1. Research question

Libretto

• Working definition: modern dramatic texts where music plays a central role

• Born in the early 17th century in Italy and rapidly exported across Europe

• Traditional scholarship focused more on music than on words

• Even librettology is still largely non-computational

Some questions

• Is it possible to consider libretti a unitary genre with its own structural features?

• Do libretti possess a peculiar "genre signal" which sets them apart from contemporary comedies and tragedies?

• How did the structure of libretti evolve compared to the other genres?

2. Corpus

Starting point: DraCor corpora

(☞ Fischer et al. 2017, dracor.org)

Initial survey and preprocessing

1. Check how many texts in our corpora were marked as  libretto (55 🇩🇪 /58 🇫🇷)
2. Disambiguate 🇫🇷 multi-label (= libretto and something else) plays → keeping only libretto as tag
3. Our initial hypothesis: the intended usage of a libretto is more distinctive than its generic alignment

Corpus enrichment

Enrichment results

🇩🇪

+ 51%

🇫🇷

+ 55%

3. Experiments

Exploratory data analysis as a methodological choice

• No strong hypothesis on how the structure of a libretto would have looked like
• "Let data speak by themselves"

A quite simple pipeline

Vectorisation of plays according to structural features

(cf. Szemes and Vida 2022)

EDA on different textual aspects

Vectorisation

1. Get almost all numeric features from the metadata tables (via DraCor API)
2. Create vectors for each play

num_of_segments, num_of_speakers,

num_of_person_groups, word_count_sp,

word_count_stage, average_degree, density,

average_clustering, max_degree,

num_of_connected_components,

diameter, average_path_length

A combination of network measures and size statistics

Experiment #1

Naïve visualisation

Procedure

1. Split the corpora into roughly 50-year spans to follow closely the genre's evolution
2. Apply dimensionality reduction methods (PCA) to the vectorised plays and plot them

3. Results were unsatisfying: no meaningful clustering, no signs of libretto being a unitary genre

Refining our operationalisation

One interesting example:

the 🇫🇷 1670-1719 timeframe

comic space

tragic zone

non-comic libretti

Experiment #2

Finding relevant features

measuring statistical significance of feature variation

• Shapiro-Wilk test
• Wilcoxon Rank-Sum test

training a Random Forest Classifier

• 5-fold cross validation on all data
• Iterative selection of the best n estimators parameter (10-1000)
• Removed highly correlated values

Pipeline

single out the most significant features for further inspection

Most discriminative features for libretti

1. word_count_stage
2. word_count_sp
3. num_connected_components
4. density
5. num_of_speakers
6. diameter

             🇫🇷                               🇩🇪

1. word_count_sp
2. num_of_person_groups
3. average_degree

Experiment #3

Plotting the evolution

of interesting features

Charting the most interpretable features as scatterplots

• four-class implementation

• plotting each play individually

• LOWESS-based smoothing curves to make trends visible

4. Findings and discussion

1. Libretti as a whole show some distinctive traits

Libretti have consistently less spoken text (above) and more stage directions (below)

🇫🇷

2. Non-comic libretti often follow an independent path

🇩🇪  num_of_person_groups / word_count_sp

🇫🇷 density / num_speakers

3. The French dramatic space is more formalised than the German one

• Looking at PCA clusterings, it seems slightly easier to discriminate between different genres in 🇫🇷
• Historical reasons:
  • 🇫🇷 corpus starts from an "age of normative aesthetics" (Boileau, d'Aubignac)
  • 🇩🇪 corpus starts from a later age where deconstruction of French classical models was underway (Lessing) → more formal freedom?

The two types of French libretti (blue) are structurally more distinct than the German ones (orange)

Limitations

• Corpora extension and markup quality
• Comparative approach: lack of 🇮🇹
• Difficulties in modelling relations between dramatic texts on the basis of formal features
• Could we do better?

Comparison: topic modelling (Schöch 2017)

• Individual structural features might be useful for distinguishing one (sub-)genre from the other

• However, it is generally not easy to distinguish between plays formalised as vectors of multiple features

• Drama often seems too homogenous, in terms of structural properties, for discriminative clustering

• Need to find better features (or construct better measures) // rethink operationalisation patterns

In lieu of a conclusion

Thanks for listening!

Questions?

🔄 plu.sh/libretti