans": [ |
| { |
| "start": 128, |
| "end": 150, |
| "text": "(Gale and Church 1991;", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 151, |
| "end": 163, |
| "text": "Kupiec 1993;", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 164, |
| "end": 177, |
| "text": "Melamed 2000;", |
| "ref_id": "BIBREF22" |
| }, |
| { |
| "start": 178, |
| "end": 197, |
| "text": "Smadja et al. 1996)", |
| "ref_id": "BIBREF28" |
| }, |
| { |
| "start": 316, |
| "end": 334, |
| "text": "(Brown et al. 1990", |
| "ref_id": null |
| }, |
| { |
| "start": 335, |
| "end": 355, |
| "text": "(Brown et al. , 1993", |
| "ref_id": null |
| }, |
| { |
| "start": 371, |
| "end": 385, |
| "text": "Melamed (2000)", |
| "ref_id": "BIBREF22" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Parallel-Corpus-based Term Translation Extraction", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "However, less attention has been devoted to automatic extraction of term translations from comparable or even unrelated texts, since such methods encountered more difficulties due to lacking parallel correlation aligned between documents or sentence pairs. Rapp (1999) proposed an approach to utilizing non-parallel corpora based on the assumption that the contexts of a term should be similar to the contexts of its translation in any language pairs. Fung et al. (1998) also proposed a similar approach that uses vector-space model and takes a bilingual lexicon (called seed words) as feature set to estimate the similarity between a word and its translation candidates. These works are important for automatic extraction of new terminology and unknown proper names in diverse domains. It is a pity that comparable corpora are easier to obtain, however, how to achieve better performance for higher translation coverage is still a challenging task.", |
| "cite_spans": [ |
| { |
| "start": 257, |
| "end": 268, |
| "text": "Rapp (1999)", |
| "ref_id": "BIBREF26" |
| }, |
| { |
| "start": 452, |
| "end": 470, |
| "text": "Fung et al. (1998)", |
| "ref_id": "BIBREF8" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Comparable-Corpus-based Term Translation Extraction", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "The Web is becoming the largest data repository in the world, which consists of huge amounts of multilingual and wide-scoped hypertext resources. A number of studies have been concentrated in the use of the Web to complement insufficient corpora (Cao & Li 2002; Kilgarriff et al. 2003) . How to utilize the Web resources to benefit translations of unknown terms is worthy to investigate.", |
| "cite_spans": [ |
| { |
| "start": 246, |
| "end": 261, |
| "text": "(Cao & Li 2002;", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 262, |
| "end": 285, |
| "text": "Kilgarriff et al. 2003)", |
| "ref_id": "BIBREF13" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Web-based Term Translation Extraction", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "As mentioned above, the conventional term translation methods suffer from the problems of the lack of large-size parallel corpora and the shortage of translation coverage of comparable corpora in medical domain. Thus, we have proposed several Web-based methods to effectively deal with translation of frequent Web query terms by exploring Web anchor text and search-result pages. Although the anchor-text-based approach has been proven effective in extracting multilingual translations (Lu et al. 2002 , it requires crawling the Web to gather sufficient training data as well as more network bandwidth and storage. For the reason to reduce such costs, this paper only adopts the search-result-based approach to extract translation candidates for term translation (describes in Section 3.2). However, many proper names are still difficult to be translated correctly using the search-result-based approach. Therefore, in this paper we intend to further explore search results to collect English-Chinese transliteration pairs, and build a good quality transliteration model which can be used to filtered out impossible translation candidates to improve translation of unknown proper names.", |
| "cite_spans": [ |
| { |
| "start": 486, |
| "end": 501, |
| "text": "(Lu et al. 2002", |
| "ref_id": "BIBREF19" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Web-based Term Translation Extraction", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "For name transliteration between Latin-alphabet languages and some Asian languages with different writing forms, such as English and Chinese, researchers have proposed phoneme-based mapping techniques (Jung et al. 2000; Knight & Graehl 1998; Lin & Chen 2002; Meng et al. 2001; Virga & Khudanpur 2003) . Knight and Graehl used an English-katakana dictionary, katakana-English phoneme mapping, and the CMU Speech Pronunciation Dictionary to deal with transliteration between English words and Katakana sequences. Lin et al. (2003) proposed a statistical transliteration model and apply the model to extract proper names and their transliterations in a parallel corpus with high average precision and recall rates. However, Li et al. (2004) have pointed out that the transliteration precision of the phoneme-based approaches could be limited by two main constraints. First, Latin-alphabet foreign names from different origins have different phonic rules (Pirkola et al. 2003) , such as French and English. Second, converting English words to Chinese characters will need two steps: converting from phonemic representation to Chinese Pinyin and from Pinyin to Chinese characters. Two cascaded converting steps may cause double errors. Taking this problem into consideration, we try to adopt direct orthographical mapping for name transliteration (described in Section 3.3).", |
| "cite_spans": [ |
| { |
| "start": 201, |
| "end": 219, |
| "text": "(Jung et al. 2000;", |
| "ref_id": "BIBREF12" |
| }, |
| { |
| "start": 220, |
| "end": 241, |
| "text": "Knight & Graehl 1998;", |
| "ref_id": null |
| }, |
| { |
| "start": 242, |
| "end": 258, |
| "text": "Lin & Chen 2002;", |
| "ref_id": "BIBREF18" |
| }, |
| { |
| "start": 259, |
| "end": 276, |
| "text": "Meng et al. 2001;", |
| "ref_id": "BIBREF23" |
| }, |
| { |
| "start": 277, |
| "end": 300, |
| "text": "Virga & Khudanpur 2003)", |
| "ref_id": "BIBREF29" |
| }, |
| { |
| "start": 511, |
| "end": 528, |
| "text": "Lin et al. (2003)", |
| "ref_id": "BIBREF17" |
