DIGITIZING AND ARCHIVING THE SPIDER ROOM COLLECTION Steven Moran University of Washington Introduction The Language Learning Center (LLC) at the University of Washington was established in the 1950s to provide students with state-of-the-art technologies for learning foreign languages. For over 30 years foreign language teaching materials were recorded at the LLC on reel-to-reel tapes so that students following the audio-lingual method could play them back for drilling purposes. As foreign language teaching methodologies evolved, so did the materials that we recorded. These recordings also include music, stories, cultural activities and song and dance from languages that are as geographically and lingual-genetically disperse as Shona, Kannada, Tibetan and Icelandic. Over the decades the LLC's collection has also grown from deposits from UW faculty and researchers. Some of these precious materials include recordings of languages spoken by very small groups of individuals. Lushootseed has less than 60 speakers in the Pugent Sound area. Haida, a language spoken in Queen Charlotte Islands in British Columbia, is also nearly extinct and current estimates are that there are less than 30 speakers in Canada and fewer than 15 in the United States. Cornish is also rapidly disappearing and has only 500 speakers, 100 of which are capable of speaking the language fluently. Navajo and Javanese, though slightly larger, are each spoken by less than 7000 in the Southwestern United States and in in New Caledonia in the South Pacific. The collection in its entirety represents at least 94 languages recorded on over 1000 reel-to-reel tapes that currently reside in our 'Spider Room', a small windowless corner of Denny Hall. The goal of this project is to develop a plan of action to examine the current condition of this collection, determine what hardware and software are needed for digitizing the tapes, to investigate best practices for digitizing these materials, and finally, to design an online database for cataloging each record item. To that extent we will outline in this paper the progress we have made so far towards these goals. We provide an inventory of the reel-to-reel tapes discovered thus far in the Spider and Server rooms. We then describe our current digital audio work station set-up and address issues of digital preservation through digital best practices for audio digitization. We will also describe our digitization preservation lab and our database implementation for cataloging and expanding the collection's metadata. Finally, we address issues of the physical preservation of the reel-to-reel tapes and propose a timeline for the digitization and cataloging of the collection. Our overall aim is to focus on archival access and preservation of the reel-to-reel recordings that make up the Spider Room collection.
Inventory A rough physical count of the Spider Room reel-to-reel collection indicates more than 1000 tapes, varying in both size and format. We have discovered 3 3/4ips and 7 1/2ips recordings on 5 and 7 inch reel-to-reel tapes. So far the following languages are representative within the collection: Amharic [amh]1, American Language2, Arabic (Eastern Arabic [avl]), Arabic (Modern Arabic), Arabic (Egyptian Arabic [arz]), Bambara [bam], Bengali [ben], Bulgarian [bul], Burmese [mya], Cambodian [khm], Catalan [cat], Chinese (Mandarin [zho]), Chinese (Cantonese [yue]), Cornish [cor], Czech [ces], Danish [dan], Dutch [nld], Dyula [dyu], English [eng], Estonian [est], Ewe [ewe], Finnish [fin], French [fra], Fulani [ful]3, Georgian [kat], German [deu], Greek (Classical) [ell], Greek (Modern), Haida [hdn], Hausa [hau], Hebrew [heb], Hindi [hnd], Hungarian [hun], Icelandic [isl], Igbo [ibo], Indonesian [ind], Irish [gle], Italian [ita], Javanese [jas], Japanese [jpn], Kannada [kan], Kazakh [kaz], Kirghiz [kir], Korean [kor], Krio [kri], Lao [lao], Lingala [lin], Lithuanian [lit], Latin [lat], Luganda [lug], Luhya [luy], Lushootseed [lut], Mende [men], Mongolian [khk], Navajo [nav], Nigerian [pcm]4, Norwegian [nob], Persian [pes], Polish [pol], Ponopean5, Portuguese [por], Punjabi [pan] (or pnb], Rhodesian6, Romansh [roh], Romania Cantat [ron], Modern Romanian, Russian [rus], Sahaptin7, Sanskrit [san], Serbo-Croatian [src], Sesotho [sot], Shona [sna], Singhalese [sin], Slavic8, Slovak [slk], Somali [som], Spanish [spa], Swahili [swh], Swedish [swe], Tagalog [tgl], Tajik [tgk], Tamil [tam], Thai [tha], Tibetan [bod], Turkish [tur], Twi [twi], Ukrainian [ukr], Uzbek [uzn], Vietnamese [vie], Wolof [wol], Yorubu [yor] (also 'Yoruba'), and Zulu [zul]. There are also other reels that are titled: Anthropology, Other Drama, Other Education, Linguistics, An Introduction to American Government, Stress Management/Relaxation Tape, Alaskan Stories and Songs, Duwamish Archaeological Dig, Venda Music, Chinese Sither Music, Osette Archaelogy Manual, Incas and Spaniards, Washington State Historical Road, 84 Conference Autumn 1984, 85 Spring Regional Conference, Adult Drug Therapy (ADT) 5/12/89, Pediatric Drug Therapy (PDT) 3/23/89, Neuropsychotropic Drug Therapy (NDT) 6/16/89, and HIV (1989). There is also a Speech and Hearing Sciences collection, including General Phonetics; Historic Speakers, SPHSC-302, Manual of American English Pronunciation 3 ed., Phonetic Transcription, General Phonetics Speech 302, Dew-Jensen, English Sound System (111), Various Accents, as well as a number of titles simply called Speech. The Linguistics collection includes the titles: Manual of Articulatory Phonetics, Kechua Indian, Naturalistic Approaches to Languages and

