In hierarchical phrase-based translation, coarse-grained nonterminal Xs may generate inappropriate translations due to the lack of sufficient information for phrasal substitution. In this paper we propose a framework to refine nonterminals in hierarchical translation rules with real-valued semantic representations. The semantic representations are learned via a weighted mean value and a minimum...

We observe that (1) it is difficult to combine transliteration and meaning translation when transforming named entities (NE); and (2) there are different translation variations in NE translation, due to different semantic information. From this basis, we propose a novel semantic-specific NE translation model, which automatically incorporates the global context from corpus in order to capture su...

translator is also dealing with two different cultures at the same time. this particular aspect seems to have been taken into consideration by the second group of theorists who regard translation equivalence being essentially a transfer of the message from the sc to the tc and a pragmatic/ semantic or functionally oriented approach to translation . finally, there are other translation scholars ...

Semantic parsing, the task of mapping natural language sentences to logical forms, has recently played an important role in building natural language interfaces and question answering systems. In this talk, I will present three ways in which semantic parsing relates to machine translation: First, semantic parsing can be viewed *as* a translation task with many of the familiar issues, e.g., dive...

Many image-to-image (I2I) translation problems are in nature of high diversity that a single input may have various counterparts. The multi-modal network can build many-to-many mapping between two visual domains has been proposed prior works. However, most them guided by sampled noises. Some others encode the reference image into latent vector, which would eliminate semantic information image. ...

Linguistic steganography (LS) aims to embed secret information into a highly encoded text for covert communication. It can be roughly divided two main categories, i.e., modification based LS (MLS) and generation (GLS). MLS embeds data by slightly modifying given without impairing the meaning of text, whereas GLS uses well trained language model directly generate carrying data. A common disadvan...