Báo cáo khoa học: "How Much Can We Gain from Supervised Word Alignment" potx

... limit of supervised word alignment. More specifically, we would like to know what magnitude of gain in MT performance we can ex- pect from supervised alignment over the state of the art unsupervised ... Oregon, June 19-24, 2011. c 2011 Association for Computational Linguistics How Much Can We Gain from Supervised Word Alignment? Jinxi Xu and Jinying Chen...

Ngày tải lên: 23/03/2014, 16:20

... empty set at iteration t = 0. We iterate for T itera- tions. We take a pool-based active learning strat- egy, where we have access to all the automatically aligned links and we can score the links based on ... We will be exploring alternative formula- tions to this strategy. We observe that confidence based metrics perform significantly better than the baseline. From the scatter...

Ngày tải lên: 30/03/2014, 21:20

... each (weighted) rule of R following the method of Graehl et al. (2008). Similarly, we can adjust the training procedure of Graehl et al. (2008), which yields that we can automatically obtain a weighted LMBOT ... utilize the additional expressive power 825 of MBOT. However, Zhang et al. (2008b) and Sun et al. (2009) demonstrate that the additional expres- sivity gained from non-conti...

Ngày tải lên: 17/03/2014, 00:20

... properties of many phonetic elements deviate significantly from their canonical form (Greenberg, 1999). In the light of the apparent discrepancy between the canonical forms and the actual spoken lan- guage, ... only to the canonical forms of words, and no syllabification of spoken word forms exists. The methods of automatic syllabification have been applied and tested exclusively on canonica...

Ngày tải lên: 20/02/2014, 04:20

... other words in the sentence as well. One typical error is found in infinitive constructions, where a phrase like to gun down is tagged as INTO NOUN IN (a prepo- sitional 'to' followed ... opposed to verbs, which results in the misclassification described above. However, when these words are fol- lowed by a particle, they are usually verbs, and in the infinitive construction...

Ngày tải lên: 20/02/2014, 21:20

... cell extracts were screened using antibodies raised against the previously purified puta- tive binding domain and six a-amylase genes from molluscan species were sequenced. Furthermore, we performed ... C-terminal domains were identified by blast search in the GenBank database. From the alignment of these domains with those of P. haloplanktis and Caenorhabditis elegans, PCR primers were d...

Ngày tải lên: 14/02/2014, 18:20

... spectroscopy, as well as for critically reading the manuscript. We are indebted to Maria Samalikova for performing the mass spectrometry. We also wish to thank the anonymous reviewer who suggested ... MO, USA). After 20 min of incubation on ice, cells were dis- rupted by sonication (four cycles of 10 s each at 50% power output). Cell extracts were centrifuged for 30 min at 10 000 g at 4 °...

Ngày tải lên: 18/02/2014, 04:20

... epitopes recognized by the detection reagents were masked, e.g. by lipoproteins. However, this scenario cannot explain the failure to detect the envelope protein by western blotting [26]. (c) As noted ... date, no CD81-negative cell line has been identified that can be significantly infected with HCVpp. However, not all CD81-positive cell lines can be infected [47,64,66]. Antibodies to CD81...

Ngày tải lên: 19/02/2014, 06:20

... of significant interest. We consider whether, and how, the way in which the information is phrased — the choice of words and sentence structure — can affect this process. To this end, we develop ... human judgments — perhaps implicit, but in a form we can extract; and (iii) we can control for the setting in which the text was used. Specifically, we use the complete scripts of ro...

Ngày tải lên: 19/02/2014, 19:20

... keystroke the correct candidate words from appearing in the list of candidates in their native scripts, thereby pre- venting them from being entered altogether. Finally, we can collect a large ... common): follow- ing previous work (Zheng et al., 2011), we ob- tained word pairs from Wikipedia 3 and SpellGood 4 . We lower-cased the entries from these sources, re- moved the du...

Ngày tải lên: 19/02/2014, 19:20