Blog report
A Short Report of the workshop "The Oral History Transcriptions Chain"
Pre-history
At a general CLARIN-Oral History workshop held in Oxford in April 2016, a number of conclusions could be drawn:
- There is a growing interest from different CLARIN member-countries to use modern Human Language Technology for opening up Oral History data in archives that has not been transcribed, and for speeding up the process of creating transcripts from one’s own interviews.
- An infrastructure containing both the tools to perform these tasks and storage to stimulate the reuse of the transcripts, that requires no extensive knowledge of ICT by its users, is very much desired.
- The accessibility of OH collections in the participating counties is widely divergent. Some have a national infrastructure with storage centres, and online catalogues with extensive metadata, on digitized content, others have vast amounts of OH collections but only in their original analogue format.
As digitalisation of analogue recordings is becoming increasingly easier, storage is affordable, and HLT performs well enough to be applied on data with speech that is clear and understandable, it was decided to explore the possibilities to create such an infrastructure.
The first call in CLARIN (EU) was used to propose two preparatory workshops one in Utrecht which was just held, and one in Arezzo in April 2017, that should lead to a concrete proposal for an infrastructure that facilitates the entire process of opening up of oral history data,: from digitisation to transcription, making use of automatic speech recognition as intermediary technology.
Targets in Utrecht
The goal of the workshop in Utrecht was setting up a workflow and division of tasks to perform in the next four months, and that will be finalized in Arezzo, :
- What are the tasks that have to be covered to create the proposal?
- Who will be responsible for the content of each part of the proposal ?
- Who are the co-authors of each topic?
- What infrastructure will be used to describe the topics (snap ik niet)
- Setting the deadlines for the tasks
- Envision the program for Arezzo, (who to invite, what kind of activities to organize)
- What should be the follow up of Arezzo?
Day 1
At the end of the morning of December 6th, the foreign guests from England (Martin Wynn) and Italy (Silvia Calamai & Riccardo del Gratta) arrived together with the Dutch participants (die ook noemen) at the SURF-offices in Utrecht. At 13:30 Henk van den Heuvel welcomed everybody, and explained the goals of the workshops and of the proposal to be realised in spring 2017.
After which, Willemien Sanders, currently involved in the follow-up of the Oxford-workshop on behalf of Clarin EU, gave a short overview of what action points had been formulated in achieved in Oxford.
This was followed by Henk van den Heuvel’s presentation of the envisioned time table, as shown below.
Time scheme for the period between the two workshops
Table with workpackages and responsibles
The outcome of this discussion was that 8 tasks were formulated for the various building blocks of the envisaged transcription chain:
| Topic |
Responsible |
Co-writers |
| AD & DD conversion |
Arjan |
John |
| ASR |
Henk |
Louis ten Bosch, Arjan,Stef, Maurizio Omologo, Piero Cosi, Francesco Cutugno, Sergio Grau Puerto |
| Transcription correction |
|
|
| - General |
Arjan |
Stef |
| - Crowdsourcing |
Oana |
Arjan |
| Transcription Guidelines |
Stef |
Silvia, Louise |
| Alignment |
John |
Silvia, Arjan |
| Metadata |
Stef |
Dieter van Uytvanck (technical), Silvia (metadata selection) |
| Export |
Tba |
|
| Hosting |
NL: René
UK: Martin
IT: Riccardo
|
NL: Roeland (NSIV, UTwente)
UK: Thomas Hain (Sheffield), John (Computer Cluster, Oxford)
IT: Silvia (Arezzo/Siena)
|
| Overview of the task and the people working on them |
After the coffee break three interesting presentations were held by:
- Rene van Horik (DANS) about potential data hosting via EUDat.
-
Oana Inel (VU) on the Crowdsourcing platform CrowdFlower. Oana showed us how easy it is to setup a Crowdsourcing environment where "other" people (you do not know them), can help you with the correction of transcriptions (made by humans or machines). A small example of a 1-minute audio segemnt and the accompanying transcription can be seen here (un/pw = /crowdtruth).
- Roeland Ordelman (NISV) who offered an overview of similar activities for a transctiption service as part of the WP5-workpackage of the infrastructure program CLARIAH
The first day was ended with a nice diner at Mr. Buscourr
Day 2
The second day was mainly focused on the organisation of the Arezzo-workshop: when, where, who, what?
Because Arezzo does not has an airport, visitors need to fly to Pisa, Firenze or Rome and than take the train to Arezzo (additional 1 to 2½ hour extra travel time).
So starting early in the morning or ending at the end of the afternoon is a bit problematic without an additional night. Stef and Willemien have offered to “solve” this issue.
Further, the responsibilities of the participants in terms of document writing and implementation of tasks were verified and settled.
At 12:30 this short, but very effective workshop was concluded and participants left with a list of tasks to perform in the coming 4 months ?.
Arjan & Henk