| }, |
| { |
| "start": 721, |
| "end": 737, |
| "text": "Li et al. (2004)", |
| "ref_id": "BIBREF16" |
| }, |
| { |
| "start": 951, |
| "end": 972, |
| "text": "(Pirkola et al. 2003)", |
| "ref_id": "BIBREF24" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Proper Name Transliteration", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "Actually, search-result page is a good resource for extracting translation of frequent unknown query terms. However, a number of unknown proper names are still not extracted correctly due to the problems of data sparseness. Thus, our idea is to integrate name transliteration techniques into the process of extracting translation of proper names in order to filter impossible transliterated candidates for improving the performance of translation extraction. To deal with the problem, first we need to extract terms from the search-result pages as translation candidates, and then filter out impossible candidates based on the name transliteration model. In fact, it is challenging to build a good quality transliteration model while lacking sufficient transliteration pairs for training. We therefore propose a Web-based semi-supervised learning algorithm to collect large amounts of English-Chinese transliteration pairs from the Web (see Section 3.3).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Problem and Challenge", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "We have proposed an effective search-result-based method to explore language-mixed search-result pages and utilize co-occurrence relation and context information for extracting unknown query term translation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "In this section, we will simply describe candidate selection methods using the search-result-based method. For more details, please refer to our previous work (Cheng et al. 2004) .", |
| "cite_spans": [ |
| { |
| "start": 159, |
| "end": 178, |
| "text": "(Cheng et al. 2004)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(1) Chi-Square Test Method: On the basis of co-occurrence analysis, chi-s q u a r e t e s t ( \u03c7 2 ) is adopted to estimate semantic similarity between the source term E and the target candidate C. The similarity measure is defined as", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(1) , ) ( ) ( ) ( ) ( ) ( ) C , ( 2 2 d c d b c a b a c b d a N E S \uf02b \uf0b4 \uf02b \uf0b4 \uf02b \uf0b4 \uf02b \uf0b4 \uf02d \uf0b4 \uf0b4 \uf03d \uf063", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "where a, b, c and d are the numbers of pages retrieving from search engines by submitting Boolean queries: \" E and C\" , \" E and not C\" , \" not E and C\" , a n d \" not E and not C\" , r e s p e c t i v e l y ; N is the total number of pages, i.e., N = a + b + c + d.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(2) Context-Vector Analysis Method: Due to the nature of Chinese-English mixed texts often appearing in Chinese pages, the source term E and the target candidate C may share common contextual terms in the search-result pages (Fung & Yee 1998; Rapp 1999) . The similarity between E and C will be computed based on their context feature vectors in the vector-space model. The conventional tf-idf weighting scheme is used and defined as", |
| "cite_spans": [ |
| { |
| "start": 225, |
| "end": 242, |
| "text": "(Fung & Yee 1998;", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 243, |
| "end": 253, |
| "text": "Rapp 1999)", |
| "ref_id": "BIBREF26" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(2) ), log( ) , ( max ) , ( n N p t f p t f w j j i t i \uf0b4 \uf03d", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "where f(t i , p) is the frequency of term t i in search-result page p, N is the total number of Web pages, and n is the number of the pages containing t i . Finally, we use the cosine measure to estimate the similarity as follows:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(3) . ) ( ) ( ) C , ( 1 2 1 2 1 \uf0e5 \uf0e5 \uf0e5 \uf03d \uf03d \uf03d \uf0b4 \uf0b4 \uf03d m i c m i e c m i e CV i i i i w w w w E S", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Method", |
| "sec_num": null |
| }, |
| { |
| "text": "(1) English Letter Substring Segmentation: Wan and Verspoor (1998) \uf0b7 Separate two consecutive vowels except the following cases: ai, au, ee, ea, ie, oa, oo, ou, etc.", |
| "cite_spans": [ |
| { |
| "start": 43, |
| "end": 66, |
| "text": "Wan and Verspoor (1998)", |
| "ref_id": "BIBREF30" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\uf0b7 Separate two consecutive consonants except the following cases: bh, ch, gh, ph, th, wh, ck, cz, zh, zk, ng, sc,ll, tt, etc.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\uf0b7 l, m, n, r are combined with the left vowel only if they are not followed by a vowel.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\uf0b7 A consonant and a following vowel are regarded as a TU.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "For example, \" a ma y a \" ( \u963f\u99ac\u96c5) is segmented into t h r e e s u b s t r i n g s \" a\" , \" ma\" , \" y a\" , and \" mobley\" (\u83ab\u5e03\u91cc) i s s e g me n t e d i n t o t h r e e s u b s t r i n g s \" mo \" , \" b \" , \" l e y \" . Currently, some English names may be segmented incorrectly, but it is easy to manually add new rules for improving English letter substring segmentation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "(2) Mixed-Syllable-Mapping Transliteration Model: To avoid double errors of converting English phonemic representation to Chinese Pinyin and from Pinyin to Chinese characters, we thus adopted direct orthographical mapping to deal with the alignment between any English name, E = e 1 e 2 \u2026e m , and its Chinese transliterated name, C = c 1 c 2 \u2026c n . Each English TU e i is mapped to a Chinese character c i with the probability P(c i | e i ). Initially, to efficiently train a Web-based transliteration model based on the collected transliteration pairs from the Web for filtering out impossible transliteration candidates, we adopt a simple name transliteration model called forward-syllable-mapping transliteration model, which computes the forward syllable mapping score between E and C using the following formula:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "(4) ], ) | ( ) 1 [( ) | ( ) , ( ) , min( 1 \uf061 \uf061 \uf02b \uf02d \uf0bb \uf03d \uf0d5 \uf03d n m i i i FSM e c P E C P C E S", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "where \uf061is the smoothing weight.