The Ethnologue's three letter language codes and ISO/DIS 639/2. We are not yet sure what the publisher/author means by 'American language'. Fulani is actually a group of related languages that are genetically related and spoken in Cameroon. We have tagged this is Nigerian Pidgin, having no further knowledge of what language this may be. We are unsure what language this recording represents. The Rhodesian collection represent numerous languages that have not been fully identified. There are five Sahaptin languages and we are not sure which this recording represents. The Slavic collection represents a language family.
Cognition, Accent Inventory Linguistics, Sounds of the Scandinavian Languages and Gullah Dialect. This list of the Spider Room's collection, however, is not exhaustive. We have collected these languages and titles from a now non-functional FileMaker Pro database, which use to reside on Niihau, one of the LLC's web servers. This machine quit working during the Winter semester of 2007 and a backed up copy of the database's contents were salvaged and we inputted its contents into an Excel spreadsheet. We then took this data and used it to populate a MySQL database table, which we built to as a Web accessible front-end via PHP. This contents of this database can be accessed through: http://students.washington.edu/stiv/spider/. The Spider Room collection is currently split between two separate rooms, both in the basement of Denny Hall. The majority of the collection is located in the Spider Room, a small window-less storage room in the building's south side. This portion of the reel-to-reel tapes represent perhaps 3/4 of the entire collection and is being stored in cardboard boxes, approximately 2x1x1 in dimension. About half of the boxes were labeled some time in the past. However, today their labels do not necessarily coincide with their contents. We have begun to sort through these materials, re-labeling the boxes with language codes to indicate what materials are presently stored in them. There are also several dozen reel-to-reel tapes which are not in boxes. They reside in the Spider Room stacked on top of each other, on shelving, or on other boxes. The second room in Denny Hall, the Sever Room, is home to the remaining 1/4 of the LLC's reel-to-reel collection. In this room the tapes are organized by language, stored on shelves built into the wall. They reside roughly six feet above the floor. Although we have not yet taken temperature or humidity samples, both rooms are climate controlled and are relatively cool due to air conditioning. The server room, in particular, is always kept at 69 degrees Fahrenheit. Each room contain both 5 inch and 7 inch reels. A sampling of the reels indicates that the collection contains, at least, these brands and models of tapes as indicated on their packaging: Scotch recording tape Standard Length 175 Exclusive Heavy Duty Coating, Silicone Lubricated 1.5 Mil Tenzar Backing 1/4 IN. x 1200 FT. (6.35mm x 360m) 7" REEL, Scotch recording tape 3M Company 5 inch, Ampex 631 1.5 MIL Polyester 1/4" x 1200 FEET SEVEN INCH REEL PROFESSIONAL, Scotch Brand Magnetic Tape 120 Plastic High output Splice Free 3M, Electicraft GOLD Label 5" Reel and Scotch Magnetic Tape RB-5 Empty 5 INCH REEL IN BOX. This list is being expanded as we encounter new reel manufacturers. For our sampling purposes, we have examined sets of reels from both rooms. To date we have found no signs of oxide shed from either room, though it is probable that it will be encountered. We have found, however, a number of reels from 1974 that display print-through. They were recorded on Scotch Brand Magnetic Tape 175 Heavy Duty Tenzar. The tapes are part of the Rhodesian Music collection. In one section two young children are whistling in a 'slinging match' against each other,

while they herd cows and converse: Boy 1: "Saka usina mombe dzese" "this is why you don't have all your cows" Boy 2: "Ndiwe wakati dzirasike" "you are the one who made them lost." Boy 1: "Iwe wa-imbo rasa waiti kune mandodo" "you used to loose them saying there are [unknown word]" Boy 2: "Kudoku tyaya bora kwai tyaya musoro" "see the way I kick the ball, I will kick your head"9 During pauses in their discourse there is a faint whistle print through. Print through has not been perceived on the other tapes that have been so far sampled. However, with over 1000 tapes there is likely to be more print through as well as some oxide shed. When oxide shed is encountered, we will seek assistance from other sources on campus, especially if cases where we need to bake the tapes. It is beyond the LLC's means to purchase a specialized oven for this task. The sample so far suggests that the tapes are otherwise in decent to good condition. More than 3/4 of the 50 tapes sampled have leader and the tapes seem to be in fairly good quality, with some crinkling at the beginning and end through wear and perhaps poor handling. Digital Audio Preservation Lab We have moved forward with setting up a digital audio workstation that will allow us to achieve our goals for digitizing the Spider Room collection. The LLC has purchased an Pioneer RT-707 tape deck unit that is connected to the patch bay in the recording studio, which is then fed