Building a Faculty Research Portal on Top of Your University's Research System
How to design a faculty-level research portal that reads from the central research information system, ORCID and Crossref — architecture, data model, sync strategy, search and the governance questions to settle first.
Most Hong Kong universities already run a central research information system. PolyU, CityU, Lingnan and HSUHK use Elsevier Pure–based scholar portals; HKU runs the HKU Scholars Hub on DSpace-CRIS; HKBU’s IRIMS includes a research portal module. These systems are the record of truth for outputs and grants.
So why do faculties keep asking for their own portal? Because a central system answers “what has this person published?” while a faculty needs to answer “what is this faculty known for, and who should I talk to?” — for journalists, NGOs, industry partners, donors, prospective students and its own Dean.
This guide describes how we approach a faculty research portal that complements the central system instead of competing with it.
1. Settle governance before architecture
Three questions decide most design choices:
- Who is the system of record? Almost always the central system. The faculty portal reads from it and never becomes a second place to enter publications.
- Who edits the story? Research themes, impact case studies and highlights need named faculty editors with a light approval flow.
- What may be public? Agree with the library and research office which fields can be shown publicly — especially grant amounts, student supervision and contact details.
The library or research office usually owns the central system. Involving them in week one turns a potential blocker into your best data partner.
2. A simple, durable architecture
Central research system ─┐
ORCID public API ────────┼──► Nightly sync jobs ──► Portal database ──► Search index ──► Public site
Crossref (DOIs) ─────────┘ │ ▲
▼ │
Data quality report Faculty editors (SSO)
- Sync jobs pull from each source on a schedule. Use the central system’s API or OAI-PMH feed where available; a scheduled export works when neither exists.
- Portal database (e.g. PostgreSQL) stores normalised people, outputs, grants and themes, plus editorial content.
- Search index (e.g. OpenSearch) powers fast, typo-tolerant expert search.
- Public site is statically generated where possible — fast, cheap to host and hard to break.
- Editors sign in with university single sign-on to manage themes and stories.
3. The data model that matters
Keep it small:
| Entity | Key fields | Source |
|---|---|---|
| Person | name (EN/ZH), title, department, ORCID iD, expertise keywords, photo | Central system + faculty editors |
| Output | title, type, year, venue, DOI, authors, open-access link | Central system, enriched by Crossref |
| Grant | title, scheme, role, period (amount only if approved for public view) | Central system |
| Theme | name, description, linked people and outputs | Faculty editors |
| Story | headline, summary, body, images, linked people | Faculty editors |
Identity is the hard part. Match people by staff ID or ORCID iD, never by name alone — Hong Kong romanisation variants (Chan / Chen, Wong / Huang) will otherwise merge or split researchers.
4. Sync without surprises
- Idempotent upserts keyed by stable IDs (staff ID, DOI, output ID), so a sync can be re-run safely.
- De-duplication across sources by DOI first, then by normalised title + year.
- A data quality report after each run: new records, unmatched authors, missing DOIs. Send it to the faculty editor, not just the log file.
- Never delete on a failed sync. If a source is unavailable, keep yesterday’s data and alert.
5. Expert search people actually use
Outsiders don’t search by publication title. They type “flood risk”, “elderly loneliness” or “VR therapy”. Good expert search:
- indexes expertise keywords, theme names, output titles and abstracts — weighted in that order;
- supports both English and Chinese queries;
- ranks people by relevance and recent activity;
- shows why a person matched (the highlighted keyword or output).
You can try this behaviour in our Research Portal demo.
6. Dashboards for the Dean’s office
Leadership wants trends, not lists: outputs by year and department, grants by scheme, theme activity, knowledge-exchange events. Two rules:
- Every number links back to the underlying records — otherwise nobody trusts it.
- Agree on definitions (for example, “output year = publication year, not acceptance year”) and print them on the dashboard.
7. Accessibility, language and performance
- WCAG 2.2 AA: keyboard navigation, contrast, alt text for researcher photos, accessible charts.
- Bilingual names and theme descriptions, with sensible fallbacks when a translation is missing.
- Fast on mobile: journalists often search from a phone.
8. Rollout plan
- Weeks 1–2: governance decisions, data access, design direction.
- Weeks 3–6: sync jobs, data model, profiles and search for one pilot department.
- Weeks 7–10: themes, stories, dashboard; remaining departments.
- Weeks 11–12: editor training, accessibility review, launch.
A single-department starter can launch in 6–8 weeks; a whole faculty typically takes three to four months.
We recently delivered a faculty research portal for a Hong Kong university faculty and offer it as a packaged service. If you are weighing whether your faculty needs one, we’re happy to look at your current research pages and give an honest opinion.
Planning a project like this?
Gigantic Work builds web platforms, research portals, AI tools, Unity/VR experiences and cloud infrastructure for Hong Kong universities, grant-funded projects, NGOs and SMEs — and hands over everything at the end.