Documentation – Overview
- Overview of the Work Packages
- Contents – Items – Data: Basic Conceptualization of the Work Packages
Overview of the Work Packages
The following overview shows the original concept at the project’s outset. Given in brackets are the respective project partners angegeben (Ffm = Musicological Institute, Goethe University, Frankfurt/Main; ER = International Audio Laboratories Erlangen; DT/PB = Musicological Seminar, Detmold/Paderborn and Department of Computer Science of the Paderborn University):
- WP 1: Building the Data Base
- WP 1.1: Adapting and integrating the Freischütz data (DT)
- WP 1.2: Digitalizing the audio recordings (DT, ER)
- WP 1.3: Encoding the music texts (DT)
- WP 1.4: Linking Edirom tools to the TextGrid repository (PB)
- WP 2: Preparing Acoustic Components
- WP 2.1: Synchronizing encoding and audio recordings (ER)
- WP 2.2: Recording selected numbers (DT)
- WP 2.3: Segmenting audio recordings based on temporal/spectral synchronization (ER)
- WP 2.4: Integrating audio elements in the presentation software (PB)
- WP 3. Freischütz Edition
- WP 3.1: Modeling the digital edition (DT/PB)
- WP 3.2: Developing the digital edition (DT)
- WP 3.3: Linking score and libretto (Viglianti)
- WP 3.4: AAdapting and enhancing the software (PB)
- WP 4: Libretto Edition and Intertexts
- WP 4.1: Genetic text edition of the libretto (Ffm)
- WP 4.2: Intertexts I: Reference points for the libretto (Ffm)
- WP 4.3: Intertexts II: Reception of the work (Ffm)
- WP 4.4: Adapting and enhancing the software (PB)
- WP 5: Integrating User Interactions
- WP 6: Demonstrating Variance as an Example for Subsequent Research Issues
- WP 6.1: Handling musical notation variance (DT)
- WP 6.2: Comparing interpretations based on performance materials (Ffm)
- WP 6.3: Comparing interpretations of textual and acoustic variance (Ffm/DT/ER)
- WP 6.4: Implementing modular research results technically (PB)
1. General Work-Package Aspects
How does the potential of digital media change our concepts of music editions? Given current technology, what could such an edition look like if planned as digital from the outset? What are the changes in workflows? What should be taken into account when planning such projects?
The Freischütz Digital project came about against the backdrop of such questions, but was based from the outset on a model of the digital music edition already outlined and published by Frans Wiering in 2009. His design of a Multidimensional Model replaces the traditional concept of an edition and translates it into a kind of “archive” of the most diverse data, correlating not only music texts and facsimiles, but other texts and annotations, audio and video sequences, together with forms of encoding for text or music, to allow the user individual access within this virtual space. Going along with such a model involves many aspects that are (of necessity) left out of traditional editions.
Carl Maria von Weber’s successful opera Der Freischütz offered the ideal vehicle to demonstrate the differences between traditional editions and future digital prospects, for it includes an abundance and diversity of items (music and verbal texts, and audio) extant for this central work in 19th-century opera history. In addition, the concurrent development of the work’s “analog” edition within the framework of the Weber Gesamtausgabe [Complete Edition] (WeGA) fortuitously enabled both enterprises to build on the same source base and to compare their different objectives more specifically. Furthermore, WeGA’s already existing digital sections in the textual area (letters, diaries, writings, performance reviews, cf. www.weber-gesamtausgabe.de) ensured that an unusual range of the work’s context could be included without considerable additional expense. Finally, the development of the Edirom software (originally coming from WeGA work and further developed within the framework of a multi-year DFG-Project) first established the prerequisites for implementing the ideas to be developed in the project.
In view of the Multidimensional Model, it was clear from the outset that work in the project should be considered from different perspectives and not just the traditional editorial ones. The project partners involved therefore introduced their own specific focus areas into the work:
- At the University of Frankfurt (in its initial phase, the University of Bayreuth), the focus was on questions about the libretto, its genesis, transmittal, as well as related reference texts.
- The project partners at the International Audio Laboratories Erlangen (a joint institute of the Fraunhofer Institute and the University of Erlangen) dealt with the processing of different audio aspects.
- At the Musicological Seminar in Detmold/Paderborn the music-text encodings were prepared and new forms of "edition" und annotations were tested. Detmold/Paderborn was also in charge of coordinating all the work.