through the audio system. We are using an Audiophile external digitization box to run the signal from the audio set up into a Mac G5. The Mac is running OS X 10.4 and we are using Peak Pro Audio Version 5 for the digitization software. Peak Pro allows for up to 48kHz and 32 bit sampling rates. The current IASA and AES standards are now up to a 96 kHz sample rate and a 24 bit rate. We are currently digitizing at CD quality, 44.1kHz and 16 bit. At this time we do not have plans to do any digital restoration of the sound files that we are creating, but it is a factor that we are investigating. Our main goal is to get the bulk of the tapes into a digital format. We are currently storing the files in AIFF and WA formats on an V external hard drive, but we will have to address a more permanent storage medium, such as more hard drives, DVDs or CDs. The LLC is committed to seeing the digitization project through to its completion. Director Paul Aoki has authorized the use of LLC funds to purchase the Pioneer RT-707 Tape Deck. Computer

9 The translation was graciously provided by Colleen Gono and Daniel Atkinson.
support is being provided by LLC Computer Specialist Bob Majors, who has set-up the digital audio workstation and provided technical documentation (see appendixes A and B). The LLC is also supporting a Research Assistant who will lead the project, develop the website and database and digitize the collection. In the future it may be possible to train an undergraduate from the LLC window staff to help with the digitization. If we digitize at 44.1kHz 16 bit 1 channel audio, the Spider Room collection, at an estimated 1000 reels at 60 minutes average per reel, would require 300GB of digital storage space. Purchasing large hard drives may be a fiscally reasonable option for the digital collection in house. Also, because many of the items are spoken recordings of language, it may be appropriate to keep a ceiling at 44.1kHz 16 bit (the human voice's dynamic range is less than than of the aural dynamic range). However, for the reels-to-reel tapes that we encounter which include music recordings, we believe we should record at higher digital sampling and bit rates. We propose 48kHz and 24 bit 1 channel. For each hour this would require 480MB of space, but it would provide a higher range and better quality for music and atmospheric recordings. If we were to digitize the entire collection at this rate, with an estimated 1000 hours of audio, we would require 480GB of space. We estimate that with the varying types of recordings, both voice and musical, that our total digital storage needs will range between 300 and 480GB. Of course we wish to back-up these digital recordings, so double that space will be needed. At the high end then, we assume that we will need approximately 800GB of digital storage. This turns out to be 200 DVDs or 1000 CDs, which seems like more hassle than purchasing a few large hard drives. Database Design We aim to leverage expertise in relational database design in developing an in-house catalog system that will be available online through a portal at the LLC's website. So far we have implemented a MySQL database with a PHP Web interface as the front end for inputting each item's metadata. This is available online at http://students.washington.edu/stiv/spider/. We are using these technologies because we plan to host the Spider Room Collections Website as part of the LLC website, which resides on departmental servers that are maintained and supported by UW's Computing and Communications. They provide these technologies, which are open source and freely available. The reel-to-reel tapes provide several challenges for collecting metadata with which to populate the database. In particular, there seems to be two ordering systems for many of the reels. The first is an older system that incorporates the language name, series number and tape number, such as M TUR 0131. We cannot at this time account for what the 'M' stands for. A later number system builds upon this order and seems to have been implemented to adjust to the growing size of the collection, e.g. M TUR

01.031. In the case of Turkish this system seems straight forward. On other tapes, however, the numbering system can be as drastically different as 'C 107' and 'M UP JPN 01.001'. We do not propose to change the ordering systems in place. We plan to include them with all metadata found on each reelto-reel and they will be entered into our catalog. This, however, poses another problem for the database design since the metadata included in each reel-to-reel varies quite dramatically. In the case of our Turkish collection, the tape spindle is labeled with permanent marker, M TUR 01.031. The reel-to-reel box is affixed with two stickers, one reads M TUR 01-31, the other M TUR 01.031. Taped to the inside of the cover of the reel-to-reel box is a note card. Typed on it is the following metadata: Turkish 2797 Lloyd B. Swift and Seiman Agrali Foreign service institute BAsic course in Turkish Unit 21 Master LL - 3.75 ips Winter 1970 3.75 ips 30min
However, when playing the tape the speaker provides the following oral metadata, 'Foreign service institute, Turkish language course, 1967.' Our goal therefore, is to catalog as much of the existing metadata provided within the collection as it is digitized. Another example from a Speech and Hearing Sciences tape is as follows: M SPE-5 M SHS 04.005 SHS B-505 (written on box)
