Illustrating Translation Ability Cognitively: Process, Parameters and Patterns

Description

This repository hosts the multi-method dataset and minimal analysis pipeline for "Illustrating Translation Ability Cognitively: Process, Parameters and Patterns" examining novice (n=11), intermediate (n=11), and advanced (n=3) translators performing bidirectional Chinese–English translation on matched news texts. It contains Translog-II keylogging logs (CE=FT, Chinese-to-English; EC=BT, English-to-Chinese), raw desktop eye-tracking streams (CSV), and fNIRS wavelength files (.wl1/.wl2), organized by Group A/B/C, plus expert stimulus evaluation and indexing files (participants.tsv, sessions.tsv). Reproducible scripts compute core process metrics (orientation duration, keystrokes per minute, pause frequency >200 ms, mean pause duration >1 s) from keylogs, generate summary plots, produce a Latin-square task order, and provide examples for fixation summaries (FC, TFD, RC) and fNIRS preprocessing toward MBLL-based HbO estimation. The data and code support the paper’s findings on ability gradients and directionality effects across behavioral and neurocognitive measures. All data are de-identified under ethics approval.

Steps to reproduce

This repository provides all materials required to reproduce the core process‑oriented analyses reported in “Illustrating Translation Ability Cognitively: Process, Parameters and Patterns.” The dataset comprises three modalities—keylogging, eye‑tracking, and functional near‑infrared spectroscopy (fNIRS)—collected from novice, intermediate, and advanced translators performing bidirectional Chinese–English translation on matched news texts. Data are organized by proficiency group (A/B/C) and translation direction (Forward: CE; Back: EC), with de‑identified participant and session indices to support aggregation across modalities under the approved ethics protocol. Construct a balanced task order for two conditions (CE and EC) using a Latin‑square rationale. Within each proficiency group, alternate the order assigned to successive participants so that half begin with CE and half begin with EC. For keylogging, segment the translation process into initial orientation, drafting, and revision/monitoring. Derive initial orientation time (To), drafting time (Td), revision/monitoring time (Tr), and total time (Tt). Compute pause‑based indicators using thresholds at 200 ms (surface‑level hesitation) and 1 s (deep pauses), including pause frequency and mean deep pause duration. Calculate keystrokes per minute over the drafting interval to index output fluency. Aggregate all indicators at the participant level by direction (CE/EC) and summarize distributions by proficiency group. Visualize results with distributional summaries and confirm that values are plausible, flagging extreme outliers or incomplete sessions for review. For eye‑tracking, establish quality criteria (e.g., validity flags, sampling stability), then derive fixation events using dispersion and minimum‑duration criteria aligned with accepted practice. Compute fixation count, total fixation duration, and revisit count within each stage defined by the keylogging segmentation. Summarize these indicators by direction and group, and, where feasible, map attention to regions of interest to interpret strategy use during orientation, drafting, and monitoring. For fNIRS, transform wavelength intensities to concentration changes in oxygenated and deoxygenated hemoglobin using a standard Modified Beer–Lambert pipeline with appropriate device parameters, followed by noise attenuation and motion handling. Average oxygenated hemoglobin over composite prefrontal‑language channels to obtain participant‑level summaries by direction. Confirm signal quality and exclude channels or sessions that fail basic criteria before group aggregation. Conduct inferential analyses for the 3×2 design (group × direction) using appropriate models (e.g., mixed analyses of variance), report effect sizes, and assess model assumptions. Compare patterns across modalities: keylogging (orientation, fluency, pauses), eye‑tracking (attention allocation), and fNIRS (neural activation).

Institutions

• Huaqiao University

  Fujian, Quanzhou

Categories

Funders

• The National Social Science Fund of China

  Grant ID: 20AYY003

Licence