Preserving interactive data visualisations from conference talks

Conference presentations increasingly depend on interactive charts, maps, dashboards and computational demonstrations. A speaker may move a slider, filter a dataset, zoom into a spatial layer or reveal a model’s changing output. Once the event ends, however, that experience can disappear even when the conference programme and presentation slides remain available.

This problem is especially visible in archived conference websites such as IASSIST 2017, which documents “Data in the Middle: The Common Language of Research”, held at the University of Kansas in Lawrence from 23 to 26 May 2017. Its programme, plenary information and practical attendee guidance remain useful, yet a visualisation shown during a talk may have relied on a temporary server, a proprietary library or an unrecorded sequence of live actions.

Archiving interactive data visualisations from conference talks therefore requires more than saving screenshots. Preservation must retain the visual appearance, underlying data, software dependencies, explanatory context and steps required to reproduce the interaction. It should also make clear what has been preserved faithfully and what has been simplified for long-term access.

For Australian libraries, universities and research offices, this work sits within a familiar environment of public research funding, open-access expectations and privacy obligations. A good archive should serve researchers in Canberra, students in Melbourne, community organisations in Perth and members of the public accessing material on a phone during a commute in Sydney or Brisbane.

Why interactive visualisations need special treatment

A static PDF can preserve a chart’s labels and colours, but it cannot show how the chart responds to a user. A web map may contain several layers that appear only after a selection. A deep-learning demonstration might display changing classifications, confidence scores or training epochs. If the archive saves only a slide image, the intellectual argument behind the visualisation may be reduced to a decorative snapshot.

Interactive material also has a fragile technical life. JavaScript frameworks are updated, external application programming interfaces are retired, domain names expire and browser security rules change. A visualisation embedded in a conference recording may fail because its data request points to an unavailable endpoint. Even a file that works today may behave differently after a library upgrade.

The challenge is interpretive as well as technical. A future researcher needs to know whether a map was filtered to a particular region, whether a dashboard displayed real-time information or whether a presenter manually changed a parameter. Capturing those actions preserves the evidence of the talk rather than merely its final appearance.

What an archive should capture

The first layer is the presentation context: speaker, title, institution, conference, date, session, abstract and relationship to the wider programme. The archive should connect the visualisation to the talk recording, slides, transcript and any supplementary paper. A data librarian’s perspective is especially valuable in sessions dealing with agricultural, environmental or social datasets; the discussion of global food security illustrates why context can be as important as the interface itself.

The second layer is the visualisation package. Preserve the source code, configuration files, data files, stylesheets, fonts and media assets where rights permit. Record the software environment, including language versions, package dependencies, database engines and build instructions. A container image or virtual machine can help retain a working environment, although it should accompany human-readable documentation rather than replace it.

The third layer is an interaction record. This may include a screen capture, narrated demonstration, event log, test script or a written sequence such as: select “Queensland”, set the year to 2015, enable the rainfall layer and hover over the Darling Downs region. For complex research software, a short video plus a machine-readable interaction log provides a useful compromise between authenticity and practicality.

Preservation metadata should identify the original files, their checksums, creation dates and relationships. It should also describe data provenance, processing steps, geographic coverage, units of measurement and known limitations. A future user should be able to distinguish an original dataset from a later normalised copy or a screenshot generated by the archive.

A workable preservation workflow

Begin before the conference, where possible. Ask presenters to submit a preservation bundle with their slides, source files, datasets, licensing information and a short demonstration script. A simple template can request the visualisation’s purpose, required inputs, external services, expected outputs and preferred citation. This approach is more reliable than trying to reconstruct a demonstration from memory months later.

During the event, record the talk and capture the visualisation at meaningful points. A full screen recording should include the browser address bar or a visible project identifier when appropriate, while a separate camera feed can preserve the speaker’s explanation. The archive team should note any live data feeds, manual interventions or unexpected failures. These notes are often essential when a later user sees a slightly different result.

After the event, create a preservation master and an access copy. The master may contain original source code, high-resolution recordings and unmodified data. The access copy can use compressed video, a static preview or a hosted replay that is safer for general visitors. Keep a manifest linking each asset to the talk and record fixity information so later integrity checks can detect accidental changes.

Validation should take place on more than one browser and operating system, including the mobile devices commonly used by Australian audiences. Test keyboard navigation, captions, alternative text and colour contrast. If the original interactive experience cannot be restored, publish a transparent fallback: for example, a series of annotated screenshots, a downloadable dataset and a narrated video showing the principal states.

Choosing formats that can survive change

There is no single preservation format for every visualisation. Open, documented formats generally offer better long-term prospects than opaque project files, but the best package usually combines several formats. A JSON data file may preserve structured values, while CSV supports straightforward inspection. GeoJSON can represent many map layers, and plain-text source code remains easier to migrate than a compiled application.

Web archives can capture HTML, CSS, JavaScript and local assets, but they may struggle with authentication, cross-origin requests, streaming services and data loaded after a user action. A self-contained Web Archive file can be useful, yet it should be tested as a replay rather than assumed to be complete. Screenshots and video remain important because they preserve visual evidence even when executable replay fails.

