In the hybrid model, given an input sentence, a lattice that consists of word-level and character-level nodes is constructed.

Word-level nodes, which correspond to

In other words, we use word-level nodes to identify known words and character-level nodes to identify unknown words.

Since we were interested in finding an optimal combination of word-level and character-level nodes for training, we focused on tuning 7“.

Ideally, we need to build a word-character hybrid model that effectively learns the characteristics of unknown words (with character-level nodes) as well as those of known words (with word-level nodes).

If we select the correct path yt that corresponds to the annotated sentence, it will only consist of word-level nodes that do not allow learning for unknown words.

We therefore need to choose character-level nodes as correct nodes instead of word-level nodes for some words.

W0 <w0> for word-level W1 <p0> nodes

Templates W0—W3 are basic word-level unigram features, where Length(w0) denotes the length of the word wo.

Templates B0-B9 are basic word-level bigram features.

Word-level alignments are then drawn based on the tree alignment.

Finally, parallel sentences are assembled from these generated part-of-speech sequences and word-level alignments.

The model is trained using bilingual data with automatically induced word-level alignments, but is tested on purely monolingual data for each language.

word-level alignments, as observed data.

We obtain these word-level alignments using GIZA++ (Och and Ney, 2003).

Finally, word-level alignments are drawn based on the structure of the alignment tree.

Assuming that trees induced over parallel sentences have to exhibit certain structural regularities, Kuhn manually specifies a set of rules for determining when parsing decisions in the two languages are inconsistent with GIZA++ word-level alignments.

0 Word-level system combination (Rosti et al., 2007) of member decoders’ n-best outputs

Word-level system combination (system combination hereafter) (Rosti et al., 2007; He et al., 2008) has been proven to be an effective way to improve machine translation quality by using outputs from multiple systems.

We also implemented the word-level system combination (Rosti et al., 2007) and the hypothesis selection method (Hildebrand and Vogel, 2008).

Word-level Comb 4045/4085 2952/3035 Hypo Selection 40.09/40.50 29.02/29.71

Most of the work focused on seeking better word alignment for consensus-based confusion network decoding (Matusov et al., 2006) or word-level system combination (He et al., 2008; Ayan et al., 2008).

We also conduct extensive investigations when different settings of co-decoding are applied, and make comparisons with related methods such as word-level system combination of hypothesis selection from multiple n-best lists.

We consider generative and discriminative models for dependency grammar induction that use word-level alignments and a source language parser (English) to constrain the space of possible target trees.

A parallel corpus is word-level aligned using an alignment toolkit (Graca et al., 2009) and the source (English) is parsed using a dependency parser (McDonald et al., 2005).

For example, several early works (Yarowsky and Ngai, 2001; Yarowsky et al., 2001; Merlo et al., 2002) demonstrate transfer of shallow processing tools such as part-of-speech taggers and noun-phrase chunkers by using word-level alignment models (Brown et al., 1994; Och and Ney, 2000).