Digitizing and Transcribing Field Recordings of Catalonian Spanish
|Robert E. Vann|
|Western Michigan University|
Advances in technology and falling prices have given ordinary computers the ability to digitize and preserve analog field recordings for later linguistic analysis. Nevertheless, standards for creating and archiving digitized texts are still emerging (Bird & Simons, 2002). This paper describes the practices chosen to digitize and transcribe analog cassette recordings in an ongoing project to document the variety of Spanish spoken in Catalonia, Spain. I hope the description of the practices used in my project contributes to the knowledge base from which recommendations of best practice in digitizing and archiving field recordings may emerge.
The variety of Spanish spoken in Catalonia (henceforth "Catalonian Spanish") has not been documented previously with textual representations of spontaneous speech such as published transcripts or digital audio recordings. After a brief historical background on the dialect itself and a discussion of the rationale for my project, I review the methodology of my project. Specifically, I discuss the practices chosen to economically digitize, transcribe, and archive 20 hours of audiocassette field recordings with maximal fidelity, portability, and potential access to the digital materials created. Possibilities for best practice are suggested. Following a report of outcomes and benefits, I conclude with a discussion of how my research relates to the E-MELD project.
|2.0 Historical Background and Rationale|
Catalan and Spanish are distinct Romance languages that have been in contact with each other for centuries in northeastern Spain. Presently, these languages are co-official in the autonomous region of Spain known as Catalonia. The prolonged contact with Catalan has influenced the Spanish of this region greatly, and most Spaniards recognize Catalonian Spanish instantly. Nevertheless, for largely political and ideological reasons having to do with the Francoist dictatorship of the mid 20th century, this publicly well-recognized dialect of Spanish has been all but ignored in academic research on Spanish dialectology. Thus, Catalonian Spanish remains largely overlooked in linguistic treatises, unrecognized as a legitimate dialect of Spanish, and uninvestigated from a social perspective (Vann, 2002).
Since Spain's transition from Francoism to democracy a quarter century ago, scholars have begun to take some interest in the Spanish spoken in Catalonia (cf. e.g. Badia i Margarit, 1980; Marsá, 1986; and Casanovas, 1996). Nevertheless, these investigations have been based on anecdotal data and casual observation. They have not provided extensive speech data to catalog the unique linguistic characteristics of Catalonian Spanish. Thus, Catalonian Spanish still remains unrecognized as a legitimate dialect of Spanish.
Prominent Spanish linguists (e.g. Boix, Payrató, and Vila 1996) have realized that to improve this situation, we need linguistic corpora of Catalonian Spanish. Linguistic corpora would enable scientific examination of this underinvestigated dialect for the first time, allowing reliable description of its distinct linguistic characteristics. Hence, the rationale for my project is simple: leading research explicitly calls for the publication of linguistic corpora to better investigate Catalonian Spanish, and no corpora of Catalonian Spanish have yet been published.
My own research has worked to improve this situation by analyzing some of the linguistic traits that distinguish Catalonian Spanish within Spanish dialectology. In my doctoral dissertation (1996), I analyzed several linguistic attributes of Catalonian Spanish based on interview data.1 Since then, I have pursued an extended line of pioneering research dealing with the linguistic description of Catalan ways of speaking Spanish and the cultural significance that such ways of speaking can convey.2 A grant I received in 1998 enabled me to employ a research assistant natively fluent in Spanish to transcribe approximately 20 hours of conversations in Catalonian Spanish that I recorded in Catalonia in 1995.3 These field recordings, which form my oral corpus of Catalonian Spanish, came from spontaneously occurring group conversations with members of two social networks in Barcelona, Catalonia's capital. Orthographic transcription of the oral corpus yielded a preliminary written corpus of 500 pages of single-spaced text.
The work I describe today concerns digitizing the taped conversations and editing the transcripts for linguistic accuracy. Eventually, I intend to publish the written corpus accompanied by the digital audio in a scholarly monograph to scientifically document Catalonian Spanish for the first time. This monograph will address significant voids in Spanish dialectology. Moreover, it will provide a foundation for research on this dialect that can subsequently be pursued by me and others, and this resulting research will provide new perspectives on Spanish language and culture.
My project involved economically digitizing the audiocassette field recordings and editing the preliminary written corpus for linguistic accuracy. Though supported by a sabbatical leave, my research was unfunded. Therefore, I was on a tight budget. I suppose you could say that I aimed for the best practices that no money could buy. The practices I chose to digitize and transcribe field recordings of Catalonian Spanish were selected to maximize fidelity, portability, and potential access to the digital materials created, given my limited resources.
