From Romance to Mysterioso: The Population of a Taxonomy of Topoi in the Eyl Collection of Silent Film Music / De la romance au Mysterioso : la population d'une taxonomie de Topoi dans la collection Eyl de musique de films muets

Abstract

This article reports the study of the population of a taxonomy of silent film music terms compared to a population of silent film music cues. The purpose of this research is to contribute to the ongoing project stream digitizing large databases of silent film cues. The three phases of research were: (1) Erno Rapée's Encyclopedia of Music for Pictures was converted to the form of a taxonomy; (2) the musical topoi in the catalogue of Ido Eyl's collection of silent film music were similarly compiled and analysed; and (3) both sources were compared to narrate the population of cues based on the taxonomy.

Résumé

Cet article présente l'étude de la population d'une taxonomie de termes de musique de film muet comparée à une population de signaux de musique de film muet. Le but de cette recherche est de contribuer au processus de numérisation en cours des grandes bases de données de signaux de films muets. Les trois phases de la recherche ont été les suivantes : (1) l'Encyclopédie Erno Rapée de la musique de films (Rapée Encyclopedia of music for films) a été convertie en une taxonomie; (2) les topoi musicaux dans le catalogue de la collection Eyl de musique de film muet ont été compilés et analysés de la même manière; et (3) les deux sources ont été comparées pour décrire la population de signaux basée sur la taxonomie.

Taxonomy as knowledge organization system

Knowledge organization systems (KOSs) often engender a multiplicity of uses despite the circumstances of their domain-centric origins. This is particularly the case with the tools known today as taxonomies, which represent limited and tightly constrained conceptual lists representing the most essential elements of a specific knowledge base in a domain. Taxonomies might be designed to classify the contents of observations in a research domain, or they might be designed to facilitate the activities of a group in a work domain. Work-based taxonomies designed to support productive activity will demonstrate both specific actions and the circumstances surrounding their origin and use. At a basic level, we can say that taxonomies contain functional vocabulary and display relationships among concepts to facilitate the goals of the domain. Once made available, a particular taxonomy might be appropriated for other uses, particularly if it has been made publicly available through publication or some other means. In knowledge organization, a taxonomy is described as "an ordered list of terms together with their definitions or other determinant characteristics. Taxonomy is a way of defining the component entities in a domain" (Smiraglia 2014a, 51). A taxonomy designed for categorizing research observations might be pressed into service as a working vocabulary or vice versa. This flexibility is one of the advantages of the taxonomy as a working KOS. Five studies address KOSs in such diverse working environments as media design (Srinivasan 2005), human relations (Marchese 2012), traditional music (Weissenberger 2015), personal information management (McKenzie and Davies 2015), and occupational classification (Hourihan 2017). A recent example of a research-based taxonomy designed for use in a work-based domain is Wu (2016), which was designed for the very critical work of coping with the Deepwater Horizon oil spill.

There have been some attempts to assess the degree to which any particular KOS represents a specific domain's concepts (e.g., Hjørland 2002; Tennis 2006, 2007), but few studies have been undertaken to ascertain how well a KOS's individual concepts are populated by real target objects. A related research stream (Salah et al. 2012; Scharnhorst and Smiraglia 2012; Smiraglia et al. 2013; Smiraglia 2013, 2014b) analysed the evolution and population over time of the Universal Decimal Classification (UDC). The importance of this technique lies in its ability to reveal the degree to which any particular KOS truly represents the conceptual content of the domain it was designed to index. In other words, over-population in segments of a KOS would tend to indicate conceptual areas where greater granularity was needed, and underpopulated segments point to areas that might usefully be compressed or deleted. This article contains a report of domain analytical study of the population of a taxonomy of silent film music cues compared to the population of cues from a specific silent film music collection. Preliminary reports of this research have appeared in Smiraglia (2015a) and Smiraglia and Henry (2016a, 2016b).

