Transcription is always the result of choices. Its adequacy depends on the type of data and the research questions being addressed. The first step is therefore to delimit the field: what aspects of the interaction are relevant to your analysis, and at what level of detail do you need to capture them? A highly granular transcription — one that captures speech, prosody, gaze, gesture, posture, and object manipulation simultaneously — makes the data reusable for many purposes, but produces a transcript that is considerably more complex to read. A more selective transcription is easier to follow but may omit features that turn out to be analytically relevant later.

This tension between comprehensiveness and specialisation is one of the central methodological considerations in CA transcription practice. A useful guiding principle: if in doubt, transcribe more rather than less — it is easier to omit tiers at the reading stage than to re-transcribe a corpus from scratch.


A. Decide on the tiers you will create in ELAN

In ELAN, different dimensions of the interaction are organised into tiers — separate annotation layers that run in parallel and are aligned to the same time axis. Tiers can be given names, associated with specific participants, and arranged into hierarchies (a gloss tier may depend on a speech tier, for instance). It is also possible to add a transcriber comments tier for observations that should not be confused with the primary data.

B. Create the project and import media
Open ELAN and create a new annotation file (.eaf). Link your audio and/or video file(s) to the project: ELAN will synchronise the media with the annotation timeline. It is possible to link multiple media streams simultaneously, which is particularly useful for multi-party interaction data. The media files are not embedded in the .eaf file but referenced by path, so keep your folder structure stable throughout the project.

C. Define the tier hierarchy
Create your tiers according to the plan established in A. For each tier, specify: its name, the participant it is associated with (if applicable), its type (time-alignable, or dependent on a parent tier), and any linguistic type or vocabulary constraint you wish to apply. ELAN supports hierarchical relationships between tiers.

Assign clear, consistent names to your tiers from the outset. Renaming tiers later is possible but can create complications if tier names are referenced in templates or export settings.

Annotate speech, prosody, non-verbal behaviour, and paralinguistic features. 

A. Transcribe speech using Jefferson conventions
Enter the verbal transcription in the speech tiers, one turn at a time, following the Jefferson system (G. Jefferson, Glossary of 'Transcript Symbols with an Introduction', in G.H. Lerner (ed.), "Conversation Analysis. Studies from the First Generation", Amsterdam/Philadelphia, John Benjamins, 2004, pp. 13-31.). The transcript is laid out vertically according to turn sequence; each turn is associated with a speaker identifier and, in ELAN, aligned to a time interval in the recording. Overlapping talk is marked with square brackets [ ] at the point where it begins and ends across the relevant tiers. It should be noted, however, that these brackets are not automatically generated by ELAN but are added manually. This point is important because conversation analysts working within the Jeffersonian transcription tradition—where the alignment of overlapping speech is visually reproduced on the page using vertically synchronized square brackets—may expect a comparable output. In ELAN, overlap is instead primarily represented through separate tiers and synchronized timestamps. Reproducing the conventional Jefferson-style layout therefore requires exporting the transcript (for instance into Word), using a monospaced font such as Courier, and often manually adjusting the alignment of the strings. 
Prosodic and voice quality features are noted inline using the standard symbols.

ELAN allows you to play short loops, slow down the audio, and zoom into the waveform display while annotating, which considerably aids the transcription of rapid, overlapping, or unclear speech.

B. Transcribe non-verbal behaviour
For audio-visual data, annotate non-verbal actions — gesture, gaze, posture, object manipulation — in the corresponding tiers, following the Mondada conventions. Each embodied action is described in plain language within angle-bracket-delimited strings, synchronised with the speech tier via time codes and shared symbols. The symbols mark the preparation, apex, and retraction phases of a movement, and indicate whether the action begins before the extract, continues beyond it, or is captured at a precise moment by a video frame reference (#).

Non-verbal actions often do not map neatly onto turn boundaries: a gesture may begin several syllables before the relevant word, or continue after the turn has ended. ELAN's multi-tier, time-aligned interface is designed precisely to capture this temporal complexity.