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "For an English-Chinese transliteration pair with different number of transliteration unit, such as \" Rusedski\"(\u9b6f\u585e\u65af\u57fa) with the five English segmented substrings \" ru\" , \" se\" , \" d\" , \" s\" , \" ki\"and four Chinese characters \" \u9b6f\" , \" \u585e\" , \" \u65af\" , \" \u57fa\" , to increase the correct mapping between English TUs and Chinese characters, we propose an alternative transliteration mapping model called reverse-syllable-mapping transliteration model, which is used to compute the reverse syllable mapping score as follows: ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\uf028 \uf029 (5) . ], ) | ( ) 1 [( ; ], ) | ( ) 1 [( , 1 ) ( 1 ) ( \uf0ef \uf0ef \uf0ee \uf0ef \uf0ef \uf0ed \uf0ec \uf03c \uf02b \uf02d \uf0b3 \uf02b \uf02d \uf0bb \uf0d5 \uf0d5 \uf02b \uf02d \uf03d \uf02d \uf02d \uf02b \uf02d \uf03d \uf02d \uf02d m", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Filtering Translation Candidates Using a Web-Based Name Transliteration Method", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "Currently, for some unknown proper names, it is still difficult to effectively extract translation by using our previous search-result-based translation extraction method. Therefore, we try to combine a new Web-based transliteration method to enhance our previous search-result-based translation extraction method.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Proposed Approach to Translation Extraction", |
| "sec_num": "3.4" |
| }, |
| { |
| "text": "Chinese seed character set C s and a general-purpose bilingual dictionary D Output: English-Chinese transliteration pair set V ec , and a transliteration model T 1. Extraction of Chinese transliterated names: 1.1. Seed character selection: select two frequent characters from the Chinese seed character set C s . 1.2. Search-result crawling: send the two selected characters to a search engine and get search-result pages. 1.3. Chinese transliterated name identification: use CKIP tagger to find unknown terms in the search-result pages, and then take the unknown terms containing the two Chinese seed characters as potential Chinese transliterated names and add them into V c . (1) Linear Combination Method: Intuitively, a simple method is to directly combine the above three different methods: the chi-square test method, the context-vector analysis method, and the Web-based transliteration method. Under consideration of the large difference of ranges of similarity values among the above methods, we would use a linear combination of inverse ranks to compute the similarity measure as follows:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "(7) , ) , ( ) , ( \uf0e5 \uf03d m m m Combined C E R C E S \uf061", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "where \uf061 m is an assigned weight for each similarity measure S m , and R m (E, C) represents the similarity rank of each target candidate C with respect to its source term E and is assigned to be from 1 to k (candidate number) according to similarity measure S m (E, C) in decreasing order.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "Note that this liner combination method is only used as baseline in comparison with our proposed hybrid translation extraction method described below in the following experiments (Section 4.2).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "(2) Hybrid Method: For some unknown proper names, the simple linear combination method might not make good improvements while these respective methods can' t obtain high ranks for possibly correct transliteration candidates. Therefore, we propose a new hybrid translation extraction method in order to obtain better performance. First, we use the search-result-based translation extraction method described above to extract k (k = 20) terms with high similarity score as transliteration candidates.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "Second, some impossible candidates included in general-purpose bilingual dictionaries are filtered out, and then each of the rest transliterated candidates is ranked according to transliteration mapping score with the test proper name which is computed based on the Web-based transliteration model (Equation (4) and (6)).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Input:", |
| "sec_num": null |
| }, |
| { |
| "text": "We conducted the following experiments to examine the performance of the proposed hybrid translation extraction method and the comparison with the simple linear combination method. Particularly, the focus of the experiments is mainly emphasized on the effectiveness of translations of unknown proper names using the proposed mixed-syllable-mapping transliteration model and hybrid translation extraction method.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results and Analysis", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Collected data: Initially, our proposed Web-based semi-supervised learning algorithm is employed to efficiently collect about 11,000 English-Chinese transliteration pairs for training a transliteration model.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results and Analysis", |
| "sec_num": "4" |
| }, |
| { |
| "text": "We constructed one test set of unknown English query terms, NTCIR proper name set, which contains 22 unknown transliteration names from a total of 100 NTCIR2 and NTCIR3 title queries that contain 175 and 183 unique query terms respectively ).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Test set:", |
| "sec_num": null |
| }, |
| { |
| "text": "Evaluation Metric: The average top-n inclusion rate was adopted as a metric on the extraction of translation equivalents. For a set of terms to be translated, its top-n inclusion rate was defined as the percentage of the terms whose translations could be found in the first n extracted translations (Cheng et al. 2004) .", |
| "cite_spans": [ |
| { |
| "start": 299, |
| "end": 318, |
| "text": "(Cheng et al. 2004)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Test set:", |
| "sec_num": null |
| }, |
| { |
