Glossary
Short explanations of the words you will meet in the ToolKit and in translation process research.
- Allocation
- Putting participants into teams. The ToolKit proposes balanced teams; a researcher checks and commits them.
- Burst (P-burst, R-burst)
- A stretch of uninterrupted writing. A P-burst ends with a pause (usually of two seconds or more); an R-burst ends because the writer went back to revise.
- Codebook
- A file that explains every column of a data table: what it means and in which unit.
- Coding
- Labelling parts of the data by hand, for example what a translator was doing during a segment. Two coders can label the same parts so that their agreement can be measured.
- Component
- A part of the ToolKit that a study switches on, such as teams, the chat, the shared document or questionnaires.
- Consent item
- One statement in the consent form that a participant agrees to separately, such as "my chat messages may be analysed".
- Deliverable
- Something a team or participant hands in: a plan, a draft, the final translation. Each hand-in is a new numbered version.
- EDA (electrodermal activity)
- Small changes in the skin's electrical conductance, measured by a wristband, often used as a sign of arousal or effort.
- Etherpad
- A shared online document that several people can write in at once. Every change is saved, which lets the ToolKit see who wrote what, when.
- IKI (inter-keystroke interval)
- The time between two key presses. Long intervals are pauses; the pattern of intervals says a lot about how someone writes.
- Indicator
- A number computed from the data, such as the number of pauses or the share of time spent searching. The ToolKit computes hundreds; a study chooses the ones it will use.
- Intervention
- Any message a researcher or the system sends to participants. Interventions are recorded so their effect can be told apart from what participants did on their own.
- Inter-rater reliability
- How much two or more raters or coders agree. Reported as statistics such as kappa, ICC or Krippendorff's alpha.
- Keystroke logging
- Recording every key press with its time while someone writes. The basis of most translation process research.
- Lock
- Fixing your choice of indicators before you look at the results, so that the analysis follows the plan. Unlocking is possible but is recorded, with a reason, and appears in the protocol report.
- MQM
- Multidimensional Quality Metrics: a way of rating translations by marking each error with a category (for example accuracy or fluency) and a severity.
- Pad
- A shared document in Etherpad.
- Pause
- A gap in typing longer than a threshold, often two seconds. Where pauses fall (inside a word, between sentences) hints at what the writer was thinking about.
- Phase
- A period of a study, such as the first two weeks. Teams can be reshuffled between phases.
- PID (participant code)
- The code that stands for a participant everywhere in the ToolKit, such as
K7M2QA. Names and email addresses are kept apart from it. - Pilot
- A rehearsal copy of a study with test participants. Its clock can be moved forward, and its data are always marked as pilot data.
- Portal
- The participant's personal page in the ToolKit, with their tasks, team, uploads and questionnaires.
- Product/process ratio
- The length of the final text divided by everything that was typed. A low ratio means a lot was deleted and rewritten.
- Protocol report
- A draft of the methods section of a paper, made by the ToolKit from the study's own settings and history.
- Realm
- Zulip's word for a separate chat space. Each team gets its own realm.
- Release (dataset)
- A packed, pseudonymised copy of a study's data for sharing, with a codebook and checksums.
- Replay
- Playing back a session so you can watch the text being written, keystroke by keystroke.
- Reshuffle
- Changing the teams between phases, to study how people adapt to new teammates.
- Retention period
- How long participants' names and email addresses are kept after the study closes. After it, they are deleted for good.
- Ruleset
- The settings the analysis uses, such as pause thresholds. Each version is kept, so a result can always be traced to its settings.
- Segment
- In the Task Segment Framework, a stretch of activity between two long pauses, labelled by what happened in it (for example adding text or changing it).
- Session
- One sitting of work recorded by Stone Guest, from start to stop. Each session becomes one uploaded file.
- Stone Guest
- The keystroke logger participants install on their computer. It works offline and saves one file per session.
- TSF (Task Segment Framework)
- A published framework for studying keylogged translation (Muñoz Martín & Apfelthaler, 2022), with a table of 513 indicators that the ToolKit computes.
- Zulip
- An open-source team chat organised in channels and topics. The ToolKit creates the teams' chats and keeps a copy of the messages for analysis.