Topoi as taxonomic objects

In the early days of film, silent films were screened in theatres equipped with pianos or organs—in fact, these performances often took place in the same venues used for roving stage shows. The musical cues used by musicians to create individual accompaniments for silent films in performance came from a list of musical terms originally settled as the vocabulary of vaudeville. Recent attempts to demonstrate the connection are discussed by Rosar (2012) and Plebuch (2012). Rosar (2012, 209) described the manner in which music was used by accompanists as a form of audio illustration for silent film, and in order to facilitate performance, "much generic 'stock music' was written … and the agitato, misterioso, and appasionato virtually defined the musical idiom." A repertoire of short pieces, mostly lasting only a minute or two and matching moods represented in the visual imagery of the film, were called "cues." The titling practice evolved over time into a form of classification in film studio music libraries, where librarians assigned generic mood musical terms to cues in newly composed scores (210). The music of the silent film era is often referred to, in English, as "photoplay" or, in German, as "Kinothek" (a contraction of Kinobi-bliothek) (210). The most prominent compendia of the repertoire were those by Rapée ([1925] 1970) and Erdmann, Becce, and Brav (1927).

Lazarus (2012) and Wilson (2012) have described the actual musical practice of the musician. Lazarus (2012, 36) shows a contemporaneous cue sheet with two examples of "hurry music": struggles and duels. Where improvisation was employed, it was strictly controlled by professional practice and the concept of "respect" for both the film and the musical illustrations utilized (Wilson 2012). Wilson (2012) shows a "cue sheet" for use with Douglas Fairbanks's movie The Thief of Baghdad. There, one can see musical incipits together with appropriate musical references alongside the actual cues "Action. Thief grabs man at well … Overture Comique … Keler Bela … 1 ½ Min." Examples provided by Beckerman and Rosar (2009) and Brooks (2012) demonstrate the practice of Kinothek in silent westerns and in linked performances of Sarah Bernhardt, respectively.

As noted, the practice grew from nineteenth-century stage melodrama, described fully by Plebuch (2012, 77); musical illustration was "largely based on stereotypical roles, generic plots, standardized gestures, and familiar locations." Plebuch related this practice to the use of what he called "musical topoi" or characteristic musical expressions that are habitual and symbolic and, therefore, easily recognizable across the profession of stage musicians. Still, topoi were sufficiently indicative that the musicians retained necessary flexibility to interpret or improvise as the situation demanded. In this sense, any list (or "taxonomy") of topoi has a kind of inherent duality to the extent that the terms are easily recognizable, but their use will vary from performance to performance and from musician to musician. Thus, the surviving "lists" are in the form of handbooks from the period in which, alongside the topoi, musicians recorded standard musical citations in the form of brief composer-title statements to indicate which bits of which compositions might most effectively correlate with the topoi. Further, there has been little research to date based on the accuracy with which the lists of cues that were actually used correlate with the topoi. Manuals from the era are available but have not been systematically studied from a domain analytical point of view to correlate terminology with actual usage.

Testing the taxonomy underlying Rapée's Encyclopedia: methodology

This article reports a study designed to inform both the analysis of film music topoi, for which information systems are being developed, and the population of work-based KOS. The central research questions, then, were:

1. 1. Can the film music cues in Rapée's ([1925] 1970) Encyclopaedia of Music for Pictures (Rapée's Encyclopedia), a working KOS, be seen to constitute a simple taxonomy?

2. 2. To what extent do the cues in the emergent taxonomy correlate with topoi used in an actual archives of silent film music practice?

In this study, which comprised three phases of research, each phase is described in detail in the sections of the article that follow. First, the Rapée Encyclopedia was digitized and then converted to the form of a taxonomy for analysis. Second, a collection of musical topoi used for the illustrations of specific silent films was similarly digitized, compiled, and analysed. Third, terms in both sources were compared to discover the extent to which the population of the taxonomy overlaps the population of topoi in recorded use. This analysis provides fodder for analytical narrative concerning the population of cues based on the taxonomy generated in the first phase. The purpose of this research is to contribute to the ongoing project stream digitizing large databases of silent film cues. This project adds understanding from the knowledge organization domain that is directly relevant to the degree of concordance between taxonomies derived from working cue handbooks and the population of actual cues used in silent film. Furthermore, this research contributes to the domain analytical research stream in knowledge organization that seeks to analyse both the population of KOSs (Smiraglia 2016) and the form and structure of occupation-based taxonomies (Hourihan 2017).