| "text": "To test the effectiveness of the mixed-syllable-mapping transliteration model, we carried out a comparative experiment with different ranking. The results are shown in Table 1 . Actually, the mixed-syllable-mapping transliteration model is effective to improve the top-n inclusion rate. For translation extraction of the NTCIR proper names, the mixed-syllable-mapping transliteration model can achieve 27% and 45% top-1 inclusion rates for the name transliteration method and the hybrid translation method, respectively. Obviously, the reason is that for many English-Chinese transliteration pairs with different number of TU, reverse-syllable-mapping transliteration model can aid in learning correct mapping between English substrings and Chinese characters. Additionally, the model has the same assist effect to many partially matching transliteration pairs collected by using our proposed", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 168, |
| "end": 175, |
| "text": "Table 1", |
| "ref_id": "TABREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Mixed-Syllable-Mapping Transliteration Model vs. Forward-Syllable-Mapping Transliteration Model", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Web-based transliteration method. Fo r t h e g i v e n p r o p e r n a me \" Michael\" ( \u9ea5\u514b) shown in Table 3 , the better rank of its correct translation can be obtained by using the mixed-syllable-mapping transliteration model.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 100, |
| "end": 107, |
| "text": "Table 3", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Mixed-Syllable-Mapping Transliteration Model vs. Forward-Syllable-Mapping Transliteration Model", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "To determine the effectiveness of the proposed hybrid translation extraction method compared with other methods, we also did several comparative experiments with different ranking. The results are also shown in Table 2 . For the NTCIR test set, surprisingly, the hybrid translation extraction method made a great improvement compared with the search-result-based translation extraction method, name transliteration method, or linear combination method. The hybrid translation extraction method with mixed-syllable-mapping transliteration model can achieve 45% top-1 inclusion rate. The main reason is that most of the incorrect translation candidates extracted by using the search-result-based translation extraction method can be filtered out by using the Web-based transliteration method. For example,", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 211, |
| "end": 218, |
| "text": "Table 2", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Hybrid Translation Extraction Method vs. Linear Combination Method", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "given the p r o p e r n a me \" Ko s o v a r \" ( s e e T a b l e 3), the corr e c t C h i n e s e t r a n s l i t e r a t e d n a me \" \u79d1\u7d22\u6c83\"can be ranked to the top one from the fourth rank using only the search-result-based translation extraction method. However, the simple linear combination method seems not effective to improve translation performance since the name transliteration method is still limited in generating correct transliterated candidates even though it can generate many pronunciation-proximate candidates.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Hybrid Translation Extraction Method vs. Linear Combination Method", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Our proposed mixed-syllable-mapping model and hybrid translation extraction method is effective to improve performance in extracting translation of unknown proper names. However, the hybrid translation extraction method sometimes performs not good as linear combination method. An example s u c h a s \" Viagra\" ( \u5a01\u800c\u525b) is shown in Table 4 . Currently, our Web-based semi-supervised learning algorithm is limited by insufficient transliteration training from our collected transliteration pairs which are still in the need of examining by large amounts of manual labor. In the future, we will develop an unsupervised learning algorithm to automatically collect much more amounts of English-Chinese Table 4 . Ineffective results of translation extraction using the hybrid translation extraction method (underlined terms indicate correct translation). are still a number of cases that are still difficult to deal with by using the simple mixed-syllable-mapping transliteration model and need to be further investigated in the future.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 330, |
| "end": 337, |
| "text": "Table 4", |
| "ref_id": null |
| }, |
| { |
| "start": 696, |
| "end": 703, |
| "text": "Table 4", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Discussions", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "We have presented a new hybrid translation extraction method that works well for improving extraction of translation of known proper names by effectively combining a previous search-result-based translation extraction method and our proposed Web-based name transliteration method. Additionally, our proposed simple mixed-syllable-mapping transliteration model and Web-based semi-supervised learning algorithm are also effective to collect English-Chinese transliteration pairs and then train a transliteration model for filtering out incorrect transliteration candidates in the process of extracting proper name translation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "5" |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Statistical transliteration for English-Arabic cross language information retrieval", |
| "authors": [ |
| { |
| "first": "N", |
| "middle": [ |
| "A" |
| ], |
| "last": "Jaleel", |
| "suffix": "" |
| }, |
| { |
| "first": "L", |
| "middle": [ |
| "S" |
| ], |
| "last": "Larkey", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "CIKM", |
| "volume": "", |
| "issue": "", |
| "pages": "139--146", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "N. A. Jaleel and L. S. Larkey. 2003. Statistical transliteration for English-Arabic cross language information retrieval. CIKM 2003: 139-146.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "A statistical approach to machine translation", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Roossin", |
| "suffix": "" |
| } |
| ], |
| "year": 1990, |
| "venue": "Computational Linguistics", |
| "volume": "16", |
| "issue": "2", |
