InterView¶

InterViewlogo placeholder

A platform for conducting survey interviews in virtual reality. Repository · Technical documentation

An interviewee wears a VR headset and sits across from an avatar in a shared virtual room; the interviewer sits at an ordinary web browser. A live audio and video link connects them, the questionnaire is embedded in the environment itself, and everything that happens (speech, gaze, posture, hand movement, answers given) is recorded into one time-aligned dataset.

First-person view from the interviewee's position: the interviewer's avatar opposite, with a panel on the desk showing five German answer options and a self-view mirror

The interviewee's view. The answer options for the current item sit on a panel on the desk, numbered, with a self-view mirror at the right. They are never read aloud, so any spoken answer containing an index is evidence the participant read the interface, which 36 % of answers did.

Components¶


InterView-VR	Unity application for Meta Quest, built on Va.Si.Li-Lab. Renders the room, drives the embedded questionnaire, and logs gaze, facial expression, hand and body tracking.
InterView-W	Static browser client: an interviewer console and a non-VR participant fallback. Device selection, virtual background, live chat, gaze overlay, answer-option and recording control.
InterView-B	Backend: a Janus SFU plus optional TURN relay, the Va.Si.Li-Lab room server, and the logging API everything is written through.
InterView-P	Post-hoc processing: Whisper transcription, gaze resolved against scene geometry, questionnaire import, and linkage into one relational dataset.

A session¶

A participant is issued a token. Entered in the VR client, it deterministically derives the media room, the VR room and the questionnaire URL, so headset and browser meet in the same place with no coordination step. The interviewer opens the same token in a browser and joins. Recordings land on the media server; all tracking and event data lands in the database, keyed by that same token.

Gaze is one of the channels captured this way, then replayed into the scene for analysis; see Studies for what it shows.

The interview room with a heat map overlaid, showing where gaze concentrated during a session — **Gaze, portrayed.** Aggregated gaze rendered back over the scene for a session.

Running it¶

git clone https://github.com/texttechnologylab/InterView.git
cd InterView
cp .env.example .env      # database credentials and an API key
docker compose up -d

Read the setup guide first

The stack requires an external database and a TLS-terminating reverse proxy, and will not work without them. See the setup guide.

Full technical documentation (architecture, services, operations) is at texttechnologylab.github.io/InterView; Unity and VR-client documentation is in the Va.Si.Li-Lab docs. This page is an overview, not a replacement for either.

In the browser¶

InterView is not only a VR application. InterView-W, the interviewer console and non-VR participant view listed above, runs in an ordinary browser with no headset at all. It was not part of the published evaluation study, but it is the interface actually in use right now.

Screenshot placeholder
InterView-W browser client

The browser client. Device selection, live chat, questionnaire and recording control, usable from a laptop with no VR hardware. This is the interface the ongoing NEPS validation study is currently fielded on; see Approach.

See Publications for how to cite it.