Skip to main content

Table 5 Statistical tests (confidence intervals) for differences between correlations reported in Table 3

From: tESA: a distributional measure for calculating semantic relatedness

tESA config

Other method

Dataset

CI

Comparison

Medline

DS (Medline)

mayo29ph

(0.09; 0.59)

+

Medline

ESA (Medline)

umnsrsRel

(0.003; 0.09)

+

Medline

ESA (PMC)

umnsrsRel

(0.025; 0.099)

+

Medline

ESA (Wiki)

umnsrsRel

(0.097; 0.2)

+

Medline

DS (Medline)

umnsrsRel

(0.13; 0.25)

+

Medline

DS (PMC)

umnsrsRel

(0.026; 0.12)

+

Medline

DS (Wiki)

umnsrsRel

(0.15; 0.26)

+

Medline

tESA (PMC)

umnsrsRel

(0.02; 0.09)

+

Medline

tESA (Wiki)

umnsrsRel

(0.11; 0.22)

+

Medline

ESA (PMC)

umnsrsSim

(0.004; 0.08)

+

Medline

ESA (Wiki)

umnsrsSim

(0.09; 0.19)

+

Medline

DS (Medline)

umnsrsSim

(0.14; 0.26)

+

Medline

DS (Wiki)

umnsrsSim

(0.11; 0.24)

+

Medline

tESA (Wiki)

umnsrsSim

(0.1; 0.21)

+

PMC

DS (Medline)

mayo29ph

(0.1; 0.61)

+

PMC

ESA (Wiki)

umnsrsRel

(0.04; 0.15)

+

PMC

DS (Medline)

umnsrsRel

(0.07; 0.2)

+

PMC

DS (Wiki)

umnsrsRel

(0.096; 0.21)

+

PMC

tESA (Wiki)

umnsrsRel

(0.06; 0.16)

+

PMC

ESA (Wiki)

umnsrsSim

(0.056; 0.16)

+

PMC

DS (Medline)

umnsrsSim

(0.1; 0.24)

+

PMC

DS (Wiki)

umnsrsSim

(0.09; 0.2)

+

PMC

tESA (Wiki)

umnsrsSim

(0.07; 0.18)

+

Wiki

DS (Medline)

mayo29c

(0.04; 0.55)

+

Wiki

DS (Medline)

mayo29ph

(0.11; 0.62)

+

Wiki

DS (Wiki)

mayo29ph

(0.01; 0.41)

+

Wiki

ESA (Medline)

umnsrsRel

(-0.18; -0.07)

-

Wiki

ESA (PMC)

umnsrsRel

(-0.15; -0.05)

-

Wiki

DS (PMC)

umnsrsRel

(-0.16; -0.025)

-

Wiki

ESA (Medline)

umnsrsSim

(-0.19; -0.086)

-

Wiki

ESA (PMC)

umnsrsSim

(-0.16; -0.06)

-

Wiki

DS (PMC)

umnsrsSim

(-0.21; -0.07)

-

  1. The CIs were constructed for pairs of correlations involving at least one tESA setup. The table provides all the information necessary to track the CI back to Table 3, i.e. the corpus of the tESA method, the method (and corpus) to which the tESA results are being compared and the reference dataset. We also provide the CI itself, additionally indicating if the result is positive or negative