Editing a preliminary corpus of transcribed speech for linguistic accuracy involves careful comparison of the word for word data entry with the actual recorded discourse. Rewinding and replaying audiotapes repeatedly as the transcripts were meticulously scrutinized would take much longer than editing the transcripts with digital audio, so I thought the best practice here would be to digitize first and edit the transcripts later.
Digitization and editing were carried out on a Bronze Powerbook G3 running system 9. Fortunately, much freeware and shareware is available on this platform to digitize audio. I considered the following audio applications for my project, among others: Audiocorder, Coaster, SoundStudio, and MicNotePadLite.4 Comparing and contrasting the value for my project of available software took much more time than I had anticipated. I had to search the applications out on-line, study pertinent read-me files to determine if each application would run on my system configuration (OS, RAM, hard drive space, etc.), download acceptable applications, install them, learn how to run them, and test them with a sample recording.
All of the applications I reviewed have the ability to digitize sound imported through the Powerbooks line-in jack, but only one of them includes transcription tools: MicNotePad Lite. This freeware application rolls digitizing audio and keyboard-controlled background audio playback into one powerful application. For these reasons, I chose MicroNotePad Lite for my audio needs. In digitization, this program offers input source control, sampling rates ranging from 7-44 kHz, sample sizes ranging from 8-16 Bits, mono and stereo channels, audio play-through, automatic or user-controlled gain control, and compression ratios ranging from none to 9:1. Supported file formats include WAV and AIF. In playback, MicNotePad Lite offers background keyboard control over the playrate (37%-271% of normal) as well as programmable keys for rewind, play, fast forward, etc.
Though software applications as well as the features they contain are constantly changing, and different projects will have different software needs, my experience suggests that the best practices regarding software selection would involve an economy of applications. It is both quicker and easier to learn well one program that does digitization and transcription than it is to learn two or more different programs well. It is also less taxing on ones OS to have fewer programs running simultaneously.
My original field recordings were made on a 1991 vintage handheld Sony TCM-89V monaural cassette recorder with a tie-pin style high-performance Sony ECM-T7 omnidirectional, battery-powered external microphone.5 The quality of the original recording equipment was certainly not professional grade, and I did not want to lose any more data than I had to when I played the cassettes for digitization. Therefore I borrowed a colleagues professional grade Marantz to play the tapes for digitization. It strikes me that a good practice in digitizing audiocassettes would be to use the highest quality tape player that one can gain access to. A $500 machine should play the same tape better than a $50 machine. To digitize the cassettes, I connected a gold-tipped monaural line from the headphone jack of the tape player to the line-in on my Mac. I then selected the computers external mic as the input source for recording.
The settings I chose to digitize my audiocassette recordings were 16 bits at 44kHz with automatic gain control (AGC) saved in uncompressed AIF format. I chose the highest sampling rate the program offered to most accurately sample in digital form the original analog recording. Obviously, the more frequently the computer digitally samples the incoming audio, the more the signal approaches its analog equivalent. The same logic applied for sample size. To standardize input levels across recordings, I enabled automatic gain control during digitization. Before recording, I played each cassette on the Marantz. By monitoring sound output levels, I was able to set the playback volume to avoid clippings.
These practices were intended to maximize the fidelity of the digital recording. My goal was to preserve the original analog recording in digital form with minimal signal loss / alteration in the process, inasmuch as my limited financial and technical resources would allow. Of course, more expensive software might offer higher sampling rates and larger sampling sizes, perhaps with a better AGC algorithm. Likewise, more expensive hardware might offer higher quality digitization. Nevertheless, my aging Powerbook and the freeware I downloaded from the Internet proved more than sufficient for the technical needs of my project. For these reasons, my experience suggests that realistic best practices in digitizing field recordings will be those that balance the technical requirements of a given project with the financial resources available to fund it.
The digital audio format I selected was chosen for its portability. AIF is a cross-platform standard that can be archived as data or music on CD. Such CDs could be read by any modern computer or played by any modern CD player. In other words, this format was chosen to maximize potential access to the digital materials created.
Many shareware audio applications offer various sound editing features including diverse filters and effects to amplify, normalize, and smooth digital audio signals. My original cassette recordings contain all manner of pops, clicks, and ambient noise recorded as part of the taped conversations. I view these background noises as part of any culturally contextualized spontaneously occurring speech. What some might view as background noise I view as potentially important linguistic data. Therefore, other than setting the AGC, I have purposefully not modified or manipulated in any way the signal I digitized from the audiocassettes. Of course, investigators interested in future phonetic analysis of the digital audio corpus may certainly choose to do so. For now I have deliberately left the digital audio as "muddy" as the analog counterpart it came from. It strikes me as a good practice in digitizing field recordings to be conservative and leave them as untouched as possible for publication, leaving retouching for later analysis on a nondestructive basis.
