CS Picks

CS Picks helps prospective CS PhD applicants, current grad students, faculty, and department chairs find the right program and research advisor. It's a JavaScript application for exploring computer science professors, universities, research strengths, publication trends, and NSF funding, and it accompanies the PhD Demystify book for prospective PhD students. CS Picks uses CSRankings and DBLP for ranking-compatible publication counts, OpenAlex plus manual corrections for estimated historical affiliations, and the official NSF Award Search for its funding beta.

Live at cspicks.roars.dev.

🚀 Features

1. Professor Search

• Search by Name: Instantly find professors by name.
• Publication Stats: View a breakdown of publication counts by research area, sorted by volume.
• Activity Summary: Consistency, momentum, and steadiness of a researcher's output over time.
• Direct Links: Quick access to the professor's:
  • Homepage
  • Google Scholar Profile
  • DBLP Profile

2. School Search

• Search by Name: Find universities by name.
• Area Analysis: View the school's top research areas with per-area rankings.
• Faculty Count: See total faculty contributing to each school.
• Further Analysis: Explore publication trends, area growth, conference trends, faculty diversity, publishing effort, and other profile highlights.

3. Historical Mode

• Toggle Historical Affiliations: When enabled, publications are credited to the institution where the author was affiliated at the time of publication (via OpenAlex data).

4. Ranking Simulator

• What-if Analysis: Model adding, transferring, or removing faculty and inspect the resulting overall and per-area rank changes.
• CSRankings and DBLP: Use existing CSRankings faculty or look up external researchers through DBLP.
• Shareable setup: the selected university and candidate names round-trip through the URL, so a link reproduces the same inputs one click away from a result.

5. Head-to-Head Comparison

• vs Search: Type A vs B in the main search box to compare two universities (CMU vs MIT), two professors, or two research areas (AI vs Security) side by side.
• Rank-Gap Breakdown: For universities, see the area-level log-score differences behind the overall rank gap.
• Area vs. area: growth versus the prior period, and who bridges both fields — the universities and researchers active in each.

6. Area & Conference Search

• Search by Area Name: Find universities and professors in a research area.
• Search by Conference: Find contributors to specific venues (e.g., PLDI, NeurIPS).

7. Manual Affiliation Overrides

• Community Corrections: Add corrections to public/manual_affiliations.csv to fix incorrect OpenAlex data.

8. NSF Funding Beta

• Nationwide Search: Explore institution-verified NSF awards for faculty at US universities in the current CSRankings roster.
• Fractional Funding Attribution: Divide NSF's estimated total award amount equally among every listed PI and co-PI, then aggregate matched faculty shares to universities.
• Funding Trends: Inspect award-year activity, programs, active awards, and individual award details.
• Conservative Matching: Retain an award only when its NSF recipient matches the faculty member's current CSRankings institution.

9. Discoveries

• On the Search page: the "🔭 Discoveries" nav link (index.html?view=discoveries) opens the same search shell — header, filter bar, search box, examples — but its default landing state shows the Discoveries insight cards where the university/faculty lists would normally sit. Typing a real search still works exactly like Search and replaces the cards with actual results.
• Notable, reproducible patterns: fastest-growing/shrinking research areas and universities, departments rising or falling, funding trends, faculty mobility (who's new to a field), regional specializations, and more — every card is computed from the live data, never hand-picked.
• University-level and subfield-level: one set of cards asks "which universities moved," a second asks "which research areas themselves grew, shrank, or changed leaders" region-wide.
• Stable, shareable per-card links: every card has its own #fragment URL and a Copy Link button; opening that link scrolls to and highlights the exact card.

10. CS Conference Schedule

• Conference and research-area search: Find conference dates and submission timelines with the same autocomplete behavior as the rest of CS Picks.
• Shared venue sets: Switch among CSRankings and CORE conference sets using the same definitions as publication search.
• Applicant-focused window: The schedule opens to the current and following conference year and understands multiple or rolling submission cycles.
• Self-contained data: Conference schedules live under csconfs/data/, are maintained through agent-assisted research of official conference sites, and deploy with this project; no external checkout or runtime service is required.