C. Add a working translation if needed
For multilingual data, or when the corpus will be shared with readers unfamiliar with the language of the interaction, it is useful to add a translation tier aligned to each speech tier. This working translation is not part of the primary transcription data but serves as an access tool. It should be clearly labelled as a gloss and not confused with the transcription itself.

A. Review the transcription against the recording
Once a first version of the transcription is complete, go through it systematically against the recording. Check the accuracy of verbal transcription, the alignment of time codes, the placement of overlap brackets, and the descriptions of non-verbal actions. Transcription is an iterative process: it is normal to revise notation choices as familiarity with the data deepens and analytical hypotheses take shape.

B. Ensure internal consistency
A well-formed transcription requires a one-to-one correspondence between symbol and phenomenon: the same symbol should always represent the same feature, and the same feature should always be represented by the same symbol. Review your transcription for inconsistencies — particularly in the notation of overlaps, pauses, prosodic features, and non-verbal descriptions — and harmonise them. Any residual inconsistency should be noted in the transcriber comments tier.

ELAN's native format is .eaf (EUDICO Annotation Format), an XML-based file that preserves the full tier structure, time codes, and media links. This is the recommended archival format. For sharing or publication purposes, ELAN also supports export to a range of other formats.

Letizia Cirillo

This workflow outlines the main steps involved in the transcription of conversational data using ELAN, a software environment designed for the annotation and synchronisation of audio and video recordings. ELAN is particularly suitable for multimodal transcription, as it allows the annotation of both verbal and non-verbal phenomena — such as gaze, gestures, pauses, and interactional dynamics — through multiple time-aligned annotation tiers.

The workflow includes the preliminary design of the transcription scheme and annotation hierarchy, the configuration of tiers and segmentation units, the automatic or manual transcription and annotation process, and the exportation of the final transcript in EAF or other formats. 

This workflow is based on Letizia Cirillo’s chapter “Trascrivere la conversazione” (pp. 181–188), part of the "Vademecum per il trattamento delle fonti orali" (Tavolo permanente per le fonti orali, 2020).

Edit actor

Other URLs relevant to the resource, e.g. ORCID, etc.

URL, ideally landing page, of the resource.

Person/group/body who helped create the resource.

Should be concise and raise interest in the entry. Allows <Link>Markdown</Link> content.

Other URLs relevant to the resource, e.g. code, DOI, etc.

Upload media

URL pointing to info about the access policy.

The activities you can do with the resource. See <Link></Link>.

Name of the conference where paper was given.

Describes the discipline covered by resource. See <Link></Link>.

Locations where the Resource is offered. See <Link></Link>.

URL to helpdesk for incidents & user requests.

Audience targeted by the resource. See <Link></Link>.

A particular published issue of a journal.

Concept or term related to MP entry. See <Link></Link>.

Language(s) in which a resource is available. See <Link></Link>.

Select license pertaining to resource. See <Link></Link>. If license is unknown, use the "Terms of use" property field to describe the usage terms.

Status of the Resource life-cycle. See <Link></Link>.

Mode of use for the resource. See <Link></Link>.

File format of the linked resource. See <Link></Link>.

Page #s of a resource, separate by commas.

Refine the resource's nature and function. See <Link></Link>.

Links to non-MP materials that are relevant.

Info about performance levels provider expected to deliver.

Standard used by the resource. See <Link></Link>.

Technology Readiness Level of the Resource. See <Link></Link>.

Resource user manual and documentation URL.

Version of the tool or service that is linked.

The volume of a journal or multivolume book.

Edit source

Name

If you are having trouble with the sign in process, please <ContactLink>contact the SSH Open Marketplace team</ContactLink>.

If you are having trouble with the sign up process, please <ContactLink>contact the SSH Open Marketplace team</ContactLink>.

Transcribing conversation in ELAN

Related items(4)

Vademecum per il trattamento delle fonti orali

Treating Oral Sources in SSH Research

Transcription de données de langage oral

ELAN

Workflow steps(5)

1 Determine what you need to transcribe

2 Configure ELAN

3 Import a transcription or transcribe

4 Review and normalise

5 Export your transcription