Rapée's Encyclopedia becomes a taxonomy

The goal of the first phase of this research was to turn Rapée's Encyclopedia into a workable term list, from which it could then be reformatted as a working taxonomy. Despite its grand encyclopaedic title, the volume is essentially a working musician's handbook, designed to sit on the music rack (or beside one on the bench) to work both as a guide to cues appropriate to specific topoi and also to serve as a repository for notes and locally generated citations to appropriate cues. It also is important to keep in mind that Rapée ([1925] 1970) used the volume to instruct new theatre musicians, so that there is a pedagogical current underlying the music recommended for specific cues. The original edition, of course, was developed in book form. The main part of the volume features cues centred in bold type, accompanied by lists of musical works by title, composer, and publisher, all in shortened form. References—both "see" and "also see"—occur as Rapée found it appropriate, often in formulations that seem peculiar today. Figure 1 provides a reproduction of page 51 of the book.

The functional aspects of the volume are visible; in the upper right corner are the first and last terms from the page, a sort of summary designed to facilitate

Figure 1.

Alabama to American from page 51 of Rapée's Encyclopedia

Source: Rapée (1925) 1970, 51

flipping quickly to a desired term. Recommended musical works (cues) are listed under each term with the dotted lines for local check marks and notations and blank lines for musicians to record their own cues for locally preferred musical works. Here, "Alpine" is not a preferred term, so the user is directed to "Austrian," "Swiss," and "Yodel."

After digitizing the volume, we transcribed all of the terminology into a spreadsheet so as to track simultaneously the preferred, non-preferred, and related terms. From this analysis, we were able to visualize aspects of the social reality of motion pictures and motion picture music from the early twentieth century. In addition to copious musical terminology, we were able to observe six topical groupings—creatures ("bees," "birds," etc.); places ("beach," "coal mine," etc.); moods ("fantastic," "gruesome," etc.); peoples ("pirates," "fireman," etc.); instruments ("clocks," "music boxes," etc.); and entities ("clowns," "ghosts," etc.). Thirteen specific states of the United States were included. Subgroupings of musical terms for dances and songs also were observed. These results are reported fully in Smiraglia and Henry (2016a). We found curious cultural points of view, starting with "hobos" and "rubes" and including the gathering of anything Latin American under the cue "Spanish," and the definition of "Oriental" as comprising "Armenian," "Persian," and "Arabian," among other Middle Eastern destinations, alongside "Desert Music." Asian cultural references point to the curious "Chinese-Japanese," which is not part of the Orient of Rapée's time. The only specifically African cues are for "Abyssinian," "Liberia," and "Zanzibar," while "African" points to "cannibal." Scandinavia is "northern," although "northern-concert" can be subdivided by "Daenish" [sic], "Finnish," "Norwegian," and "Swedish." Even the United States is represented in a granular and incomplete manner with "Illinois," "Indiana," "Mississippi," and "New York" all specified, reflecting no doubt the locales of films of the period.

