Challenges of Speech Data in Swedish Archives and Agencies

Speech data in Swedish archives and government agencies face obstacles like lack of collaborations with users, suitable analyses, and methods. Despite a rough inventory in 2015, materials remain unused due to IP rights, lack of responsible individuals, and sheer size issues.

• Speech data
• Swedish archives
• Government agencies
• Challenges
• Data inventory

emmacl Follow

Uploaded on Sep 20, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

Presentation Transcript

1. Speech data in Swedish national archives and government agencies Jens Edlund, KTH Royal Institute of Technology Dept. of Speech, Music and Hearing

2. About me About me Full time researcher Mainly human face-to-face interaction (humanities) so in everyday life, a CLARIN user But also Speech technology (technology/computer science) And for the purposes of this talk, a CLARIN representative

3. About KTH Speech, Music and Hearing About KTH Speech, Music and Hearing A CLARIN K-centre a centre of expertise providing an information service offering technical advice on speech analysis A research institution One of the oldest speech labs in the world (founded by Gunnar Fant in 1951)

4. Structure of this talk Structure of this talk Background Speech in general Speech in Swedish archives and agencies CLARIN obstacles (with a speech focus) Lack of collaborations with users Lack of suitable analyses and methods

5. Speech vs writing Speech vs writing Speech is often, but not always perceived as a special case of writing Speech is consistently treated like a special case of writing But Speech predates writing Speech is the most commonly occurring form of language Writing is a special case of speech? In practice, there are many similarities, but the differences are huge Speech Writing

6. Some salient differences Some salient differences Speech is transient It exists only in the present This is true, in a sense, even if recorded Speech is largely interactive and emergent It is created, edited, and undrstodd dynamically This is true for read speech as well

7. Speech in Swedish archives and agencies Speech in Swedish archives and agencies Rough inventory performed in 2015 Direct interviews with ~25 data holders, indirect with another ~25 Key finding There is a lot of materials around Nobody uses these materials at all Virtually none of it can be made fully public (easily) Obstacles IP rights is just the beginning (other legislation) Lack of responsible people Lack of descriptions Sheer size

8. How to find users How to find users To start with, don t call them users ? Start with existing methodologies and the needs that come out of this Autumn 2015 workshop in Stockholm Discussions in triplets : a researcher, a data holder, and a language tech representative Resultet in several pretty mature project ideas A large project application combining three research tracks oral history, language change, human behaviour in interaction (under review)

9. How to find users (2) How to find users (2) Interdisciplinary themes Similar to the CLARIN + efforts Initial theme for SWE-CLARN: food Enormous interest from a wide range of researchers We only just started

10. How to analyse archive speech How to analyse archive speech Turn it into text Automatic transcriptions Manual correction Annotation Text storage And dig the audio back down again

11. But But Current methods are not designed for this type of speech Text is not speech Different studies call for different analyses There is a very real danger in standardizing too soon

12. Analysis as an iterative process Analysis as an iterative process Automatic transcription Produces (erroneous) machine transcriptions Manual correction Produces (correct) manual transcription So we have the sound, a negative example, and a positive example This is very good for training Take home message: don t throw away materials that can be used to improve the automatic methods

13. Data for speech analysis development Data for speech analysis development For example Parametrized speech Statistics Word (and n-gram) frequencies Pronunciation variation Huge amounts are needed Expensive to get, even with data access Suggestion: build in data taps in various processes

14. Tapping training data from processes Tapping training data from processes Example 0: Automatic transcription Example 1: Transliteration Example 2: Digitization of text Example 3: Digitization of speech

15. Thank you! Thank you! Questions (or feel free to ask me in the breaks)