Automatic Proposition Extraction from Dependency Trees: Helping Early Prediction of Alzheimer's Disease from Narratives

摘要: Idea Density (ID) was originally proposed as a way of measuring the memory load narratives, by representing underlying content text series semantic units, called propositions or ideas. From clinical perspective, this notion has been shown to correlate with several cognitive aspects, such memory, readability, aging, and dementia onset progress. Traditionally, are extracted manually from texts. There is tool that can automate ID extraction [1], but it uses shallow information input, doesn't produce themselves output. We propose novel approach obtaining automatically text. Our method an automation Chand et al.'s manual [2], consists rule-based system acting upon dependency trees. Initially, for each sentence in text, parser used elicit relations between words. Then, set rules recursively applied order process these yield corresponding propositions. analyze preliminary results our using well-formed journalistic speech transcriptions patients.