Textual Signatures: Identifying Text-Types Using Latent Semantic Analysis to Measure the Cohesion of Text Structures - Natural Language Processing and Text Mining - page 113

Table 7.3. Pairwise comparison ofabstracttosimilar-themedbody

Method (AM) Results(AR) Discussion (AD)

Introduction (AI) Diff=.150(.032)*Diff=.135 (.032)*Diff=-.015 (.020)
Method (AM) Diff=-.015 (.026) Diff=-.165 (.036)*

Results(AR) Diff=.150(.032)*

Notes: Diff denotes the average difference betweenthecosines; * p<.01

7.10 Experiment 2a

One potential weakness of Experiment 2 was the relatively small sizeofthecorpus
(i.e., 20 texts). To alleviate theconcern that the results were a function ofthesize
oftext corpora, wesplit theoriginal 67-text corpus from Experiment 1 into three
random groupsof 20 texts and re-analyzedthe results. If 20 texts were a su ciently
sized corpus,thenthe analysisshouldyieldthesame pattern asobservedinExper-
iment 1. This analysis producedthree setsof scores for each ofthe four comparison
types (AI, AM, AR, and AD). Ascan beseenfrom Figure 7.3, the three new signa-
tures map almostperfectlyto theoriginalsignature from Experiment 1. There were
also nosignificant differences within thecorrespondingsection comparisons fromthe
three 20-text corpora.

Fig. 7.3. Comparison of Experiment 1 to three setsof 20 texts takenfromthe
originalcorpora.