🔗 URL architecture & sharing

Every page keeps the URL in sync with what's on screen, so any view is a link that reproduces itself — no login, no server-side state.

Page	What's encoded
Search (index.html)	q (the search text, including A vs B), target/targetType (the selected analysis target), plus region/years/venue set/rankings/history/per-capita from the shared filter bar
Discoveries (index.html?view=discoveries)	view=discoveries, region/years/venue set/history/per-capita, plus a #fragment per card (#discovery-fastest-growing-subfields, etc.) that scrolls to and briefly highlights that card on load — or q/target once the visitor searches for something
Simulator (simulator.html)	Filters, univ (selected university), and candidates (the raw candidate names/DBLP links) — opening the link pre-fills the setup one click from a result, without re-querying DBLP on load
CS Confs (csconfs.html)	q, conference-year range, venue set, and whether only upcoming conferences are shown
NSF Funding (nsf.html)	q (search or A vs B) plus the year-range filter
Awards & Grants (grants.html)	q, audience, sponsor, topic, deadline, and sort filters

Filter choices also persist across page navigations via localStorage, so switching between Search and Discoveries, or clicking into Simulator, NSF Funding, or Awards & Grants, doesn't silently reset the region or year range.

Every page has an unobtrusive Copy Link button in the header (src/share.js): the Web Share API's native sheet where the browser offers one, a clipboard copy otherwise. Discoveries cards each get their own copy of the same control, scoped to that card's fragment. src/seo.js keeps <title>, the meta description, canonical link, and OpenGraph/Twitter tags in sync with the same state, so a shared link's title and social preview describe the actual view, not just the generic homepage.

📈 SEO & social previews

• Every page ships baseline <title>, meta description, canonical link, and OpenGraph/Twitter card tags in its HTML <head>, so a crawler that never runs JS still sees something accurate.
• src/seo.js sharpens those tags client-side once a specific view (a university, a comparison, a Discoveries filter set) is on screen.
• public/og-image.png is the site-wide social preview image, regenerated with npm run og:image (uses Playwright's already-installed Chromium to screenshot scripts/og-card-template.html — no new dependency). A true per-page dynamic OG image isn't possible on a static GitHub Pages deploy without a server, so this is the "best static alternative": one well-designed card, with per-page title/description still set dynamically in the tags above.
• public/sitemap.xml lists the eight static pages plus one deep link per university straight into its Search-page research profile. Regenerate it after a meaningful CSRankings roster change with npm run sitemap (scripts/generate-sitemap.mjs).
• public/robots.txt allows all crawlers and points at the sitemap.

📊 Analytics

No analytics are wired to a real account by default — src/analytics.js's calls are safe no-ops until one is configured, and nothing here can invent credentials for you. To enable lightweight, cookie-free tracking:

1. Sign up at plausible.io (or self-host it) and register cspicks.roars.dev.
2. Uncomment the <script defer data-domain="cspicks.roars.dev" src="https://plausible.io/js/script.js"> tag near the bottom of each page's <head> (index.html, simulator.html, csconfs.html, nsf.html, grants.html).
3. Deploy. Plausible's dashboard then answers: visits, popular pages (via its own pathname-based pageviews), and referral sources out of the box.

src/analytics.js additionally fires custom events — View (by page and kind: school/researcher/area/search-results), Comparison, and Discovery Share, each tagged with page: 'search' | 'discoveries' where relevant — at the same points the URL updates, so "popular university pages," "popular research fields," "comparison usage," and "Discoveries traffic" are answerable from Plausible's custom-event breakdowns even though those views share one static HTML file per page. Swap the calls in analytics.js for another tool's API (e.g. GoatCounter) if preferred; nothing else needs to change.

🛠️ Technologies Used

• Frontend: Vanilla JavaScript (ES6+), HTML5, CSS3.
• Build Tool: Vite.
• Charts: Chart.js.
• CSV Parsing: PapaParse.
• HTML Encoding: he.

