Generating bilingual lexical equivalences from parallel texts

Automatic generation of Japanese–English bilingual thesauri based on bilingual corpora

“…Most of the methods proposed so far for extracting translation pairs from bilingual corpora are frequency-based methods (Ahrenberg et al, 1998;Boutsis et al, 1999;Brown et al, 1993;Chen et al, 1999;Daille et al, 1994;Fung, 1995;Fung & McKeown, 1994;Gale & Church, 1991;Gaussier, 1998;Hiemstra, 1998;Hull, 1998;Kageura et al, 2000;Kitamura & Matsumoto, 1996;Kumar & Byrne, 2002;Kupiec, 1993;Lopez et al, 2002;Melamed, 2000;Shen & Dorr, 1997;Smadja et al, 1996;Sun et al, 2000;Toutanova et al, 2002;van der Eijk ,1993;Vogel et al, 1996;Wu & Xia, 1994;Yamamoto & Matsumoto, 2000). In this section, we will show that frequency-based methods are not effective for extracting low-frequency translation pairs; in other words, we will validate assumption (b) in the previous section.…”

Plant‐insect interactions: what can we learn from plant lectins?

“…Most of the methods proposed so far for extracting translation pairs from bilingual corpora are frequency‐based methods (Ahrenberg et al, 1998; Boutsis et al, 1999; Brown et al, 1993; Chen et al, 1999; Daille et al, 1994; Fung, 1995; Fung & McKeown, 1994; Gale & Church, 1991; Gaussier, 1998; Hiemstra, 1998; Hull, 1998; Kageura et al, 2000; Kitamura & Matsumoto, 1996; Kumar & Byrne, 2002; Kupiec, 1993; Lopez et al, 2002; Melamed, 2000; Shen & Dorr, 1997; Smadja et al, 1996; Sun et al, 2000; Toutanova et al, 2002; van der Eijk ,1993; Vogel et al, 1996; Wu & Xia, 1994; Yamamoto & Matsumoto, 2000). In this section, we will show that frequency‐based methods are not effective for extracting low‐frequency translation pairs; in other words, we will validate assumption (b) in the previous section.…”

“…If these clusters can be extracted automatically, it would be very useful. While some methods have been proposed for automatically extracting bilingual clusters from bilingual corpora (Kageura, Tsuji, & Aizawa, 2000) and many methods have been proposed for automatically extracting translation pairs from bilingual corpora (Ahrenberg, Andersson, & Merkel, 1998; Boutsis, Piperidi, & Demiros, 1999; Brown et al, 1993; Chen, Kishida, Jiang, Liang, & Gey, 1999; Collier, Hirakawa, & Kumano, 1997; Daille, Gaussier, & Lange, 1994; Fung, 1995; Fung & McKeown, 1994; Gale & Church, 1991; Gaussier, 1998; Hiemstra, 1998; Hull, 1998; Imamura, 2002; Jeong, Myaeng, Lee, & Choi, 1999; Ker & Chang, 1997; Kitamura & Matsumoto, 1996; Kumar & Byrne, 2002; Kupiec, 1993; Lopez, Nossal, Hwa, & Resnick, 2002; Melamed, 2000; Meyers, Yangarber, & Grishman, 1996; Shen & Dorr, 1997; Smadja, McKeown, & Hatsivsilloglou, 1996; Sun, Jin, Du, & Sun, 2000; Toutanova, Ilhan, & Manning, 2002; van der Eijk, 1993; Vogel, Ney, & Tillmann, 1996, Watanabe, Kurohashi, & Aramaki, 2000; Wu, 1995; Wu & Xia, 1994; Yamamoto & Matsumoto, 2000), most of these methods depend heavily on word frequency and are thus ineffective for extracting low‐frequency bilingual clusters (see The Problem With Using a Frequency‐Based Method for Extracting Low‐Frequency Pairs subsection). As the frequencies of newly coined words in the corpora are low by nature, most of these methods cannot properly extract newly coined clusters 1…”