Rapée ([1925] 1970) also used a form of facets both explicitly and implicitly to express mood or performance difficulty, suggesting familiarity with film music library practice of his day (Beynon 1921). In knowledge organization, facets have been described variously as "component parts … [that] completely describe the domain" and also serve as "the faces—the presentation characteristics … of a phenomenon" (Smiraglia 2014a, 44, 82). A full analysis of the faceted aspects of Rapée's Encyclopedia appears in Smiraglia and Henry (2016b). Mood is conveyed under certain cues (e.g., agitatos, love themes, mysteriosos, etc.) as "heavy (H)," "medium (M)," "light (L)," "dramatic (DR)," and "dramatic-neutral (D)." Implicit facets are conveyed as qualified terms (e.g., "andante [happy]," "robbery [serious]"), although the lead terms do not appear in unqualified form. Clarifying qualifiers "neutral" and "light" appear with certain cues as well. The cue "Suites" is qualified with terms for countries and people, conforming to the idea of facets for space and time. We can make an analogy between Rapée's ([1925] 1970) partial use of facets in highly individualized locations and the use of faceted auxiliaries designed for specific concepts in the UDC, which is familiar to the knowledge organization community (McIlwaine 2007).

It is important, of course, to remember that Rapée ([1925] 1970) was not developing a taxonomy but, rather, was creating a commercially viable (i.e., publishable) working tool for film musicians and pedagogical tool for students of film music accompaniment. The published list can be seen as a base-line visualization of music for film of the time, particularly from the North American marketplace. But the mechanics were not exactly in line with standard taxonomic structure. For example, Rapée used only two kinds of references—"see" and "also see." The "see" references accord with standard "used for (UF)" references. But his "see also" references were used for "broader term (BT)," "narrow term (NT)," and "related term (RT)" references, without designation. A final step in this phase of research was to coordinate the terms taxonomically, using standard definitions of narrower and broader terms alongside Rapée's "also see" references. The resulting taxonomy has 346 lead terms ranging from "aeroplane" to "zoo." Appendix 1 shows the arrangement in the final taxonomy for a smattering of terms.

Figure 2 shows how the terms "fox-trots" and "mysterioso" appear in Rapée's Encyclopedia. "Fox-trots" appears without suggested cues but is followed by six pages of blank lines (not shown) for musicians to use to record locally preferred musical works—a clue to the ubiquity of the foxtrot. The entry for "mys-terioso" demonstrates how facets function in the Encyclopedia, with the mood designations appearing just to the left of the cue titles.

Extracting cues from the Eyl collection

The second phase of this research involved analysis of actual cues used in silent films of the era. Our source was the catalogue of the silent film collection of the EYE Film Institute in Amsterdam, Netherlands. The EYE Film Institute preserves Dutch films and also foreign films that have been screened in the Netherlands, dating from 1895. A 1992 catalogue of silent film music by van Houten (1992, 51) gives details of the musical resources collected by Ido Eyl, who built the working collection of the Utrecht Rembrandt Theater. The catalogue contains entries for 3,235 musical works from the Eyl collection. An initial report of this phase of research appeared in Smiraglia (2015a). The catalogue was digitized, and its contents transported to a spreadsheet containing composer, title, arranger, medium of performance, publication details, and various notes associating each cue with the film it accompanied. The title data were entered into the ProSuite™ from Provalis Research. The WordStat™ module can then be used to generate frequency distributions of keywords and multi-word phrases, which can be combined and sorted to yield a basic taxonomy. (This technique, frequently employed in domain analysis for knowledge organization, is described in full in Smiraglia 2015b.) In this case, terms used to identify the musical works were sorted by frequency of occurrence, which generated a list of 1,792 words. The top end of the distribution included 15 keywords that each occurred 40 times or more (see table 1).

Linguistic variants are common in the dataset (e.g., fantaisie, fantasie). Each keyword can be considered a potential anchor term in a set of clusters, which are

Figure 2.

"Fox-trots" and "mysteriosos" from pages 206 and 351 of Rapée's Encyclopedia

Source: Rapée (1925) 1970, 206, 351

Table 1.

Top tier of keyword frequency distribution

Table 2.

Clustered terms associated with "Romance"

musically related by reference to the specific keyword. For example, table 2 shows the set of such clusters associated with the keyword "Romance."