M SHS 04.005 (written on tape spindle) Non standard pronunciations (title written on back of box) From the Spanish collection a tape which contains a musical play contains the following metadata: SPA-B506 (on box) Mi Bella Dama Musical Play (Span. Ed) Reel II (on box) (taped to inside cover) SP-V 2263 55Min 7 1/2 ips
no author Mi Bella Dama Musical Play (Spanish Edition) Reel 2 Reproduction of Record L.L. M SPA 39.002 (tape to tape spindle) Mi Bella Dama Musical Play (Span. Ed.) Reel 1/2 (taped to spindle) And the Hindi collection has a 'temporary box label' taped to the outside of the box, which includes Title, Language, Speed and Track. The digital audio workstation has Internet access and a Web form has been built which allows digitizers to add metadata directly to the database as they digitize each reel-to-reel. For this we are using the following data fields:

Language Language Code (will be automatically populated through the language name via ISO codes) Accession Number (e.g. format TUR 01-31) Accession Number2 (e.g. format TUR 01.031) Author Year Publisher Tape speed Length Audio metadata Comments Condition of tape Date entered (will automatically generated by the system)
We may need to be expand these fields as new metadata columns are encountered from other tapes or sources. Another obstacle that we plan to tackle at a later date is the integration of other structural standards, such as MARC 21, METS and Dublin Core. Our current plan is to provide access points at each data field, which would be expanded by adopting another structural standard. The problem we currently are facing through, is that we have such little metadata form these recordings to work with.

Plan for Physical Preservation Regarding physical preservation of the the archival recordings, there are two issues at hand. In 2009 Denny Hall is projected to be remodeled and made more earth quake-safe. These plans will cause the LLC to relocate to Condon Hall for an estimated period of three years. The storage of these materials will have to be addressed within the coming year when plans are made for relocating. It may be at this time that materials from the collection are offered to other archives or they are disposed of. The author suggests that all material of special value be treated with great care. Lushootseed recordings should be deposited with the Vi Hilbert collection in the Ethnomusicology Archives at UW. The remainder of the collection, once offered to Special Collections at UW but rejected, needs to be evaluated. We are of the opinion that several language collections within the Spider Room collection are of great value, especially those that represent endangered language recordings. The support of the LLC is needed to properly store the physical Spider Room collection. Attempts have been made to communicate with senior staff to address these concerns, but they have not yet been reciprocated. The Server Room is normally a cooled 69 degrees Fahrenheit 24/7/365 and it provides an ideal location for the reel-to-reel tapes until future storage of these tapes is addressed. Although it has had AC for about six years, tapes have resided in that room from as early as the mid1950's, so they will not to be moved. There are no temperature readings for the Spider Room and there are no humidity readings for either room. The Spider Room, however, seems to remain fairly cool in temperature. It also has no windows, very low traffic and some climate control. It feels cooler than the hallway it resides in when entering it. It is also very dusty, filled with cob webs and overall pretty dirty. All materials of value will be transported from the Spider Room to the Server Room as they are digitized. Timeline for implementation We have already begun to digitize the Spider Room collection and are making strong progress. For the oral presentation of this paper, on March 9th, we had digitized 10 tapes that represented 2GB of storage space. On March 14th we have 40 tapes (12GB) digitized. Our plan is to set forth the following time line for project completion:

Before Spring quarter have the front end data entry in place Digitize 12-15 real hours of audio per week Aim for total project completion within a year During Summer employment, digitize 30 hours of audio per week By Summer have the online catalog and web presence coded For Summer of 2008 address concerns of physical long term storage
Conclusion and Final Comments This document is a welcome first step in the preservation of the Language Learning Center's Spider Room collection. With over 1000 reel-to-reel tapes, we have much work to do digitize this huge amount of audio recordings, but this document presents an overall project outline, some motivation and a work flow. We have already worked towards our goal of creating an online catalog for these materials, for cataloging their existence in a web accessible database, and for digitizing and physically storing the tapes for long term access. Copyright and rights are marginally outside of this document, though we acknowledge that they will be need to be addressed at a later date, which we plan to do. This manuscript will also become part of the LLC Wiki, its internal knowledge base for Research Assistants and Staff. We have also begun producing a Website for this project and for the online catalog. It is available at: http://students.washington.edu/stiv/spider/.