Preservation need	Useful capture	Long-term value	Common limitation
Visual appearance	PNG, SVG, PDF or high-resolution video	Shows labels, layout and styling	Cannot reproduce interaction
Source and structure	HTML, CSS, JavaScript and configuration files	Supports migration and technical study	Dependencies may age quickly
Underlying data	CSV, JSON, GeoJSON or database dump	Enables analysis and verification	Sensitive or licensed data may be restricted
User behaviour	Screen recording, transcript and event log	Records filters, sequences and responses	Manual descriptions can omit detail
Reproducible environment	Container, virtual machine or lockfile	Helps rebuild the original system	Requires maintenance and specialist skills

A preservation description should explain which layer is authoritative for which purpose. The video may be the authoritative record of what the audience saw, while the source repository may be authoritative for how the visualisation was built. Clear distinctions prevent users from treating a reconstructed replay as an unchanged original.

Rights, privacy and community authority

Copyright applies to code, written explanations, graphics, recordings and some datasets. Conference organisers should obtain permissions that cover preservation, migration, institutional hosting, public replay and educational reuse. A presenter may permit the slides to be open while restricting the source code or commercial dataset. Those conditions should be recorded in machine-readable rights metadata and displayed plainly to users.

Australian institutions also need to consider the Privacy Act 1988 and applicable health, education and research requirements. A dashboard may expose names, precise locations, small-area statistics or identifiable participant information even when the presenter originally treated it as a demonstration. Redaction, aggregation, access controls or a synthetic dataset may be necessary for the public version.

Indigenous data requires additional care. Australian archives should consult relevant Aboriginal and Torres Strait Islander communities and apply principles such as authority, responsibility and ethics in data governance. A technically successful replay can still be inappropriate if it exposes cultural knowledge, community locations or data outside the permissions under which it was collected.

International material can create further complications. A project involving food security or agricultural change may include data supplied by overseas agencies, local communities or partners with different expectations. A cross-border example such as Peru-based data work should be preserved with its original attribution, licence and cultural context rather than detached from the people and institutions represented.

Making the archive usable after the event

A replay should open with a clear description of what the user is seeing. State whether the original server is running, whether the data has been frozen at a particular date and whether a substitute dataset is in use. A visible “preservation status” note can distinguish live, reconstructed, partially functional and documentary-only versions.

Accessibility belongs in the preservation design, not as a later enhancement. Provide captions for recordings, transcripts for spoken demonstrations, text alternatives for important charts and keyboard-accessible controls where possible. For a map, describe the geographic pattern in words and provide the underlying values or a downloadable data file. This supports users who rely on screen readers, low-bandwidth connections or printed material.

Australian users may encounter practical constraints that differ from those of the original conference audience. Regional researchers can face slower connections, while people using public Wi-Fi may avoid large downloads. A lightweight static preview, a compressed recording and an option to download assets for offline use make the archive more resilient. The same principle helps a student accessing material on a train or a researcher working during a temporary network outage.

Search and citation features also matter. Index the talk title, presenters, datasets, locations, methods and subject terms. Assign a persistent identifier to the visualisation package and cite the version date. A researcher should be able to refer to “the 2017 replay using rainfall layer B” rather than an unstable page whose content changes without notice.

Maintaining a living preservation record

Interactive archives need scheduled review because their dependencies and hosting arrangements change. An annual or biennial check can test links, playback, file integrity, accessibility and metadata. Record the result of each check, including failures. A broken replay is more useful when its last working date and available fallback are documented than when the archive silently presents an error page.

Use a layered storage strategy. Keep at least one managed institutional copy, a geographically separate backup and a preservation package in a trusted repository where possible. Australian universities may coordinate through library repositories, research data services or national infrastructure, while smaller organisations can use documented export packages and offline checksums to reduce dependence on a single vendor.

A maintenance record should identify the responsible organisation, review interval, storage locations, software environment and escalation path. It should state whether future migration is permitted and who can approve changes to restricted material. When a library restructures its website or a conference archive moves to a new platform, this information prevents valuable visualisations from being separated from their metadata.

Practical priorities for conference organisers

The most effective preservation plan is modest enough to use routinely. Organisers do not need to build a bespoke digital repository for every event, but they do need consistent collection practices, clear permissions and a defined handover between the conference team, presenters and archivists.

• Request source files, data, licences and a short interaction script before the presentation.
• Record the talk and capture the visualisation’s important states, filters and outputs.
• Preserve an open, machine-readable package alongside screenshots, captions and a transcript.
• Document privacy, Indigenous data governance, copyright restrictions and any redactions.
• Test the replay on desktop and mobile connections, then publish a static fallback when necessary.
• Assign a persistent citation, checksum the files and schedule regular preservation reviews.

A conference archive becomes substantially more valuable when it preserves the decisions and actions behind a visualisation. Slides may show what a speaker wanted the audience to remember, while a documented replay shows how evidence was explored, qualified and interpreted. Together, those materials allow future researchers to study the original argument without mistaking a fragile web interface for the whole scholarly record.