Textual Signatures: Identifying Text-Types Using Latent Semantic Analysis to Measure the Cohesion of Text Structures - Natural Language Processing and Text Mining - page 111

Fig. 7.1. Textualsignature formedfrommeansofthe abstracttoother sections for
Experiments 1,2and 3.

the number ofwords in eachtext sectionincluding the within-text factorsof Intro-
duction ( M=1598.015, SD=871.247 ), Method ( M =1295.791, SD =689.756), Results
( M =1408.627, SD =841.185), and Discussion ( M =1361.284, SD =653.742). There was
a main effect of comparisontype, F (3,66)= 2.955, MSE =382691.182, p=.034. Pair-
wise contrasts(seeTable 7.2) indicated that theonly significant differences were
betweenthe AI/AM comparisontypes and the AI/AD comparisontype. The re-
sultsconfirmed that thesection length signature does notreflectthe LSA signature
(see Figure 7.2). Removing words that LSA does notaccount forfromthis analysis
(suchas numbers) made nosignificant difference to the results.

Table 7.2. Pairwise Comparisonsofthe Relatedness ofText Sections to the Ab-
stractfortext length

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

Introduction (AI) Diff=302.22 (113.13)*Diff=189.39 (107.91) Diff=236.73 (94.37)*
Method (AM) Diff=-112.84 (120.59) Diff=-65.49 (105.67)

Results(AR) Diff=47.34 (97.41)

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