Discourse Topics by Watson Todd Richard;

Author:Watson Todd, Richard; [Watson Todd, Richard]
Language: eng
Format: epub
Publisher: John Benjamins Publishing Company
Published: 2016-11-22T12:53:29+00:00

Table 34. MI scores for co-occurrences of concepts in Text 1

The primed conceptual associations obtained from analysing a large corpus are valuable in allowing us to identify coherent moves between concepts in discourse, but need to be supplemented in two ways. First, as with topic-based analysis using relations, if two concepts appear in the same T-unit, we can assume that the speaker considers them to be associated. Second, even with a large corpus such as COCA, not every possible association will appear as a co-occurrence, especially for relatively infrequent words. For instance, earlier I argued that intuitively C3PO is likely to be associated with R2D2, Star Wars and robots. In Table 34, based on co-occurrences in COCA, C3PO is associated with the first two of these, but not with robots. The problem of limitations of coverage even in a large corpus can be solved by looking at how the associations of concepts cluster together.

Although C3PO is not directly associated with robots in COCA, C3PO is associated with R2D2 and Star Wars, both of which in turn are associated with robots. We might then argue that there is an indirect association between C3PO and robots. Such indirect associations are similar to the link sets of Berber Sardinha (2001), who argued that, in conducting a lexical cohesion analysis, two sentences could be considered bonded when they are both bonded to two or more other sentences even if they share no direct links. To identify such indirect associations, we can use the data in Table 34 to create an extended tree diagram that shows how closely associated all of the concepts are (see Corter 1996). To do this, I used the EXTREE program (Corter 1986) and gave empty cells in the table an arbitrary MI score of −20.0 (roughly the negative equivalent of the maximum positive MI score) to indicate a lack of co-occurrence in COCA. The extended tree diagram is shown in Figure 16. The leftmost vertical line covers all 18 concepts and I have taken concepts to form clusters of associations if they share a branch at least a distance of 2.00 away from this line. These clusters are circled in Figure 16, and moves between concepts in the same cluster can be considered coherent since they are indirectly associated.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.