After digitizing the audiocassette field recordings, I edited the preliminary transcripts that had been word-processed by my research assistant in 1998. Mistakes and oversights in the preliminary transcriptions were painstakingly corrected at this stage, and material previously not included was added where necessary. In the case of my particular corpus, for example, while most of the speech data are in the Spanish language, albeit a Catalanized dialect, some passages are actually spoken in Catalan, a different Romance language altogether, and some words are coined from a mix of both languages. Since my research assistant was not able to completely transcribe some of these bilingual passages in the preliminary corpus, I had to add some material as I edited. Finally, because audiotaped voices sound similar and can be easily confused in transcriptions of group conversation, editing the preliminary written corpus required review of videotapes to verify the accurate identification of the different speakers.
I undertook the task of editing the preliminary transcripts with the aforementioned goals of maximizing fidelity, portability, and potential access to the digital materials created. Regarding the fidelity of the transcription, I determined that the best practice for my project was to make indications in the edited transcripts for pauses, false starts, repeated words, punctuation, and interruptions, but not for pronunciation. Here is where fidelity meets access (and where for me lies the difference between creating a linguistic resource and creating a linguistic analysis). I wanted the transcripts to acknowledge the aforementioned discourse phenomena, but I also wanted the transcripts to be accessible to the lay reader not trained in linguistics. In other words, I wanted to avoid idiosyncratic representation and make the transcripts useful for a variety of different purposes. Furthermore, I had to consider the project budget and timeline.
For these reasons I decided to do all transcription following simple orthographic conventions with minimal formatting and mark-up. All words were spelled out as they would appear in a dictionary6 and discourse phenomena were indicated through simple and familiar keyboard combinations. For example, pauses were indicated by three dots " " and false starts were indicated with hyphens "-". No linguistic mark-up system needs to be learned to follow my written corpus. Of course, nothing precludes future investigators interested in conversational analysis from linguistically marking up the corpus on an individual basis for the purposes of specific research. Likewise, future investigators interested in phonetic or phonological analysis are free to use the transcripts as a basis for phonetic / phonemic transcriptions of the corpus. Since at this time I am interested only in creating the digital texts as a resource for general use, I have deliberately left the transcripts as generic as possible. I leave phonetic transcription and conversational mark-up for those who would use my resource in future linguistic analyses.
To maximize the portability of the written corpus, transcripts were saved as WORD documents with the intention of creating print copies for publication. I am considering publishing the transcripts as digital text files as well to provide a point of departure for future investigators who might want to digitally mark-up the written transcripts. Future investigators with access to these documents as WORD files would have powerful editing tools at their disposal. As Bird & Simons (2002, p. 2) point out, aside from being "readily available, often pre-installed, and familiar," word processors have the advantages of WYSIWYG, find/replace, and cut/paste editing. Bird & Simons also point out, however, that proprietary file formats such as WORDs may limit the useable lifespan of digital texts. They suggest (p. 7) that best practices would permit the use of proprietary formats for delivery provided that resource storage were nonproprietary or open. Perhaps if I choose to publish the transcripts as digital text files as well as in print, I will export them to an open format application like Transcriber and store them as XML or Unicode, which should remain more readable than WORD format across platforms and over time. Either way, the digital text files containing the written transcripts will be archived on CD along with the digital audio materials I have created and some of these digital materials may be published alongside the transcripts in print in the form of a scholarly monograph. This monograph will serve to formally document Catalonian Spanish as a distinct dialect within Spanish linguistics based on the corpus as primary source data.
|4.0 Outcomes and Benefits|
The project is beneficial in its own right for several reasons. First, it addresses a significant problem in the traditional literature of Spanish dialectology (cf., e.g. Zamora Vicente, 1989), which is the lack of professional recognition of dialects of Spanish spoken in contact with (and influenced by) Catalan. Even sociolinguistic investigations that have specifically addressed Spanish in contact with Catalan in Catalonia have not recognized Catalonian Spanish as a distinct dialect of Spanish (cf. e.g. Badia i Margarit, 1980). Therefore, the idea of recognizing this contact variety as a legitimate Spanish dialect is a novel concept in the field.
