Analysis of the Interpretation at the Relational and ASF Level

Abstract

Two important aspects of understanding a text are the ability to skim it, extracting important elements (a coarse-grain view of comprehension), and the ability to read it “deeply” (a fine-grain view of comprehension). A computational analogue that mimics skimming should include a representation of a set of semantic relationships about the text that can be used to summarize it and extract what is important. A computational analogue that supports a deep reading of the text should be able to represent the background details (non- systematic relationships) associated with the concepts in the text, including the larger frame in which the text concepts are situated. Consider the Fillmore (1982) example discussed in Chapter 1: understanding the difference between the concepts “land” and “ground” lies in recognizing how “land” is used in contrast to “sea,” and how “ground” is used in contrast to “air.” How can these nonsystematic relationships be represented? One approach is to represent them via statistical associations gathered from on-line text corpora.