Constraint based and hypothetical reasoning for human and molecular biology languages

点击翻译按钮获取中文摘要

Recent advances in the computational study of language grammars and rules, a branch of Artificial Intelligence, have the potential for practical, real-world applications. Most notably, the discovery of the human genome, partly through methodologies I helped develop (logic grammars), has highlighted the potential benefits of viewing molecular biology problems as specialized linguistic problems: DNA can be seen as a language with an alphabet of four letters (the nucleotides A, C, T and G), and  the genome can be viewed as the 'book of life'.      The proposed research aims at extending our understanding of not only human languages, but the language of life, by further interleaving these fields. I will explore the potential of promising methodologies for language processing, in particular HYPROLOG, our novel logic programming language which extends the possibilities of classical logic by allowing the user to explore "what if" scenarios, and to hypothesize possible explanations using abduction (the unsound but useful inference of p as a possible explanation for q given that p implies q).      I will also further explore a promising new linguistic paradigm, that of Property Grammars, which allows incomplete and even incorrect input to be parsed. Research will continue on my recently developed parser that invisibly endows descriptions made by non-computer specialists (e.g. linguists) with direct and automatic, while efficient, executability. Semantic aspects, and relation to current computational linguistic theories, will be examined in the medium term. In the long term, this research is expected to result in industrially useful expert systems for linguists, and perhaps even to have an impact on linguistic theory itself.      Finally, I will continue cross-fertilizing the natural language processing field with that of knowledge based systems, focussing on knowledge discovery. My recent research has uncovered the usefulness of my methodology for parsing property grammars as applied to knowledge discovery, in particular, for extracting concepts from arbitrary text found in the internet. This has many interesting short-term potential spin-offs, e.g. applications to phylogenetic classification.