| "pages": "79--85", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Roossin. 1990. A statistical approach to machine translation. Computational Linguistics, 16(2):79-85.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Base noun phrase translation using Web data and the EM algorithm", |
| "authors": [ |
| { |
| "first": "Y.-B", |
| "middle": [], |
| "last": "Cao", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Li", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "Proc. of COLING", |
| "volume": "", |
| "issue": "", |
| "pages": "127--133", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Y.-B. Cao and H. Li. 2002. Base noun phrase translation using Web data and the EM algorithm. In Proc. of COLING 2002: 127-133.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "The Chinese Text Retrieval Tasks of NTCIR Workshop 2", |
| "authors": [ |
| { |
| "first": "K.-H", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| }, |
| { |
| "first": "H.-H", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| } |
| ], |
| "year": 2001, |
| "venue": "Proc. of the Second NTCIR Workshop Meeting on Evaluation of Chinese & Japanese Text Retrieval and Text Summarization", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "K.-H. Chen and H.-H. Chen. 2001. The Chinese Text Retrieval Tasks of NTCIR Workshop 2. In Proc. of the Second NTCIR Workshop Meeting on Evaluation of Chinese & Japanese Text Retrieval and Text Summarization.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Creating Multilingual Translation Lexicons with Regional Variations Using Web Corpora", |
| "authors": [ |
| { |
| "first": "P.-J", |
| "middle": [], |
| "last": "Cheng", |
| "suffix": "" |
| }, |
| { |
| "first": "Y.-C", |
| "middle": [], |
| "last": "Pan", |
| "suffix": "" |
| }, |
| { |
| "first": "W.-H", |
| "middle": [], |
| "last": "Lu", |
| "suffix": "" |
| }, |
| { |
| "first": "L.-F", |
| "middle": [], |
| "last": "Chien", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "535--542", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "P.-J. Cheng, Y.-C. Pan, W.-H. Lu, L.-F. Chien. 2004. Creating Multilingual Translation Lexicons with Regional Variations Using Web Corpora. In Proc. of ACL 2004: 535-542.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Free Resources and Advanced Alignment for Cross-Language Text Retrieval", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [ |
| "W" |
| ], |
| "last": "Davis", |
| "suffix": "" |
| }, |
| { |
| "first": "W", |
| "middle": [ |
| "C" |
| ], |
| "last": "Ogden", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "Proc. of the Sixth Text Retrieval Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "385--394", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "M. W. Davis and W. C. Ogden. 1998. Free Resources and Advanced Alignment for Cross-Language Text Retrieval. In Proc. of the Sixth Text Retrieval Conference (TREC6): 385-394.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "An IR approach for translating new words from nonparallel, comparable texts", |
| "authors": [ |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Fung", |
| "suffix": "" |
| }, |
| { |
| "first": "L.-Y", |
| "middle": [], |
| "last": "Yee", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "414--420", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "P. Fung and L.-Y. Yee. 1998. An IR approach for translating new words from nonparallel, comparable texts. In Proc. of ACL 1998: 414-420.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Identifying Word Correspondances in Parallel Texts", |
| "authors": [ |
| { |
| "first": "W", |
| "middle": [ |
| "A" |
| ], |
| "last": "Gale", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [ |
| "W" |
| ], |
| "last": "Church", |
| "suffix": "" |
| } |
| ], |
| "year": 1991, |
| "venue": "Proc. of DARPA Speech and Natural Language Workshop", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W. A. Gale and K. W. Church. 1991. Identifying Word Correspondances in Parallel Texts, In Proc. of DARPA Speech and Natural Language Workshop.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Phoneme-based Transliteration of Foreign Names for OOV Problem", |
| "authors": [ |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Gao", |
| "suffix": "" |
| }, |
| { |
| "first": "K.-F", |
| "middle": [], |
| "last": "Wong", |
| "suffix": "" |
| }, |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Lam", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Proc. of IJCNLP", |
| "volume": "", |
| "issue": "", |
| "pages": "274--381", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W. Gao, K.-F. Wong and W. Lam. 2004. Phoneme-based Transliteration of Foreign Names for OOV Problem. In Proc. of IJCNLP 2004: 274-381.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Lexicon-based orthographic disambiguation in CJK intelligent information retrieval", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Halpern", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "Proc. of Workshop on Asian Language Resources and International Standardization", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "J. Halpern. 2000. Lexicon-based orthographic disambiguation in CJK intelligent information retrieval. In Proc. of Workshop on Asian Language Resources and International Standardization.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "An English to Korean Transliteration Model of Extended Markov Window", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [ |
| "Y" |
| ], |
| "last": "Jung", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [ |
| "L" |
| ], |
| "last": "Hong", |
| "suffix": "" |
| }, |
| { |
| "first": "E", |
| "middle": [], |
| "last": "Paek", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "Proc. of COLING", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "S. Y. Jung, S. L. Hong and E. Paek. 2000. An English to Korean Transliteration Model of Extended Markov Window. In Proc. of COLING 2000.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Introduction to the special issue on the web as corpus", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Kilgarriff", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Grefenstette", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Computational Linguistics", |
| "volume": "29", |
| "issue": "3", |