- The University of Paderborn was responsible for questions of software development, user interfaces, and integrating user interactions.
The first attempt at approaching such an extensive work with authentically digital edition concepts led to a multitude of minor and major difficulties, unforeseen, but hardly surprising; given the limited time available, these in turn meant modifying many objectives that had perhaps been formulated all too lightly. The effort and expense that such a pilot project must incur, especially pertaining to innovative and untested handling of data, was considerably greater than expected, particularly in the area of in-depth indexing of music data or combined text and music data. Bearing that in mind, we hope that with the tools developed to aid the user and with existing documentation, a crucial guide can be provided to new editorial enterprises in order to get comparable results more quickly (and to move even further from the traditional book conventions than we succeeded in doing). Thus, from the project’s outset, basically supporting the editorial work and demonstrating new possibilities, as well as developing generic solutions, were more important than completing an “edition” in its own right, which in this respect was used in many places more as an object of demonstration, even though, in the sense of a traditional edition concept, it might occasionally be referred to as a misuse of data. Solutions were developed in this context, particularly for the concept of variance in the area of the music text, which can be more fruitfully applied to other works than in the tradition of the largely stable Freischütz music text (once the secondary parameters of the music text particularly emphasized here were disregarded).
The project’s problem-oriented approach and the changing perspectives on the object of attention – Weber’s opera variously manifested as music facsimile, newly-engraved text, audio-example, etc. – led to the fact that, beyond implementing technical software solutions, many concepts could initially be presented only in the form of so-called demos, and that documenting the developments, but also, in particular, describing the problems that had arisen played an important role. In this respect, the project should be perceived as a kind of “work in progress” based not only on its current (certainly deficient) external manifestation, but above all on the data made available in different areas for further research and on the documented workflows. From the current point of view, Freischütz Digital is “paradigmatic” mainly because established standards are used in all parts of the project, and the community’s results are made freely available to facilitate opening up new supporting or follow-up researches and delivering results, in both senses of the word, inconclusive or even incomplete. The project is in this respect only a small contribution towards new forms of cooperative research.
General Work-Package Aspects in Critical Retrospect
The first consortium meeting in Detmold at the beginning of October 2012 already highlighted the need for a continuous intensive exchange between the participating project partners, which was advanced not only by a total of six consortium meetings, but also by ten further meetings in smaller groups. The need for the most effective possible communication, supported also by joint online project documentation (Redmine), increased significantly in the second half of the project and became even greater when a shared platform for publishing the results was conceived; then collaborators finally established regular two-week Skype sessions often following up the joint large discussion or were even continued meantime in smaller get-togethers. Stated in retrospect is that mutual understanding of the project partners’ questions and problems could only be gradually established and that the complex encoding problems in the text and music fields, together with problems related to presentation, were not from the outset obvious to both sides. A problem generally affecting the digital humanities appeared here: for one thing, methods and ways of thinking, hence learning processes, differ in computer science and the humanities. This fact means that each side must understand the other side’s position and requirements in solving problems in order to dovetail respective solutions towards a common end, in itself, a complex learning process. After a first exchange between the two sectors to advance initially substantially independent developments in the different work packages and to exclude in the encoding, for example, later presentation questions – as actually called for by the philosophy of XML languages –, the concept, given the complexity of the subject matter works only to a very limited extent. Even in the case of a clean content concept in dealing with the encoding of genetic developments in the libretto texts, for instance, the much discussed problems of overlapping hierarchies subsequently produce headaches related to converting the many individual and special cases into a uniform presentation.
Thus, in this case both humanities scholars and computer scientists found it a most valuable experience that from the outset they jointly had to develop the concepts to reflect the respective consequences of decisions on both sides. When adequate encoding of difficult issues implicitly requires a high level of presentation, this can lead to the need to adapt encodings in such a way that compromises the content requirements and their feasibility - even more so for music encoding. It is by no means a matter of reducing the complexity required by the musical object at hand, but the important thing is that detailed encoding guidelines should be defined at an early stage in order to limit the various possibilities of capturing the same facts offered by both MEI and TEI, since this range creates great hurdles for machine processing. Both sides must therefore know each other’s exact requirements, so that from the outset a very close cooperation is recommended. Since, furthermore, in terms of text and music encoding in the FreiDi project, the IT competence was concentrated in one place, the need for an intensive exchange proved all the more important.
