Shared-Packed Parse Forests and Logical Form

为了同时处理句法和语义的歧义，我们建议使用Seo和Simmons(1989)的共享打包句法森林，并结合Harper(1990)提出的句子的逻辑形式表示。如果单独使用压缩语法分析林，则在处理句法歧义时不能做出有关语义歧义的明智决定。如果只使用逻辑形式，则必须在为句子提供逻辑形式之前解决所有句法歧义。在给定解析树的情况下，可以精确地指定逻辑形式，但是使用一次提供一个解析树的解析器，找到正确的解析树需要指数级的时间。该算法将逻辑形式与密集森林相结合，将在多项式时间内为歧义语句产生紧凑的中间表示。由于句法和语义歧义是相互制约的，在中间表征中保持两者应该集中在它们的相互有效性上。随着更多信息的处理，该中间表示将被递增地更新，直到解析森林被缩减为包含一种逻辑形式的单个树。此外，这种方法的模块化将降低自然语言问题的概念复杂性。

To simultaneously handle syntactic and semantic ambiguity, we propose using the shared-packed parse forest of Seo and Simmons (1989) in conjunction with the logical form representation of a sentence developed by Harper (1990). If the packed parse forest is used alone, informed decisions about semantic ambiguities cannot be made while addressing syntactic ambiguities. If logical form alone is used, all syntactic ambiguities must be resolved before providing the logical form for the sentence. Logical form can be specified precisely given a parse tree, but with a parser that provides one parse tree at a time, finding the correct parse tree requires exponential time. The algorithm resulting from the wedding of logical form to packed forests will produce a compact intermediate representation for an ambiguous sentence in polynomial time. As syntactic and semantic ambiguity are mutually constraining, maintaining both in the intermediate representation should focilitate their inter- effectiveness. As more information is processed this intermediate representation will be incrementally updated until the parse forest is reduced to a single tree containing one logical form. In addition, the modularity of this approach will reduce the conceptual complexity of the natural language problem.