Searching

The main search page at /arksearch is the primary entry point for discovering research funding. It combines semantic vector search with optional AI-powered analysis to deliver results ranked by relevance to your research description.

The Search Box

The search box is an auto-expanding textarea that grows with your input (up to 300px height). As you type, it displays a real-time character count and a clear button. The character limit depends on whether ArkAssist is enabled:

Search Suggestions

When you begin typing (2 or more characters), a dropdown appears with dynamic suggestions based on your input. When the search box is focused but empty, a panel shows your recent searches and trending keywords across the platform.

ArkAssist AI Mode

ArkAssist is an optional AI enhancement layer that requires login. When toggled on, it activates full LLM-powered analysis of your query. An animated banner indicates that ArkAssist is active.

With ArkAssist enabled, you gain access to two additional capabilities:

Bio Integration

If you have a research bio configured in your profile, ArkAssist can use it as context for personalized results. This means the search understands not just what you typed, but who you are as a researcher -- your disciplines, methods, and focus areas all factor into result ranking. Bio integration requires both ArkAssist and login.

Document Upload

Upload research documents (PDF, DOCX, or TXT) to provide additional context for your search. Uploaded files appear as chips below the search box. The system extracts key themes from your documents and uses them alongside your query text to find the most relevant opportunities. Document upload requires ArkAssist to be enabled.

Semantic Search

ArkSearch uses EmbeddingGemma-300M to convert your query into a high-dimensional vector that captures its meaning, then compares it against pre-computed vectors for every opportunity in the catalogue. A query like "renewable energy storage materials" surfaces opportunities about battery technology, hydrogen storage, and advanced photovoltaics -- even when those exact words do not appear in your query.

The practical benefit is that you do not need to guess the funder's vocabulary. Describe your research in your own words, and the AI handles the translation.

Understanding Search Results

After submitting a search, a progress indicator displays the current processing stage and completion percentage. When results arrive, the page shows:

• Result statistics -- Total number of matches with an active/expired breakdown, so you immediately know how many opportunities are still open.
• Ranked results -- Opportunities ordered by relevance, blending text-match strength with semantic similarity.
• Pagination -- Navigate through large result sets with page controls.
• Share URL -- Copy a shareable link to the current search results so you can send them to colleagues.

Tips for Effective Searching

• Be specific about your domain. "Sustainable urban drainage systems" returns more targeted results than "sustainability" alone.
• Use natural language. Describe what you are looking for as you would to a colleague -- for example, "funding for early-career researchers in computational neuroscience".
• Enable ArkAssist for complex queries. When your research sits at the intersection of multiple fields, the longer character limit and LLM analysis produce better results.
• Attach documents for precision. If you have a draft proposal or published paper that captures your research focus, uploading it gives the search engine richer context than a short text query alone.
• Iterate. If your first query returns too many results, add more detail. If it returns too few, broaden the description or remove qualifying terms.
• Combine with filters. Run a broad semantic search and then apply filters to narrow the results to what is actionable right now.