From these experiences, new digital edition projects will be urgently advised to plan the IT side as a truly integral part of the editing enterprise and to maintain the constant exchange between the different areas – with no conceptual decision on one side, representatives may miss the other sides. Detailed organized communication and its documentation are essential prerequisites for implementing such a project.
Flexibility in Planning
A second important general point concerns how flexibly such a project should be able to react to new findings or developments. Three issues can explain this:
- Formulated at the outset and regarded as the necessary prerequisite for designing new methods and features, it was argued in envisioning the project that the MEI-encoded notes should be linked with the temporal positions of the corresponding audio data streams in order to achieve a new kind of music-text and audio synchronization (see WP 2.1). It was therefore foreseen that in order to carry out their work, Erlangen's colleagues should adopt the Detmold encodings of the music data; this led to corresponding delays due to the complicated preparation of these codes. However, it turned out that the experiments for temporal segmentation could just as well take place based on MIDI data generated from the graphic notation files not yet transferred to MEI, so that a link to MEI would become necessary only in a second step.
- The greatest flexibility was needed in the case of music encoding. To date, no piece of music of that size had been presented in MEI. The sheer volume of encoding, barely recognizable as humanly readable (the autograph alone runs to about 1.4 million lines of MEI), required new approaches to some form of checking this encoding. Developed for this purpose in the project’s middle phase were special proofreading tools of our own that gave new prospects to the by now almost hopeless attempt to check this flood of data. These revisions were, however, dependent on “visualizing” the encoding, i.e., transforming it into human-readable music notation. To this end in the initial phase, presentation libraries such as abcjs, VexFlow, or MEISE could be tested and included without these procedures actually being satisfactory. When, in the last phase of the project, the Verovio presentation library published by Laurent Pugin showed great promise in implementing MEI data, this library was adopted in the further presentation attempts and integrated into the Edirom display in consultation with Pugin.
It is important that projects such as FreiDi actively pursue the latest developments in the relevant fields and revise technology decisions at an early stagein order to integrate powerful new tools without, on the other hand, prematurely settling on possibly only short-lived developments. This implies that such projects are actively involved in the exchange within the community, which is why the project has made every effort to regularly attend important conferences and to draw attention to its own developments through publications.
If in this respect there were deviations from the originally formulated procedures or objectives, a rigid adherence to the once formulated would have been contrary to the spirit of the project and have contributed to the failure of the overall objectives.
This also applies to objectives in the area of software development. Priority was given to prototypical solutions in line with the project's requirements, although the long-term use on a broader source basis or transferability to other projects were as far as possible to be taken into account. Since, however, identifying new possibilities should have priority in case of doubt, the project team (as mentioned above) decided to cover sub-areas by developing so-called demos. These reveal new functions without presenting them in the form of non-adaptable ready-to-use software components. Important in these cases, however, are the perspectivesthat can be demonstrated by these demos, though having available appropriate developer capacities would be desirable since having such tools practically available can lead to a real change in research methods.
This preface to the documentation cannot be concluded without a warm word of thanks: this project could only have been realized with the support of numerous libraries, archives, and private individuals willing in this project to make their sources available to the scholarly public; likewise important was the generous permission by audio firms and broadcasting institutions to use their audio recordings. We would like to take this opportunity to thank all of you!
2. Contents – Items – Data: Basic Conceptualization of the Work Packages
The focus in the six work packages jointly worked out by the four project partners was the work’s graphic, logical, and acoustic (or performative) domains. Graphic and logical domains concern both the musical and textual portions of the edition. In the following, the individual components of the edition are briefly presented; further explanations can be found in the respective access areas.
2.1 Music Texts
The transmittal of the music portions of the Freischütz, aside from several marginal sketches not considered within this context, begins with Weber’s autograph score and then initially continues with the 41 copies whose production or dispatch the composer documented in his accounts book between November 1820 and April 1823. Not recorded there is, in addition, a copy for the Berlin premiere. Of these altogether 42 copies authorized by Weber, only eight prepared by two of his chief copyists are still known to be extant. All of these sources are accessible number-by-number and bar-by-bar in Edirom online. For an overview, see “Data and Tools” on the menu or the heading “Music Edition: Sources in Edirom online.
