Splitting Complex Sentences for Natural Language Processing Applications: Building a Simplified Spanish Corpus

This paper presents a new Spanish parallel corpus of original and syntactically simplified texts. The simplification carried out basically consists of opportunistically splitting a complex original sentence into several simple ones. This parallel corpus is envisioned as a first step in order to create an automatic syntactic simplification system to be used as a preprocessing tool for other Natural Language Processing tasks such as Text Summarization, Information Extraction, parsing or Machine Translation. The corpus has been evaluated by human annotators regarding its grammaticality and preservation of meaning. The results suggest that the meaning of simplified and original sentences is almost identical.