Academia.eduAcademia.edu

Figure 2 – uploaded by Giovanni Tapang

See full PDFdownloadDownload figure
Fig. 2. (Color online) Text boundaries and clustering coefficient of a synthetic article from equally sized (150 words) random extracts of Edgar Allan Poe’s works. The shaded regions straddle the poem/prose boundaries in the text with S = 75 and I = 1. (a) Clustering coefficient, (b) the difference in clustering coefficient per increment, and (c) prose and poem locations throughout the text (black = poem, white = prose).

Figure 2 (Color online) Text boundaries and clustering coefficient of a synthetic article from equally sized (150 words) random extracts of Edgar Allan Poe’s works. The shaded regions straddle the poem/prose boundaries in the text with S = 75 and I = 1. (a) Clustering coefficient, (b) the difference in clustering coefficient per increment, and (c) prose and poem locations throughout the text (black = poem, white = prose).

Related Figures (8)

Table 1. Calculated network parameters and their description. boundaries were not taken into account in determining the adjacent words. The last word of a sentence or paragraph is connected to the first word of the next sentence or paragraph.
Table 1. Calculated network parameters and their description. boundaries were not taken into account in determining the adjacent words. The last word of a sentence or paragraph is connected to the first word of the next sentence or paragraph.
Fig. 1. (Color online) Word adjacency network of (a) the poem The Summer Sun Shone Round Me, with N = 80, C = 0.02, D = 2.69, and (b) the short story The Tadpole and the Frog, with N = 53, C = 0.18 and D = 2.93, by Robert Louis Stevenson. The short story has greater clustering coefficient and average degree even though it has lesser words indicating a more connected network compared to that of the poem.
Fig. 1. (Color online) Word adjacency network of (a) the poem The Summer Sun Shone Round Me, with N = 80, C = 0.02, D = 2.69, and (b) the short story The Tadpole and the Frog, with N = 53, C = 0.18 and D = 2.93, by Robert Louis Stevenson. The short story has greater clustering coefficient and average degree even though it has lesser words indicating a more connected network compared to that of the poem.
Table 2. Calculated network metrics average over 34 prose anc 60 poems. Table 3. Effect size criterion and the classification function coefficients obtained using linear discriminant analysis.
Table 2. Calculated network metrics average over 34 prose anc 60 poems. Table 3. Effect size criterion and the classification function coefficients obtained using linear discriminant analysis.
the number of features that they used as predictor for hit song prediction and music genre classification to avoid overfitting of their classifier.1° The effect size effectively gives a normalized measure for the difference of two means on a standardized scale. The effect size is measured based on Cohen’s d!° as given by:
the number of features that they used as predictor for hit song prediction and music genre classification to avoid overfitting of their classifier.1° The effect size effectively gives a normalized measure for the difference of two means on a standardized scale. The effect size is measured based on Cohen’s d!° as given by:
Table 4. Confusion matrix for poem-—prose disam- biguation. The gray-filled cells (either poem vs poem or prose vs prose) represent the classification rates for a particular case (poem or prose). The other cells are the false negatives/positives of the classifier. the effect size shown in column 2. This can be utilized to determine which of these parameters can best disambiguate the two classes.
Table 4. Confusion matrix for poem-—prose disam- biguation. The gray-filled cells (either poem vs poem or prose vs prose) represent the classification rates for a particular case (poem or prose). The other cells are the false negatives/positives of the classifier. the effect size shown in column 2. This can be utilized to determine which of these parameters can best disambiguate the two classes.
Table 5. Classification rate for different word count. 150 words is needed to successfully disambiguate the two literary forms above the proportional chance criterion.
Table 5. Classification rate for different word count. 150 words is needed to successfully disambiguate the two literary forms above the proportional chance criterion.
Fig. 3. (Color online) Threshold plots of |AC| for different window sizes of the same article as in Fig. 2. Window sizes shorter than 75 have many false positives.
Fig. 3. (Color online) Threshold plots of |AC| for different window sizes of the same article as in Fig. 2. Window sizes shorter than 75 have many false positives.
Table 6. Success rate in detect- ing the text boundary for different window sizes.
Table 6. Success rate in detect- ing the text boundary for different window sizes.
Related topics:
Language networksGenre detection
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved