Improve text splitting and error handling in summarizer

Fix NLTK data paths and tokenizer loading

Initial commit: Add ParseAI document processor application