The second reason why this project is beneficial in its own right is that my digital corpus registers extensive examples to facilitate the linguistic description of Catalonian Spanish in the academic literature. A similar small-corpus monograph has been published for Spanish in contact with English (Lope Blanch, 1990 on US Spanish), yet no such study currently exists for Catalonian Spanish. Unlike previous studies that rely on anecdotal data (cf. e.g. Badia i Margarit, 1980; Marsá, 1986; and Casanovas, 1996), my corpus project represents a scientific way to document Catalonian Spanish in the literature, opening the door to its future linguistic description and analysis. Furthermore, to my knowledge, publications of linguistic corpora involving other dialects of Spanish have never before included any audio component. Digital audio of the conversations transcribed in my monograph on Catalonian Spanish will allow colleagues, students, and other interested individuals to actually hear for themselves the conversations they are reading in the transcripts.
In addition to aiding in the linguistic description of Catalonian Spanish, the corpus will function as a resource for investigators interested in carrying out sociolinguistic analysis of Catalonian Spanish. My own studies with the individuals recorded in the corpus suggest that statistically significant correlations exist between Catalan ways of speaking Spanish and the amount of exposure individuals had to the Catalan language in their youth (Vann 1998, 1999a, 1999b). Moreover, through the metalinguistic commentary it contains, the corpus provides rich data for qualitative sociolinguistic analysis regarding community ways of speaking. My own studies with the individuals recorded in the corpus (Vann 1995, 2000) suggest that they have developed unique ways of speaking Spanish that reveal patterns of cultural distinction.
The third reason why my project is beneficial involves its value as a useful resource in other fields of linguistics and beyond. For example, my digital materials provide abundant conversational data for the study of language contact phenomena. Moreover, because the use of Catalonian Spanish can reflect and even constitute important sociocultural differences between Catalans and other Spaniards, the corpus materials may be of interest in other fields that are concerned with cross-cultural communication, including but not limited to discourse analysis, cultural anthropology, and public affairs. The corpus materials also have value in the classroom, where they provide new source material for teaching with technology.
Finally, and perhaps, most importantly, my project provides knowledge that has considerable social value for the community studied. Within the community investigated in Barcelona, people's use of language is influenced by their history of exposure to Catalan and Spanish. Patterned linguistic experiences can sometimes cause a speaker to react linguistically without thinking. At other times, these same experiences can also lead to more conscious linguistic acts designed to reflect the speaker's social identity. Of course, both possibilities can result in the same interpretation on the part of others, regardless of the speaker's intent. Public interpretations of linguistic behavior are very important because they can affect socioeconomic relationships. Many monolingual Spaniards view Catalonian Spanish as "bad" or "broken" Spanish and hold low opinions of people who speak this dialect. Therefore, documenting Catalonian Spanish as a legitimate dialect of Spanish has the potential to improve intercultural relations and perceptions at both the regional (Catalonia) and national levels in Spain.
In describing the rationale for digitizing and transcribing field recordings of Catalonian Spanish, the practices used, and the benefits attained therein, my paper contributes a case study to the knowledge base concerning digital text creation and archiving in linguistic research. Though my project itself as well as and its rationale and objectives are unique, the practices involved in creating and archiving digital texts for my project may be useful to others involved in similar endeavors. Accordingly, I hope my experiences are of use in working towards "best practice" recommendations for digitizing and annotating texts and field recordings in linguistic research on a more general level.
On this more general level, my research is consonant with the E-MELD projects attempt to build digital standards for the preservation of language data. The goals of my project to offer the widest possible access to the digital materials created and to provide information in a maximally useful form support the stated goals of the E-MELD project. Nevertheless, I submit that interpretation of the terms "access" and "useful" will vary from project to project, and so will associated practice. The E-MELD project description itself recognizes the need to be flexible in formatting and practice, all the while advocating comprehensive publication of linguistic data, metadata, and markup in a platform independent, open format. Given enough time and funding, the publication of metadata and linguistic markup may well be desirable and attainable for large ditigal corpus projects directed to a linguistic audience. Not all investigations, however, have a need or use for such linguistic details or the temporal and financial resources required to include them. Furthermore, such details could actually make texts less accessible to certain audiences. My project, for example, is directed to a lay audience as well as a linguistic one, and it was constrained by time and funding. Linguistic markup and metadata in my digital corpus would not be accessible or useful to my entire potential audience. Conversely, all audiences will find the common denominator of orthographic transcription useful and accessible.
My own research is part of the growing movement to create digital archives of language data and documentation. I agree that common standards will be necessary if we are to avoid a degree of variation in archiving practices and language representation that seriously inhibits data access, searching, and scientific investigation. Yet, the flexibility we must allow in research practice may inhibit the adoption and use of umbrella standards based on predetermined definitions of usefulness and accessibility. I submit that best practices should creatively respond to flexible definitions of usefulness and accessibility to be determined in individual projects.
