英文摘要
|
Humanists are often interested in people and events occurred in certain time and space. Binary relation is the simplest form to represent the relationships among objects (people, events, time, and space). In this paper, we use a binary relation to describe the relationship between two objects. For instance, a binary relation is_father_of(x, y) describes that the person x is the father of the person y, and occurs_at(x, y) says that the term x occurs at the text y. Analyzing a collection of binary relations can help humanists find properties hard to get from individual relations. For instance, one can infer the family relationship from a collection of is_father_of relations, and analyze the co-occurrence of terms occurred in texts from a collection of occurs_at relations. DocuSky is a research platform developed by the National Taiwan University (NTU) Research Center for Digital Humanities. It allows a user to build personal databases which support fulltext retrieval, post-classification, text analysis, and data visualization. With the help of text-annotation tools, one can tag terms in some text, mark out a binary relation between two tags, and specify a type name to characterize the relation. Then she can convert the annotated result to build a DocuSky database. Once the database is constructed, the user can make use of various tools in DocuSky to search terms over the database, to analyze tagged terms and generate a statistical report, and to invoke visualization tools to illustrate analyzed results. Visualization helps people realize complicated data relationships. The goal of this research is to follow the guidelines of DocuSky Lite Tools to develop a visualization tool, BinRelLite, to help one plot a graph from binary relations. The primary goal of this tool is to smooth the process from text annotation to data visualization. BinRelLite draws a graph from a collection of binary relations (come from text annotation or other data resources). In the graph, a binary relation R(x, y) is represented by two nodes x, y and an edge R between these nodes. BinRelLite allows its users to plot the relations as a directed or an undirected graph. It provides simple editing functions to allow one to modify node titles and edge titles without changing their values in the data source. It also offers filtering functions to help one select a desirable subset from a large collection of binary relations. This paper discusses the data application process as well as some design and implementation issues of BinRelLite. We use the dialog markups of Sanguo Yanyi (三國演義) chapters 42 to 51 (第 42-51 回) to show that, with the help of BinRelLite and graph visualization, it can be easier to spot interesting properties from a large set of relations.
|