There is also a small assortment of several printed sources, starting with the first print of the piano reduction that Weber made, through the posthumously published score edition of 1849 (in several issues), up to the popular Peters print (see the overview as above).
For the coverage of these items in detail, see the section "The Core Concept".
2.1.2 Encoding the Music Texts
A consistent encoding of the music texts (to implement the logical domain) was paramount for linking the various items within the project, particularly so that the real advantages of digital editions could be clarified. This encoding consists of two components: Captured in a formal manner in the so-called Metadata are crucial details of the traditional source description, whereas in the ideal case the encoding of the music texts themselves reproduces a precise image. Created with that are both the prerequisites for comparing sources semi-automatically, as well as also providing, on the other hand, the foundation for documenting these differences in formalized critical-apparatus entries, the prerequisite in turn for presenting variants as variants or as variants within the music text. The concepts of this encoding using the standard Music Encoding Initative (MEI) are exemplified in Documenting the Music Text.
Within the context of the work on the MEI encoding and regarding the necessity of creating, proofreading, and redacting these comprehensive XML data as expeditiously as possible for the most diverse specifications, numerous transformation scenarios had to be conceived, developed, and the results of the created XSLT scripts audited. Developed for this purpose were an impressive series of new tools enabling, for example, the proofreading of pitch and duration, slur placement, dynamics, articulation, etc., and making them available, adapted in form as needed, for other projects. The formidably complex access with the MEI data becomes clear in the section “Werkzeuge,” describing the individual tools in detail.
3. Text Sections - Part A: The Project’s Particular Work Areas
3.1 The Freischütz Libretto
The Freischütz textual sources can be said to offer an ideal source situation: Extant are nearly all the essential sources for documenting the work’s development. There is, in addition, an early libretto copy (from still before the work’s premiere), together with the several revised editions published by the librettist himself. Thus, a major concern of this module was that the genesis of the work’s text and its further development up to the death in 1843 of the librettist Friedrich Kind be made transparent in the digital edition. In this case, prerequisites for this and for a further use of the data created was the encoding of the texts with the XML standard Text Encoding Initiative (TEI).
ZFor an overview of the manuscript and printed libretto sources considered within the edition’s context, see the items “Data” and “Tools” in the menu.
All transcriptions are based on using the TEI encoding standards. Since the manuscript sources reveal many traces of revision, respectively, capturing the addenda and corrigenda within the manuscript directly related to the text’s compositional process was of special interest, and so its own unique presentation form was considered in its presentation (cf. on this the description of the tools, Genetic Text Stages – Text-Layer Display). But to be retained at the same time are also both the differences among the (manuscript) librettos related to the compositional process, as well as, on the other hand, the variants handed down in the copies and printed librettos. In the light of the consequently very complex demands for presenting these different variances related to the work’s genesis and transmittal, the idea of designing a diverse construct – a text core file – comparable to a music edition was obvious. The operating principle of this text core file is fundamentally different, however, from that of the music core file (cf. on this the section Presenting the Text Genesis).
In this edition’s context, however, it would be a hybrid endeavor to make all variants among all sources visible in appropriate synopses (though this would be possible with the available data within the context of a corpus linguistics investigation). For this reason, presenting orthographic variants was omitted (the orthography of the sources is accessible in the diplomatic text transcriptions – with the exception of a few characteristics of historical types of notation) and merely meaningful variants (admittedly defined subjectively) are visualized using the core file. The possibilities of displaying automatable assay procedures have been clarified with individual demos (see Variance Display).
The same goes for the complex linking of score and libretto.
3.2 Reference Texts of the Freischütz Libretto
There have been frequent arguments in the Weber literature about possible precedents to Kind’s libretto, including direct accusations of plagiarism. Undisputed is that the Gespensterbuch (vol. 1, Leipzig, 1810) published by August Apel and Friedrich Laun was the starting point for several dramatizations of the Freischütz material. Freischütz Digital creates the prerequisite for comprehensively examining the question of how possible models are interdependent and exemplarily attempts to correlate not only motivic and historical subject matter, but to display general textual interdependencies. In the process, the evidence of intertextual references is not restricted to the reference points historically preceding this libretto, but includes (at least in some respects) also the libretto’s repercussions. From the abundance of texts in question for both points of view, a number (predominantly in print) were selected and reproduced in transcription and if possible also in facsimile. Included was also Friedrich von Schiller’s Wallenstein trilogy, often referred to both in Kind’s and in several reference texts. The Schiller texts were taken from the TextGrid repository.
