Welcome   About eScribe   For linguists   For users

About eScribe

What is eScribe?
Field testing
Comparing transcribing and respeaking
Contact
Acknowledgments
What is eScribe?

Linguistic corpora are valuable resource, both for linguists and for speakers of the world's 7,000 languages. Collection of these corpora is limited by the number of linguists willing and able to dedicate long hours collecting and transcribing naturally occurring speech. But what if corpus building was something that everyone could participate in? Imagine how much linguistic data we could collectively collect.

eScribe is an Android app designed to facilitate collecting and transcribing linguistic corpora. eScribe is designed to be easy to use, by linguists and nonlinguists alike, to collect corpora, of any size, in any language. It was designed with the goal of recording and transcribing corpora of child directed speech with the help of native speakers of understudied languages, but will be able to be used for many kinds of language documentation.


Field testing

We are still actively developing eScribe, and we are interested in feedback. We are particularly interested in feedback from linguists who try out eScribe with our intended users - non linguist native speakers of less commonly studied languages. If you try it out for your own project that involves recording and transcription of either adult or child speech, please let us know how it goes!


Respeaking versus Transcribing

As respeaking is a relatively new and untested methodology, we wanted to be sure that it was as reliable a method as text transcription in collecting corpora. In particular, we wanted to know whether this method introduced any different error patterns than traditional transcription. To find out, we conducted an experiment in which native English speakers listened to recordings of child produced and child directed speech, and either transcribed or respoke the recordings. Their transcriptions and respoken recordings are being tagged for different kinds of errors (e.g. omissions, lexical substitutions, grammatical changes), and the error rates across the two data collection methods will be compared. While data analysis is still in progress (9/2013), preliminary findings suggest that the two methods are comparable. Check back soon for a full report on this experiment.


Contact

Please get in touch with Annie Gagliardi with any feedback, comments or questions about eScribe.


Acknowledgments

eScribe was conceived of and designed by Annie Gagliardi and developed by Jessica Yao. It was inspired by Aikuma. Invaluable input on the design and functionality has come from Maria Polinsky, Keith Plaster, and members of the Polinsky Language Sciences Lab. Katryna Cadle and Jacob Louis Hoover assisted in the experiment to compare respeaking and transcription. This project was funded by the NEH-NSF Documenting Endangered Languages Fund, by funding from CASL at the University of Maryland, by the National Science Foundation, and by the Insitiute for Quantitative Social Sciences at Harvard.