Multilingual Sentence Generation

This paper presents an overview of a robust, broad-coverage, and application-independent natural language generation system. It demonstrates how the different language generation components function within a multilingual Machine Translation (MT) system, using the languages that we are currently working on (English, Spanish, Japanese, and Chinese). Section 1 provides a system description. Section 2 focuses on the generation components and their core set of rules. Section 3 describes an additional layer of generation rules included to address applicationspecific issues. Section 4 provides a brief description of the evaluation method and results for the MT system of which our generation components are a part. 1 System Description We present a natural language generation method in the context of a multi-lingual MT system. The system that we have been developing is a hybrid system with rule-based, example-based, and statistical components. Analysis and generation are performed with linguistic parsers and syntactic realization modules, the rules of which are coded by hand. Transfer is accomplished using transfer rules/mappings automatically extracted from aligned corpora. The MT process starts with a source sentence being analyzed by the source-language parser, which produces as output a syntactic tree. This tree is input to the Logical Form module, which produces a deep syntactic representation of the input sentence, called the LF (Heidorn, G. E., 2000). The LF uses the same basic set of relation types for all languages. Figure 1 gives the syntactic tree and LF for the simple English sentence, “I gave the pencils to John”.