ICON 2019 tutorial

Recording and Analyzing BEHAVIORal Data in Interpretative Text Production

ICON-2019 Tutorial, Hyderabad, India, 18th December 2019.

The analysis of eye-tracking and keyboard logging protocols have become established methods for the investigation of interpretative text production, including text copying, paraphrasing and translation of texts, and have recently been adopted for spoken translation and interpretation tasks. However, this branch of empirical (translation) process research is not well known in India and little research has been conducted with Indian languages. This tutorial addresses researchers with a mix of lectures and hands-on sessions.

The analysis of independent, multi-modal behavioral data streams (reading, writing, hearing, speaking) is a challenging task that requires advanced technologies for data acquisition and synchronization, but promises deeper insights into the human cognitive processes during text production. It allows for better grounded and more general theories of the human mind. With the integration of multi-modal data streams, it is possible to investigate universal cognitive processes in written and spoken text reception and text production and to assess individual differences.

The tutorial presents methods for collecting user activity data from monolingual and bilingual interpretative text production sessions, including paraphrasing, summarization, translation and machine translation post- editing sessions. It addresses issues in the design of translation experiments, as well as recording, processing and analyzing of human behavioral data. The tutorial also introduces the CRITT Translation Process Research Database (TPRDB), a large repository of available legacy data, against which results of individual experiments can be matched and compared. We will use R (and/or python) to demonstrate some features of the TPRDB.


Session plan:

  • introduction and rational (10 mins)

  • setting up, recording, replay of a translation experiment (50 mins)

  • word alignment with YAWAT and TPRDB summary tables (30 mins)

  • rational of TPRDB tables, product and process features (30 mins)

  • correlation and regression with TPRDB data in R (60 mins)


Tutorial Material:


Organizers:

  • Michael Carl (Kent State University)

  • Jahfar Ali, (IIIT, Hyerabad)