| "pages": "333--348", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. Kilgarriff and G. Grefenstette. 2003. Introduction to the special issue on the web as corpus. Computational Linguistics 29(3): 333-348.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "An algorithm for finding noun phrase correspondences in bilingual corpora", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [ |
| "M" |
| ], |
| "last": "Kupiec", |
| "suffix": "" |
| } |
| ], |
| "year": 1993, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "17--22", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "J. M. Kupiec. 1993. An algorithm for finding noun phrase correspondences in bilingual corpora. In Proc. of ACL 1993: 17-22.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "A Joint Source-Channel Model for Machine Transliteration", |
| "authors": [ |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Li", |
| "suffix": "" |
| }, |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Zhang", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Su", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "160--167", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "H. Li, M. Zhang and J. Su. 2004. A Joint Source-Channel Model for Machine Transliteration. In Proc. of ACL 2004: 160-167.", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Word-Transliteration Alignment", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| }, |
| { |
| "first": "C.-C", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| }, |
| { |
| "first": "J.-S", |
| "middle": [], |
| "last": "Chang", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Proc. of ROCLING XV", |
| "volume": "", |
| "issue": "", |
| "pages": "1--16", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "T. Lin, C.-C. Wu, J.-S. Chang. 2003. Word-Transliteration Alignment, In Proc. of ROCLING XV, 1-16.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "Backward machine transliteration by learning phonetic similarity", |
| "authors": [ |
| { |
| "first": "W.-H", |
| "middle": [], |
| "last": "Lin", |
| "suffix": "" |
| }, |
| { |
| "first": "H.-H", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "Proc. of CONLL", |
| "volume": "", |
| "issue": "", |
| "pages": "139--145", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W.-H. Lin and H.-H. Chen. 2002. Backward machine transliteration by learning phonetic similarity. In Proc. of CONLL 2002: 139-145.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "Translation of Web Queries using Anchor Text Mining", |
| "authors": [ |
| { |
| "first": "W.-H", |
| "middle": [], |
| "last": "Lu", |
| "suffix": "" |
| }, |
| { |
| "first": "L.-F", |
| "middle": [], |
| "last": "Chien", |
| "suffix": "" |
| }, |
| { |
| "first": "H.-J", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "ACM Transactions on Asian Language Information Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "159--172", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W.-H. Lu., L.-F. Chien and H.-J. Lee. 2002. Translation of Web Queries using Anchor Text Mining, ACM Transactions on Asian Language Information Processing (TALIP), 159-172.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Anchor Text Mining for Translation of Web Queries: A Transitive Translation Approach", |
| "authors": [ |
| { |
| "first": "W.-H", |
| "middle": [], |
| "last": "Lu", |
| "suffix": "" |
| }, |
| { |
| "first": "L.-F", |
| "middle": [], |
| "last": "Chien", |
| "suffix": "" |
| }, |
| { |
| "first": "H.-J", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "ACM Transactions on Information Systems", |
| "volume": "22", |
| "issue": "2", |
| "pages": "242--269", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W.-H. Lu., L.-F. Chien and H.-J. Lee. 2004. Anchor Text Mining for Translation of Web Queries: A Transitive Translation Approach. ACM Transactions on Information Systems 22(2): 242-269.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "A Bottom-up Merging Algorithm for Chinese Unknown Word Extraction", |
| "authors": [ |
| { |
| "first": "W.-Y", |
| "middle": [], |
| "last": "Ma", |
| "suffix": "" |
| }, |
| { |
| "first": "K.-J", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Proc. of ACL workshop on Chinese Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "31--38", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "W.-Y. Ma and K.-J. Chen. 2003. A Bottom-up Merging Algorithm for Chinese Unknown Word Extraction, In Proc. of ACL workshop on Chinese Language Processing 2003: 31-38.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "Models of translational equivalence among words", |
| "authors": [ |
| { |
| "first": "I", |
| "middle": [ |
| "D" |
| ], |
| "last": "Melamed", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "Computational Linguistics", |
| "volume": "26", |
| "issue": "2", |
| "pages": "221--249", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "I. D. Melamed. 2000. Models of translational equivalence among words. Computational Linguistics, 26(2):221-249.", |
| "links": null |
| }, |
| "BIBREF23": { |
| "ref_id": "b23", |
| "title": "Generate Phonetic Cognates to Handle Name Entities in English-Chinese Cross-Language Spoken Document Retrieval", |
| "authors": [ |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Meng", |
| "suffix": "" |
| }, |
| { |
| "first": "W.-K", |
| "middle": [], |
| "last": "Lo", |
| "suffix": "" |
| }, |
| { |
| "first": "B", |
| "middle": [], |
| "last": "Chen", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Tang", |
| "suffix": "" |
| } |
| ], |
| "year": 2001, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "H. Meng, W.-K. Lo, B. Chen and K. Tang. 2001. Generate Phonetic Cognates to Handle Name Entities in English-Chinese Cross-Language Spoken Document Retrieval, ASRU 2001.", |
| "links": null |
| }, |
| "BIBREF24": { |
| "ref_id": "b24", |
| "title": "Fuzzy Translation of Cross-Lingual Spelling Variants", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Pirkola", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Toivonen", |
| "suffix": "" |
| }, |
| { |
| "first": "H", |
| "middle": [], |
| "last": "Keskustalo", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Visala", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Jarvelin", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Proc. of SIGIR", |
| "volume": "", |
| "issue": "", |
| "pages": "345--352", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "A. Pirkola, J. Toivonen, H. Keskustalo, K. Visala and K. Jarvelin. 2003. Fuzzy Translation of Cross-Lingual Spelling Variants, In Proc. of SIGIR 2003: 345-352.", |
