Testimony Transcription at the Archive

A note from Christy Bailey-Tomecek, AV Access Services Librarian, Beinecke Rare Book and Manuscript Archive:

Since 2017, the Fortunoff Archive has been steadily transcribing our collection of over 4,500 testimonies. Transcription is inherently an imperfect representation of a video testimony: it leaves out visual cues, what Laurel Vlock called “demeanor evidence,” as well as inflections of speech, tone, and emotion. The archive feels strongly that researchers should refer to the original document, the video testimony, not the transcripts as the definitive source, but it is impossible to deny that transcripts allow a layer of searchability that goes beyond keywords and controlled vocabularies. Moreover, with over a dozen different languages, being able to search in the native language is an enormous benefit to researchers with the necessary language skills. It is also important that we make testimonies accessible to viewers that are deaf or hard of hearing, or have auditory processing disorders.

The process of transcription at Fortunoff Archive is a complex, multi-step effort. Much of the initial work is done using speech-to-text AI applications, also known as audio speech recognition (ASR). These applications have provided us with accessible transcripts, but not perfect ones. ASR applications can struggle with video testimonies due to poor audio quality, strong accents, and the inevitable switching between languages so common to how survivors recall life events. This will cause “hallucinations,” or text being created that isn’t what the survivor said. Therefore, the second stage is for the transcripts to be corrected. Corrections are performed either by archive staff (both permanent and student staff with knowledge of the necessary languages), fellows and senior scholars, consultants, or by classes working on transcription for assignments. For instance, classes at the College of New Jersey were assigned a set of testimonies to correct, which helped them learn about the survivors’ life stories, and their historical and cultural contexts. Some transcription has been supported by grant projects, such as a grant from the Material Claims Conference, which funded the transcription of dozens of Hebrew testimonies. To date, the archive has transcripts for over 1,200 testimonies in 15 languages, with more being produced regularly.

Some testimonies are in languages that are not covered by the conventional ASR programs, including Yiddish, Romanes, and Ladino. Moreover, some of these languages, like Ladino, have relatively few native speakers. Their “endangered status,” is a direct reflection on the cultural and linguistic toll of the Holocaust. The archive is therefore prioritizing projects to transcribe these testimonies – which are not only important primary sources for Holocaust research, but linguistic treasures in their own right.

We have now produced transcripts in all three languages. Yiddish is a Jewish vernacular language that has its origins in Central and Eastern Europe. Yiddish has a mixture of German, Slavic, and Semitic (Hebrew and Aramaic) elements. Prior to the war, there were roughly 11 million Yiddish speakers, both in small towns and large cities. Around 5 million of those speakers were murdered in the Holocaust. Fortunoff has 40 testimonies in Yiddish and is partnering with scholars fluent in Yiddish to manually transcribe these testimonies. An ongoing effort, a few of the testimony transcripts are already available here.

Ladino, also known as Judeo-Spanish, is a Jewish vernacular language that is spoken by Sephardic Jewish communities throughout the Mediterranean. Ladino is an offshoot of Castilian Spanish, with Turkish, Portuguese, and Semitic influences. In the 20th century, use of Ladino was already declining due to nationalist and antisemitic movements, but the Holocaust accelerated this cultural destruction. It is estimated that there are 60,000-300,000 present-day Ladino speakers. Fortunoff partnered with Ladino language scholars at the University of Thessaloniki to manually transcribe all 19 Ladino testimonies in the collection. They are available here.

Romanes is spoken by Roma communities around the world. Since those communities are diasporic, Romanes in Europe has strong regional differences based on the surrounding languages, and multiple writing systems. Roma communities have long faced persecution. Between 250,000-500,000 Roma were murdered in the Holocaust. There are roughly 5 million Romanes speakers today and Fortunoff Archive is cooperating with the European Roma Institute for Arts and Culture (ERIAC) to manually transcribe all of the archive’s Romanes-language testimonies, as well as translate them into English. You can view the results here.

Fortunoff Archive is dedicated to transcribing its entire collection to enhance accessibility and searchability of our testimonies, as well as meet the needs of international scholars working with these important sources in their own languages. But we are also proud to promote research and education with and in these languages, as well as the legacy of the survivors who chose to speak in their native languages when they shared their story with the archive.

Photograph above is of physical indexes.