The technique of the so-called “Topic Maps” (see on this the explanations below…), with motives or topoi in common visualized through graphics generated ad hoc, was used to enable more quickly grasping and therefore verifying the motivic references among these texts – likewise diplomatically transcribed in TEI Standard. This also allowed the references to be visualized even when the items are differently described verbally or orthographically. This possibility could only be exemplarily modeled here (capturing motives, terms and names), but it should be clear that this is a powerful tool whose importance increases when the texts mentioned under 2.3 are included.
4. Text Sections - Part B: Integration of WeGA’s Digital Text Components
Weber’s letters, diaries, and writings are compiled in a WeGA digital edition, see http://www.weber-gesamtausgabe.de. Documented in these ego documents is obviously also the genesis of Freischütz and its transmission during Weber’s lifetime. Moreover, a wealth of further information is available on this website: thus, biographical data for the network of persons in contact with Weber, also reviews of his published works, and especially reviews of their performances (mainly during the composer’s lifetime); currently there are, in fact, over 300 reviews for Freischütz alone. This data, all of it captured using the TEI encoding guidelines, is available for the project Freischütz Digital , thus considerably expanding the data spectrum. All the items of this website are provided with identification (IDs) for unambiguous referencing.
A separate demo illustrating the links to the WeGA website was omitted due to lack of personnel resources: integrated in the texts for the genesis of the libretto and music are relevant links that can be implemented currently only by opening new windows.
5. Audio Contributions
When the project was conceived, including the audio domain in the work was planned right from the outset, but the by now already over 50 existing recordings of the opera show that this is an ideal field for interpretive research, making new forms of knowledge possible in conjunction with the music texts accessible online. As the project partners Audio Laboratories in Erlangen possessed the relevant experience, one of the objectives was to synchronize music encoding and audio recording. This was less about the synchronization itself than about the development of semi-automatic methods for horizontally segmenting the audio data. The results of the model experiments, leading to the improvement of the segmentation algorithms are described in the documentation, since they are not directly apparent to the project’s end user. However, the user benefits from these methods insofar as the tool Score follows Audio is used to make the music image and audio recording parallel and to access any desired bar of the number.
In order to enable experimentation with segmenting not only the horizontal, but also the vertical structure of the sound, several Freischütz numbers (nos. 6, 8, 9) were selected in collaboration with the Hochschule für Musik Detmold and its Tonmeisterinstitut [audio-engineering institute] for use in a complex recording process (see the report on the final audio production at ETI of the HfM Detmold)), in which both a total mixing as well as the one-off acceptance of the vocal groups and soloists took place. With these individual tracks, reference data were available for a spectral segmentation of the overall mixture, allowing in turn the optimization of corresponding algorithms. Intended in the long term is using these experiences to experiment with historical recordings in the direction of vertical segmentation – for instance, emphasizing certain instruments or voices in such a recording in a manner facilitating interpretive comparison. To demonstrate the results of the segmentation, a multi-track demo was developed with which the user can filter individual voices from the overall sound, thus demonstrating the parameters under consideration, such as separation precision and shifts in running time (see here, Multitrack Data Set.)
The recording, however, subserved not only the researches of the Erlangen project partners, but also permitted a discussion about considering the characteristics and variants of the score text in the configuration of the orchestral parts. Of course, with the large orchestral apparatus, interpretation alternatives could be tried out only to a limited extent during rehearsals, based on uncertainties in the score, yet this led to a series of interesting findings, albeit to be applied with extreme caution. A few selected passages are therefore presented in Freischütz Digital (see Conclusion of the Audio Production at ETI of the HfM Detmold). The hope is that the work achieved here only for the selected numbers will have an impact on later examination of the whole work’s music text.
This recording is also particularly important in that all parties were willing to provide the data for this recording for future scholarly research free of charge.