The E-MELD project description states in its introduction that "any proposed solution must allow for some continuing variation in individual practice. Not only will different languages and theories always call for different analytical categories, but different research questions will always call for different types of data manipulation and display." My experience supports this caveat, demonstrating that research objectives, projected audiences and budgetary constraints will all influence actual methodological practices.
1. I would like to acknowledge the aid of the Program for Cultural Cooperation between Spains Ministry of Culture and the United States Universities, subvention #1490, which facilitated the collection of these data.
2 This research has produced fifteen journal articles and book chapters during the last seven years, to my knowledge the largest number of publications in this area by any scholar during this timeframe.
3 I would like to acknowledge the aid of the Faculty Research and Creative Activities Support Fund, Western Michigan University.
4 At the time I had not yet run across Transcriber.
5 The frequency response of the external mic was 100-15,000 Hz, exceeding that of the cassette recorder itself, which was 150-6,000 Hz.
6 Neologisms were spelled out according to standard rules of Spanish orthography.
Badia i Margarit, A. (1980). Peculiariadades del uso del castellano en las tierras de lengua catalana, Actas del I simposio para profesores de lengua y literatura españolas (pp. 11-31). Madrid: Castalias.
Barras, C. (2000). Transcriber (Version 1.4.3) [Computer freeware]. Arcueil Cedex, France: DGA.
Bird, S. & Simons, G. (2002). Seven dimensions of portability for language documentation and description. In Proceedings of the Workshop on Portability Issues in Human Language Technologies, Third International Conference on Language Resources and Evaluation (pp. 23-30). Paris: European Language Resources Association.
Boix, E., Payrató, L., & Vila, F.X. (1996). Spanish and Catalan. In H. Goebl et al. (Eds.), Contact linguistics: An international handbook of contemporary research (pp. 90-104). Berlin: De Gruyter.
Casanovas Català, M. (1996). Algunos rasgos propios del español en las comunidades de habla catalana: Fonética, morfosintaxis y léxico. Analecta Malacitana, 19(1), 149-160.
E-MELD Proposal. (2002, July 17). Retreived April 16, 2003 from http://linguist.emich.edu/%7Eworkshop/E-MELD.html
Kwok, L. (2001). SoundStudio (Version 1.5.4) [Computer shareware]. Philadelphia, PA: Felt Tip Software.
Lope Blanch, J. M. (1990). El español hablado en el suroeste de los Estados Unidos: Materiales para su estudio. México (D. F.): Universidad Nacional Autónoma de México.
Marsá, F. (1986). Sobre concurrencia lingüística en Cataluña. In M. Alvar et al. (Eds.), El castellano actual en las comunidades bilingües de España (pp. 93-104). Salamanca: Junta de Castilla y León.
MicNotePad Lite (Version 4.5) [Computer freeware]. (1998). Santa Cruz, CA: Nirvana Research.
Roth, C. (2001). Coaster (Version 1.1) [Computer freeware]. http://<www.visualclick.de/>: VisualClick Software.
Smolinski, C. (2000). Audiocorder (Version 1.9.8) [Computer shareware]. Westminster, MD: Black Cat Systems.
Vann, R. (1995). Constructing Catalanism: Motion verbs, demonstratives, and locatives in the Spanish of Barcelona. Catalan Review, 9(2), 253-274.
Vann, R. (1996). Pragmatic and cultural aspects of an emergent language variety: The construction of Catalonian Spanish deictic expressions (Doctoral dissertation, The University of Texas at Austin). Dissertation Abstracts International (University Microfilms No. 9633318).
Vann, R. (1998). Aspects of Spanish deictic expressions in Barcelona: A quantitative examination. Language Variation and Change, 10, 263-288.
Vann, R. (1999a). An empirical perspective on practice: Operationalizing Bourdieu's notions of linguistic habitus. In M. Grenfell & M. Kelly (Eds.), Pierre Bourdieu: Language, Culture and Education. Theory into Practice (pp.73-83). Bern & London: Peter Lang.
Vann, R. (1999b). Language exposure in Catalonia: An example of indoctrinating linguistic ideology. WORD, 50(2), 191-209.
Vann, R. (2000). Constructing reality in bicultural communication: Catalan ways of speaking Spanish. Intercultural Communication Studies, 10(1), 113-124.
Vann, R. (2002). Linguistic ideology in Spains ivory tower: (Not) Analyzing Catalonian Spanish. Multilingua, 21(2/3), 227-246.
Word (Version Macintosh Edition 98) [Computer software] (1998). Redmond, WA: Microsoft Corporation.
Zamora Vicente, A. (1989). Dialectología española (2nd edition, augmented). Madrid: Gredos.