Teaching Computers to Read, Understand, and Write Human Languages

This paper presents a contemporary linguistic challenge: how to teach computers to read, understand, and write human languages. Researchers in linguistics and artificial intelligence need to work together to address this challenge. This paper outlines an attempt to address this challenge by creating an artificial intelligent system that can compose new English sentences to summarize documents. To process the documents conceptually to create abstractive summaries, the system makes use of one of the world's largest knowledgebase and one of the most powerful inference engines. The resultant AI system first uses natural language processing techniques to extracts syntactic structure of the documents and then maps the words of the sentences and their parts of speech into related concepts in the knowledgebase. It then uses the inference engine to generalize and fuse concepts to form more abstract concepts. The system then composes new sentences based on the key concepts by linking subject concepts with their related predicate concepts. The system has been implemented and tested. The test results showed that the system can create new sentences that include abstracted concepts not explicitly mentioned in the original documents and that contain information synthesized from different parts of the documents to compose a summary.