Including historical recordings of the work was possible only in an exemplary way within the project’s context: thanks to the support of the owners of audio media, several recordings of the numbers 6, 8, and 9 could be used as demos of a tool – still very rudimentary – for comparing interpretations. The recordings were by the following:
- Otto Ackermann (1951)
- Wilhelm Furtwängler (1954)
- Erich Kleiber (1955)
- Eugen Jochum (1959)
- Carlos Kleiber (1973)
- Marek Janowski (1994)
- In addition, the recording of the HfM Detmold under Karl-Heinz Bloemeke (2014)
In this case, too, it was merely an exemplary demonstration of new possibilities for interpretive comparisons; further audio recordings (from, among others, the archives of the parts of the SLUB Dresden) are referenced in the section on the question of the appoggiaturas (article in preparation).
6. Technical Prerequisites
Technical prerequisites for implementing the project include among other things:
- The XML encoding formats of the Text Encoding Initiative TEI and the Music Encoding Initiative MEI: Whereas in the conceptual area, besides using the TEI module Performance Texts, concepts also for encoding genetic textual details could be consulted, in particular (see guidelines, the section Representation of Primary Sources), for the music encoding, besides using the current MEI-Release 2013, the MEI technical team needed continuous discussions with the MEI community within the context of the problems encountered. The close collaboration between project and this team led to a series of concrete proposals for the upcoming release of MEI.
- dThe current development of the Edirom tools and especially of Edirom online: These tools, initially conceived within the framework of a DFG project, have been continuously developed over the last few years and were an essential prerequisite for drafting suitable processing and display forms.
- Various presentation libraries for rendering MEI data, in particular, the Verovio library developed by Laurent Pugin, which could be integrated within the last phase of the project in Edirom online. Single features needed for FreiDi were developed and implemented in close consultation with Pugin’s project.
- [Further prerequisites in the narrower technical sense will be added at this point.]
7. Presentation Concept
Since the start of the project, a twofold strategy has been pursued for presenting the project results: In addition to making the results available via various forms of publicly accessible websites or repositories (to be described here), articles, presentations, posters or workshops have conveyed the important addendum to the scholarly community, for the project’s aim was not to present finished results, but to present, above all, new possibilities and to stimulate discussion about suitable new concepts in the digital medium combined with promoting development of relevant workflows, to optimize necessary standards, and to develop further suitable tools. This is and will remain a central prerequisite for successfully and sustainably establishing new edition forms and techniques in our specialist disciplines. (An overview of the activities to be continued in this field is given by the FreiDi website under Publications & Presentations.)
Since to date no overall architecture has existed for the closely linked publication of results from a wide range of different areas, though such was also not feasible in this project, the results are accessible in the most varied and flexible form possible:
- The tools of Edirom online have been designed to provide the highest level of integrative capacity. All important sources and most visualizations are combined here under one interface: explanatory text, all digitized music and text portions, audio files, annotations, and metadata. Furthermore, this interface provides targeted access to the datasets and to a variety of possible combinations. Thus, in the music, e.g., individual numbers or bars are selected and collated with other sources or presented in text layers used in the manuscript libretti for better identifying additions in the manuscripts. The parallel opening of windows from the areas of work, music, texts, reference texts, and recordings also contributes significantly to utilizing the data flexibly. Finally, integrated in this interface are annotations opening both internal windows in music or texts, together also with external pages of the Weber-Gesamtausgabe (WeGA) with their extensive information about people, as well as linking the most diverse d ocument types (letters, diaries, writings, performance reviews, reviews etc.). Even a (very preliminary and still deficient) presentation of the music text using the rendering of the Verovio library is possible for Weber’s autograph and for the selected numbers (6, 8, 9) through Edirom online.
- All documentation and the specific aspects of the music and text editing, together with the handling of the demos illustrating audio data are available on the project’s website Freischütz Digital. This should clarify the fact that these are solutions to specific problems whose complete implementation was no longer possible on the entire data set within the project’s context. However, these are not mere mockups, but prototypes based on the data and they are currently not really able to consistently handle the diversity of the data structures, needed for which would be a much more advanced programming effort than was possible in our project. The solutions should therefore be applicable in projects equipped with appropriate capacities.
- Text, Music, and acoustic data (in the form of XML data or audio files) are accessible in several places: on the GitHub pages for Freischuetz-Digital, on the other hand, within the TextGrid Repository. Thus managed is the prerequisite for an active reuse of the data.
(Translation: Margit McCorkle ©2017)