| "links": null |
| }, |
| "BIBREF25": { |
| "ref_id": "b25", |
| "title": "Finding Ideographic Representations of Japanese Names Written in Latin Script via Language Identification and Corpus Validation", |
| "authors": [ |
| { |
| "first": "Y", |
| "middle": [], |
| "last": "Qu", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Grefenstette", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "184--191", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Y. Qu and G. Grefenstette. 2004. Finding Ideographic Representations of Japanese Names Written in Latin Script via Language Identification and Corpus Validation In Proc. of ACL 2004: 184-191.", |
| "links": null |
| }, |
| "BIBREF26": { |
| "ref_id": "b26", |
| "title": "Automatic identification of word translations from unrelated English and German corpora", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Rapp", |
| "suffix": "" |
| } |
| ], |
| "year": 1999, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "519--526", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "R. Rapp. 1999. Automatic identification of word translations from unrelated English and German corpora, In Proc. of ACL 1999: 519-526.", |
| "links": null |
| }, |
| "BIBREF27": { |
| "ref_id": "b27", |
| "title": "The N-best algorithm: An efficient and exact procedure for finding the N most likely sentence hypothesis", |
| "authors": [ |
| { |
| "first": "R", |
| "middle": [], |
| "last": "Schwartz", |
| "suffix": "" |
| }, |
| { |
| "first": "Y.-L", |
| "middle": [], |
| "last": "Chow", |
| "suffix": "" |
| } |
| ], |
| "year": 1990, |
| "venue": "Proc. of ICCASP", |
| "volume": "", |
| "issue": "", |
| "pages": "81--84", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "R. Schwartz and Y.-L. Chow. 1990. The N-best algorithm: An efficient and exact procedure for finding the N most likely sentence hypothesis. In Proc. of ICCASP 1990: 81-84.", |
| "links": null |
| }, |
| "BIBREF28": { |
| "ref_id": "b28", |
| "title": "Translating collocations for bilingual lexicons: a statistical approach", |
| "authors": [ |
| { |
| "first": "F", |
| "middle": [], |
| "last": "Smadja", |
| "suffix": "" |
| }, |
| { |
| "first": "K", |
| "middle": [], |
| "last": "Mckeown", |
| "suffix": "" |
| }, |
| { |
| "first": "V", |
| "middle": [], |
| "last": "Hatzivassiloglou", |
| "suffix": "" |
| } |
| ], |
| "year": 1996, |
| "venue": "Computational Linguistics", |
| "volume": "22", |
| "issue": "1", |
| "pages": "1--38", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "F. Smadja, K. McKeown, and V. Hatzivassiloglou. 1996. Translating collocations for bilingual lexicons: a statistical approach. Computational Linguistics, 22(1):1-38.", |
| "links": null |
| }, |
| "BIBREF29": { |
| "ref_id": "b29", |
| "title": "Transliteration of Proper Names in Cross-Lingual Information Retrieval", |
| "authors": [ |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Virga", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Khudanpur", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "ACL 2003 workshop MLNER", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "P. Virga and S. Khudanpur. 2003. Transliteration of Proper Names in Cross-Lingual Information Retrieval. ACL 2003 workshop MLNER.", |
| "links": null |
| }, |
| "BIBREF30": { |
| "ref_id": "b30", |
| "title": "Automatic English-Chinese name transliteration for development of multilingual resources", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Wan", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [ |
| "M" |
| ], |
| "last": "Verspoor", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "Proc. of ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "1352--1357", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "S. Wan and C. M. Verspoor. 1998. Automatic English-Chinese name transliteration for development of multilingual resources. In Proc. of ACL 1998: 1352-1357.", |
| "links": null |
| }, |
| "BIBREF31": { |
| "ref_id": "b31", |
| "title": "Extracting pronunciation-translated names from Chinese texts using bootstrapping approach", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Xiao", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Liu", |
| "suffix": "" |
| }, |
| { |
| "first": "T.-S", |
| "middle": [], |
| "last": "Chua", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "the 1st SIGHAN workshop on Chinese Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "J. Xiao, J. Liu and T.-S. Chua. 2002. \"Extracting pronunciation-translated names from Chinese texts using bootstrapping approach\", the 1st SIGHAN workshop on Chinese Language Processing , Taipei, Taiwan, Aug 2002.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "text": "To cover all possibly correct mapping between English TUs and Chinese transliterated characters for the distinct types of English-Chinese transliteration pairs with the same or different transliteration units, we propose a simple mixed-syllable-mapping transliteration model, which combine the forward-syllable-mapping and reverse-syllable-mapping transliteration models, to estimate the mapping score as follows: Web-based Semi-Supervised Learning Algorithm: We intend to take advantages of abundant language-mixed texts on the Web to collect English-Chinese transliteration pairs and then train a good quality transliteration model. Thus, we design a semi-supervised learning process of transliteration mapping. The process is composed of three main stages: extraction of Chinese transliterated names, extraction of English original names, and learning of transliteration mapping, and described below as well as the algorithm inFigure 1.\uf09f Extraction of Chinese Transliterated Names:Xiao et al. (2002) have proposed a bootstrapping algorithm that uses only five frequent Chinese transliterated characters as initial seed character set: {\u963f, \u723e, \u5df4, \u65af, \u57fa} to automatically collect over 100,000 of Chinese transliterated names by utilizing search-result pages. Inspired by Xiao et al., we design a different bootstrapping algorithm which uses the same seed character set to automatically find large amounts of Chinese transliterated names from search-result pages. Initially, we select two frequent Chinese transliterated characters from the seed character set, and then send them to search engines for getting search-results pages. To efficiently extract more Chinese transliterated names from the search-result pages, we use the CKIP tagger(Ma & Chen 2003), which is a representative Chinese POS tagger with the ability of segmenting Chinese texts into meaningful words and extracting unknown words.