Six specific phrases occur adjacent to "Romance," some such as "Celebre romance" or "Petite romance" are potential topoi. We also can follow this analysis back to the catalogue to see how specific topoi might be associated with particular film cues, such as "Romance à l'etoile du soir," which is used for the end of the second and third scenes in the associated film. Table 3 shows an excerpt from a more complex example in which different linguistic variants of the keyword "waltz" are collocated and sorted by frequency of occurrence. Sixty-four separate iterations of "waltz" occur, the most frequent is "walzer," which occurs 91 times in the collection.

These simple illustrations demonstrate the complexity of the task of term extraction as key musical terms, the potential topoi, can occur in a variety of multi-word terms.

Wordstat™ also has a phrase-finder feature that extracts phrases of any length specified. By using this feature, 5,414 two-to-five word phrases were extracted and arrayed by frequency of occurrence, yielding 117 unique phrases in the dataset. Table 4 shows the top tier of this frequency distribution (phrases occurring 10 times or more) as an example, and it has been edited to remove phrases such as "aus der" that occur within other phrases (e.g., Fantasie aus der Oper).

Table 3.

Excerpt from collocated "waltz" clusters

Table 4.

Top tier of phrases from Eyl cues

This example is taken directly from the frequency distribution, thus no form of vocabulary control has been applied. As a result, we can see how core terms like "potpourri" or "valse" could form clusters with various modifiers, as we saw above in table 2. Ultimately, the term and phrase lists were sorted and deduplicated to yield a working taxonomy of 353 terms.

Comparing cues to topoi

The third and final phases of the project was to compare the two working taxonomies to discover the extent to which the actual cues found in the Eyl collection match those in the taxonomy developed from Rapée's Encyclopedia. Prior research summarized in Scharnhorst et al. (2016) focused on analysing the population of a KOS has involved classifications. Specifically, the population of the UDC was analysed by comparing UDC strings derived from several bibliographic sources to the compiled classification itself. This was accomplished by matching the UDC's nine main classes or its six common auxiliaries (language, form, place, ethnic grouping, time, and properties/materials/relations/persons). An extension of the study compared the topical content of Wikipedia at a specific point in time to the UDC main classes and first-level divisions (e.g., 5xx mathematics-natural sciences, 55x earth sciences-geological sciences). For this study, all terms in the taxonomy derived from Rapée's Encyclopedia were sorted into a single alphabetical list but coded by source (lead term or facet). This list was sorted together with the working taxonomy compiled from the cues in the Eyl collection. Linguistic variants were collocated (e.g., dancetanz, valse-waltz, and even love-liebe). The overlap between the two represents the population of the Rapée taxonomy by actual cues used and recorded in the Eyl collection.

There were 354 terms in the working taxonomy from the Eyl collection, and 540 terms in the working taxonomy from the Rapée Encyclopedia. There were 87 terms in the Rapée taxonomy that were populated by cues from the Eyl collection. Quantitative measures are of arguable use in this situation; the term lists used to sort were of different types. Although it seems that a small proportion of the terms in the Eyl collection matched terms in the Rapée taxonomy, there were many uncoordinated clusters in the term list from Eyl (e.g., "air" and "airs," six different terms for "American," five forms of "tragic," etc.) From the coordinated Rapée taxonomy, 16% of the terms were populated by cues in the Eyl collection. Of those, 31 terms were associated with various facets having to do with the level of difficulty (easy-difficult), intensity (light-heavy), or ethnicity (English, French, Swedish, etc.). Ten terms in the Eyl collection matched terms from those facets; "heavy" and "light" matched the intensity facet; all of the rest matched the ethnicity facet ("American," "Chinese," "French," "Indian," Italian, Japanese, Russian, and Spanish).

It is interesting to see the geo-political-ethnic emphasis of the films in the Eyl collection, which we must remember represented performances at the Utrecht Rembrandt Theater. In other words, we see that the theatregoers of Utrecht were likely to have encountered musical illustrations of neighbouring European cultures (e.g., English, French, Italian) alongside a limited scope of more exotic culture (e.g., Chinese, Indian, Japanese). The musicians at Utrecht, not surprisingly, were quite experienced so we find no level of difficulty facet represented. Searching the topical divisions, we observed no usage of terms from categories "creatures," "instruments," or "entities." One term was used from the category "place" ("western"), and one from "people" ("cowboy"). Twelve of the 25 terms from the "moods" category were used ("dramatic," "emotional," happy, joyous, love, melancoly, agitato, pathetique, sinister, tragic, caprice, hurry).