REFENENCES Bucknum, Mary Russell. 2001. Music Sound Archives in the United States. Fontes Artis Musicae. October-December 48/4. Gordon, Raymond G., Jr. (ed.), 2005. Ethnologue: Languages of the World, Fifteenth edition. Dallas, Tex.: SIL International. Online version: http://www.ethnologue.com/. Hunter, Gregory S. 2003. Developing and Maintaining Practical Archives: A How-To-Do-It Manual. Second Edition. How-To-Do-It Manuals for Librarians. Number 122. Neal-Schuman Publishers. New York. IASA. 2005. The Safeguarding of the Audio Heritage: Ethics, Principles and Preservation Strategy. Technical Committee: Standards, Recommended Practices and Strategies. IASA-TC 03. Version 3, December 2005. McWilliams, Jerry. 1979. The Preservation and Restoration of Sound Recordings. American Association for State and Local History. Nashville, Tennessee.
Appendix A Reel-to-reel operations instructions (under development) Machine: Pioneer RT-707 for 3-3/4 inches per second (aka ips) Unit is connected to patch bay. (Currently swapping cable between Pioneer and 7 1/2" ips) Turn on power (upper left) Thread machine with tape align rocket ship flanges to slip tape on (if tape were stored properly; tails out -- then put tape on right reel) put reel on push all the way pull rocket ship flange out and adjust unit in place under cap stand over pinch roller - make sure it doesn't go over drive mechanism (can potentially scrap tape on drive) across heads over pinch roller under cap stand SETTINGS 19/9.5 CMS 19 = 7 1/2" 9.5 = 3 3/4" TAPE (out) SOURCE (in button) want to monitor TAPE to listen BIAS -- EQ **NEVER WANT RECORD MODE ON** Make sure pitch control is in the center detant - it clicks (Mic and line are inputs and have no effect - for recording, ganged pot) Gain control: The two output level pots in the back of the recorder should be left in the center detant position (Technically, gain is electronically measurable signal strength; volume is the human hearing system's interpretation of sound in the air.

Always mute channel 1 (the mic input) on the mixert. If the recording is stereo, unmute channels 3 and 4 on the mixer and pan channel three hard left, and channel four hard right; if mono (even if signal is present on both channels), only unmute channel 3, and set the PAN to the center position. See Audio Studio Procedures for level control setting (basically, set the slider on the mixer to unity gain (0) -- unity gain means no additional gain is being added to the signal, and the signal is not being attenuated (the signal is unaffected in strength). Then set the GAIN pot to a level that is below clipping (clipping will show up when the red PEAK LED by the fader lights up. The GAIN pot may end up being around 1pm, but it depends on the source strength. You never want the PEAK LED to light up. See mixer manual for more on level setting. Set the MAIN MIX fader/s to the appropriate level for Peak (see the Audio Studio Procedures manual). The LLC set up inappropriate procedures for storing tapes heads out. A tape is ideally played through all the way and stored on a reel tails out. This makes print through occur after the sound, and provides an even tension and pack on the tape. Since LLC tapes are stored heads out, when done playing, rewind the tape back to the supply reel by alternately pressing the rewind and fast forward buttons to reduce the speed a bit to help a bit with tape pack and tension. Press reset at first bit of audio and reset to zero When you stop tape there will be slack in tape, pull right reel Only use pause for short periods of time to keep the tape from wearing out. Maintenance: the tape recorder heads and capstan / tape path need routine, proper cleaning. The pinch rollers need proper cleaning with a proper substance. The tape heads and tape path need routine, proper demagnetization. Critical This is an example of three types of metadata found with or on the one of the Turkish tape. Please document as much of the metadata as possible. (sticky on tape)