\uf09f Extraction of English Original Names:We first use the search-result-based translation extraction method (Section 3.2) to find possible candidates of English original names, and then filter out the impossible candidates which are included in general-purpose bilingual dictionaries. Finally, to collect English-Chinese transliteration name pairs with high quality, we may need to take some manual efforts to examine the correct transliteration pairs. \uf09f Learning of Transliteration Mapping: On the basis of the English letter substring segmentation rules and the proposed mixed-syllable-mapping transliteration model described above, we will train a Web-based transliteration model based on the collected transliteration pairs from the Web.", |
| "uris": null, |
| "type_str": "figure", |
| "num": null |
| }, |
| "FIGREF1": { |
| "text": "1.4. Seed character set updating: update C s by adding the new characters from the new Chinese transliterated names. 1.5. Repeat step1 until the desired number of the Chinese transliterated name in the V c is reached. 2. Extraction of English original names: for each potential Chinese transliterated name in V c , perform the following sub-steps: 2.1. Potential English name extraction: use search-result-based translation extraction method (Section 3.2) to find potential candidates of English name. 2.2. Candidate filtering: filter out impossible English name candidates included in D. 2.3. English name identification: take some manual efforts to examine the correct original English names.2.4. English-Chinese transliteration pair updating: update V ec by adding the new transliteration pair. 3. Learning of English-Chinese transliteration mapping: use the proposed mixed-syllable-mapping transliteration model (equation (6)) to train a Web-based transliteration model T based on the extracted English-Chinese transliteration pairs.", |
| "uris": null, |
| "type_str": "figure", |
| "num": null |
| }, |
| "FIGREF2": { |
| "text": "Algorithm for collecting transliteration pairs and training a transliteration model.", |
| "uris": null, |
| "type_str": "figure", |
| "num": null |
| }, |
| "TABREF0": { |
| "content": "<table/>", |
| "type_str": "table", |
| "html": null, |
| "text": "have developed a fully rule-based algorithm to transliterate English proper names into Chinese names. We simplify their syllabification techniques to generate a few simple heuristic rules of segmenting an English name into letter substrings. Each English substring is regarded as a transliteration unit (TU) in this paper and had at most one corresponding character of the Chinese transliterated name. Initially, we used only five rules listed below: \uf0b7 a, e, i, o, u are vowels, and y is also regarded as a vowel if it appears behind a consonant. All other letters are consonants.", |
| "num": null |
| }, |
| "TABREF1": { |
| "content": "<table><tr><td/><td/><td colspan=\"3\">Forward-Syllable-Mapping</td><td colspan=\"3\">Mixed-Syllable-Mapping</td></tr><tr><td colspan=\"2\">Translation Method</td><td colspan=\"3\">Transliteration Model</td><td colspan=\"3\">Transliteration Model</td></tr><tr><td/><td/><td>Top-1</td><td>Top-3</td><td>Top-5</td><td>Top-1</td><td>Top-3</td><td>Top-5</td></tr><tr><td colspan=\"2\">Name Transliteration</td><td>14%</td><td>27%</td><td>27%</td><td>27%</td><td>32%</td><td>32%</td></tr><tr><td>Hybrid</td><td/><td>36%</td><td>41%</td><td>45%</td><td>45%</td><td>50%</td><td>55%</td></tr><tr><td>Table 2.</td><td colspan=\"6\">Comparison of translation results between the different translation methods.</td></tr><tr><td/><td/><td/><td colspan=\"5\">Mixed-Syllable-Mapping Transliteration Model</td></tr><tr><td colspan=\"2\">Translation Method</td><td/><td colspan=\"2\">Top-1</td><td>Top-3</td><td colspan=\"2\">Top-5</td></tr><tr><td colspan=\"2\">Search-Result-based</td><td/><td>36%</td><td/><td>36%</td><td colspan=\"2\">45%</td></tr><tr><td colspan=\"2\">Name Transliteration</td><td/><td>27%</td><td/><td>32%</td><td colspan=\"2\">32%</td></tr><tr><td colspan=\"2\">Linear Combination</td><td/><td>32%</td><td/><td>41%</td><td colspan=\"2\">55%</td></tr><tr><td>Hybrid</td><td/><td/><td>45%</td><td/><td>50%</td><td colspan=\"2\">55%</td></tr><tr><td colspan=\"8\">Table 3. Effective results of translation extraction using the hybrid translation extraction method</td></tr><tr><td/><td colspan=\"5\">(underlined terms indicate correct translation).</td><td/></tr><tr><td/><td/><td/><td colspan=\"4\">Top 5 Translation Candidates</td></tr><tr><td>Test Query</td><td>Translation</td><td colspan=\"3\">Forward-Syllable-Mapping</td><td colspan=\"3\">Mixed-Syllable-Mapping</td></tr><tr><td/><td>Method</td><td colspan=\"3\">Transliteration Model</td><td colspan=\"3\">Transliteration Model</td></tr><tr><td colspan=\"2\">Search-Result-based</td><td colspan=\"3\">\u9ea5\u53ef\u5e03\u96f7,\u9ea5\u514b\u5091\u514b\u68ee,\u65bd\u6587 \u5f6c,\u83ef\u7d0d,\u500b\u4eba</td><td/><td/></tr></table>", |
| "type_str": "table", |
| "html": null, |
| "text": "Comparison of translation results between the forward-syllable-mapping model and the mixed-syllable-mapping model.", |
| "num": null |
| }, |
| "TABREF2": { |
| "content": "<table><tr><td>Test Query</td><td>Translation Method</td><td>Top 5 Translation Candidates</td></tr><tr><td/><td/><td>Mixed-Syllable-Mapping Transliteration Model</td></tr><tr><td/><td>Search-Result-based</td><td>\u5049\u54e5,\u98df\u54c1\u85e5\u7269,\u5a01\u800c\u525b,\u85e5\u7269\u7ba1\u7406\u5c40,\u85e5\u7269</td></tr><tr><td>Viagra</td><td>Name Transliteration Linear Combination</td><td>\u8587\u963f\u683c\u62c9,\u8587\u4e9e\u683c\u62c9,\u8587\u827e\u683c\u62c9,\u8587\u963f\u845b\u62c9,\u8587\u4e9e\u845b\u62c9 \u5049\u54e5,\u98df\u54c1\u85e5\u7269,\u8587\u963f\u683c\u62c9,\u5a01\u800c\u525b,\u85e5\u7269\u7ba1\u7406\u5c40</td></tr><tr><td/><td>Hybrid</td><td>\u842c\u827e\u53ef,</td></tr></table>", |
| "type_str": "table", |
| "html": null, |
| "text": "\u85e5\u7269\u7ba1\u7406,\u98df\u54c1\u7ba1\u7406,\u8f1d\u745e,\u5a01\u800c\u525b transliteration pairs from the Web for training good quality transliteration model. Besides them, there", |
| "num": null |
| } |
| } |
| } |
| } |
