Figure 1
Processing steps of the approach, illustrating the parallel analysis of text and audio features.
Table 1
Overview of the resulting topics found within the corpus of metal lyrics () and their correlation to hardness obtained from the audio signal (see Section 4.2), *: : (Bonferroni-corrected significance level). Terms are presented in their stemmed format (e.g., “fli” instead of “fly” or “flying”).
| Topic | Interpretation | Salient Terms (Top 10) | Hardness |
|---|---|---|---|
| 1 | personal life | know, never, time, see, way, take, life, feel, make, say | −0.238** |
| 2 | sorrow & weltschmerz8 | life, soul, pain, fear, mind, eye, lie, insid, lost, end | −0.005 |
| 3 | night | dark, light, night, sky, sun, shadow, star, black, moon, cold | −0.114* |
| 4 | love & romance | night, eye, love, like, heart, feel, hand, run, see, come | −0.260** |
| 5 | religion & satanism | God, hell, burn, evil, soul, Lord, blood, death, satan, demon | 0.144** |
| 6 | battle | fight, metal, fire, stand, power, battle, steel, sword, burn, march | 0.129* |
| 7 | brutal death | blood, death, dead, flesh, bodi, bone, skin, cut, rot, rip | 0.292** |
| 8 | vulgarity | fuck, yeah, gon, like, shit, little, head, girl, babi, hey | 0.054 |
| 9 | archaisms & occultism | shall, upon, thi, flesh, thee, behold, forth, death, serpent, thou | 0.144** |
| 10 | epic tale | world, time, day, new, end, life, year, live, last, earth | 0.009 |
| 11 | landscape & journey | land, wind, fli, water, came, sky, river, high, ride, mountain | −0.113** |
| 12 | struggle for freedom | control, power, freedom, law, nation, rule, system, work, peopl, slave | 0.090* |
| 13 | metaphysics | form, space, exist, beyond, within, knowledg, shape, mind, circl, sorc | 0.057 |
| 14 | domestic violence | kill, mother, children, pay, child, live, anoth, father, name, innoc | 0.067 |
| 15 | dystopia | hum, race, disease, breed, destruct, machin, mass, seed, destroy, earth | 0.245** |
| 16 | mourning rituals | ash, word, dust, stone, speak, weep, smoke, breath, tongu, funer | 0.043 |
| 17 | (psychological) madness | mind, twist, brain, mad, self, half, mental, terror, urg, obsess | 0.103* |
| 18 | royal feast | king, rain, drink, fall, crown, sun, rise, bear, wine, color | −0.028 |
| 19 | rock ’n’ roll lifestyle | rock, roll, train, addict, explod, wreck, shock, chip, leagu, raw | 0.042 |
| 20 | disgusting things | anim, weed, ill, fed, maggot, origin, worm, incest, object, thief | 0.087 |
Table 2
Inter-rater reliability in terms of mean pairwise correlation between raters ( in each experiment), Cronbach’s , and Krippendorff’s .
| Mean pairwise correlation | Cronbach’s | Krippendorff’s | |
|---|---|---|---|
| Study 1 | 0.772 | 0.992 | 0.698 |
| Study 2 | 0.665 | 0.987 | 0.590 |
Table 3
Audio features included in the linear model for hardness, along with their unstandardized and standardized coefficients and p values. Sorted by absolute standardized coefficient descending.
| Audio feature | (std) | p | |
|---|---|---|---|
| −0.190 | −0.250 | < 0.001 | |
| HPCP Entropy | 0.741 | 0.196 | 0.003 |
| Spectral Complexity | 0.031 | 0.192 | 0.002 |
| Dissonance | 13.856 | 0.163 | 0.026 |
| Pitch Confidence | −8.943 | −0.162 | 0.002 |
| Diatonic Strength | −2.439 | −0.159 | < 0.001 |
| Scale | −0.202 | −0.112 | 0.003 |
Table 4
Performance measures of the audio hardness model on the training and test sets.
| Training | Test | |
|---|---|---|
| 0.832 | 0.822 | |
| (adjusted) | 0.823 | 0.802 |
| MSE | 0.376 | 0.387 |
| MAE | 0.488 | 0.486 |
Figure 2
Scatterplot showing the comparison between the mean hardness ratings given by human participants on the x-axis and the results of the audio feature–based hardness model on the y-axis (Pearson’s for the training set and for the test set).
Figure 3
Comparison of the topic distributions for all included albums by the bands Manowar and Cannibal Corpse, showing a prevalence of the topics “battle” and “brutal death”, respectively.
Figure 4
Topic configuration obtained via multidimensional scaling. The radius of the circles is proportional to the percentage of tokens covered by the topics (topic weight). The x-axis has been termed PC1—personal/impersonal, and the y-axis has been termed PC2—profane/transcendent.
Figure 5
Correlations between lyrical topics and the audio hardness dimension; : : (Bonferroni-corrected significance level).
Figure 6
Heatmap of topic prevalence by genre, z-standardized. Subgenres and topics are sorted by established relation to audio hardness in decreasing order.