The terms from the Eyl collection not found in the Rapée Encyclopedia taxonomy were sorted and regularized linguistically leaving 126 unmatched terms. These terms were of the types discussed above—places that are both specific sites and locales ("cabaret," "Naples," "Wien,"); people ("Ziguener," "peasants"); moods ("wild," "violent," "uncanny"), and several that could be characterized as events ("fights," "convalescence," "riots," "stampedes"). There were also several musical terms ("scherzino," "paso doble," "rigaudon," etc.) and the dance form "shimmy." Finally, there were a number of uniquely German terms ("Rheinlie-der," "Standchen," "Volkslied").

Discussion and conclusions

The purpose of this research was to contribute to the ongoing project stream that is digitizing large databases of silent film cues. We approached this purpose by using methods from the knowledge organization domain that are directly relevant to the degree of concordance between taxonomies derived from working cue handbooks and the population of actual cues used in one collection of silent films. Obvious limitations of this study are the limitation to one working handbook (Rapée's Encyclopedia) and one silent film collection (Eyl). On the other hand, if we consider this as a case study, we can see that it has been revelatory in several ways. We were able to plumb the catalogue of the Eyl collection to reveal a set of terms matching cues the use of which were recorded in a theatre of the period. In that sense, our data are grounded in actual historical practice. We were able to convert the terminology in the Rapée Encyclopedia into a working taxonomy that we could then use for this research by matching terms from the taxonomy to terms taken from the used cues. Both results represent significant contributions to the research. In particular, both phases of this research helped create visualizations of the way in which musical terms were used as cues to illustrate silent film and, in turn, the ways in which topoi can be used to form working taxonomic KOSs. In addition, the analysis of the early twentieth-century cues reveals multiple examples of shifting cultural values. In this, our work adds to the stream of historical analysis of KOSs such as studies of eugenics (e.g., Tennis 2012) or sex and gender (e.g., Fox 2015).

Furthermore, we sought to contribute to the domain analytical research stream in knowledge organization that seeks to analyse both the population of KOSs and the form and structure of occupation-based taxonomies. We were able to match a substantial portion of the cues from the Eyl collection to the terms in the Rapée taxonomy. This demonstrates both the actual utility of the taxonomy as a working musician's tool and its capability to serve as one of several KOSs applicable to the digitization of music from the Kinothek period. We also were able to see the difference between the cues in the Eyl collection and the Rapée taxonomy. The Eyl cues contained many Eurocentric terms that did not appear in Rapée's Encyclopedia list, which points, in turn, to the North American emphasis of that list. It is possible that the same research conducted with a taxonomy derived from Erdmann, Becce, and Brav (1927) would match a larger proportion of the terms found in the Eyl collection. That said, we were able to demonstrate how terms might have been added to the existing structure of the Rapée taxonomy—specifically, terms representing moods, locales, and entities. We also were able to observe an additional facet in the Eyl cues—for events—that does not emerge from the Rapée Encyclopedia list.

Further case studies using different collections, musician-annotated handbooks, or combinations of them would yield a more complete picture of KOSs for silent film music. Such a picture might reveal consistencies in practice across time and geographic region as well as whether there were more formal knowledge organization structures, such as explicitly faceted systems. We have approached this research from the standpoint of knowledge organization and the population of KOSs. But understanding silent film music KOSs can also lay the foundation, or at least provide a point of departure, for understanding KOSs for sound film music because just as silent film music borrowed conventions from theatre, so too did sound film music utilize practices from silent film music. This stream of research can ultimately also inform general film music scholarship.