📦 Installation & Setup

1. Clone the Repository

   git clone https://github.com/dynaroars/cspicks.git
   cd cspicks

2. Install Dependencies

   npm ci

3. Run Development Server

   npm run dev

   The application will be available at http://localhost:5173/.

4. Run Tests and Build

   npm test
   npm run test:e2e
   npm run build

5. Deploy to GitHub Pages

   npm run deploy

🔁 Routine Maintenance

Two upstream sources move at very different speeds, so refreshing them is two different jobs.

How often	Command	Cost	What it does
Monthly, or after any CSRankings roster update	npm run sync:nsf:names	2 CSV downloads, seconds	Re-resolves NSF investigators to the name CSRankings' publication table uses, and rewrites public/nsf-name-crosswalk.csv
Quarterly, or when award data looks stale	npm run sync:nsf:all	Thousands of NSF API queries, hours	Re-queries the NSF Award Search API for every faculty/institution pair
After changing institution or name matching	npm run sync:nsf:rebuild	Local cache only, seconds	Rebuilds the dataset from .nsf-sync-cache.json with no API access

Run npm run sync:nsf:names regularly. CSRankings spells some faculty differently in csrankings.csv than in generated-author-info.csv, and the site matches on the latter. Without this refresh, faculty hired or renamed since the last award sync silently show no funding. It needs no NSF API access, so it is safe to run any time — commit the resulting public/nsf-awards.json and public/nsf-name-crosswalk.csv.

The crosswalk is meant to be read: each row records a name that needed resolving. Correcting a wrong row by hand is a legitimate fix.

npm run sync:nsf:names   # then review the diff in public/nsf-name-crosswalk.csv
npm test && npm run build

💰 Refreshing NSF Funding Data

The browser does not query NSF directly. NSF rejects browser-origin requests, and live per-user requests would make results dependent on API availability and unstable name matching. Instead, the funding page lazily loads a synchronized static dataset only when someone opens funding.html.

To synchronize all US institutions in the current CSRankings roster:

npm run sync:nsf:all

The synchronizer:

• queries the official NSF Award Search API for each unique faculty/institution pair;
• accepts awards only when the NSF recipient matches the current CSRankings institution;
• retains all listed PIs and co-PIs for fractional attribution;
• finds exact-title sibling awards for collaborative projects and deduplicates institution-transfer records;
• checkpoints progress in the ignored .nsf-sync-cache.json file;
• resumes incomplete runs without repeating completed queries.

To force a targeted faculty refresh while diagnosing a name variant:

npm run sync:nsf:all -- --faculty "Hoang-Dung Tran"

• writes the deployable dataset to public/nsf-awards.json, including explicit coverage totals.

NSF records awards under legal names (Regents of the University of Michigan - Flint), informal ones (Georgia Tech Research Corporation), and expansions of names CSRankings abbreviates (Massachusetts Institute of Technology vs Massachusetts Inst. of Technology). The synchronizer normalizes those forms, keeps an alias list for names it cannot derive, and assigns each awardee to the most specific matching institution so a flagship never claims its branch campus's awards.

To build a deliberately scoped dataset for one institution instead:

npm run sync:nsf -- --school "George Mason University"

Run the nationwide command again before deployment when you want to refresh NSF data. npm run deploy does not contact NSF automatically.

Funding interpretation

An award's estimated total amount (its intended amount) is divided equally among every listed PI and co-PI. University totals sum the shares assigned to matched current CSRankings faculty. These are matched-faculty statistics—not complete university NSF portfolios, annual expenditures, fiscal-year obligation totals, or measures of research quality.

Awards made to a professor's former institution are intentionally excluded. Name variants, missing co-PIs, transfers, supplements, and NSF data changes can still cause omissions.

🌱 Growing CSPicks

Adding a new Discovery

Discoveries must be reproducible from the same data every other page uses — never hardcoded or hand-picked. Adding one is usually two small pieces:

1. Compute it. Add a case to calculateDiscoveryInsights (university-level movement) or calculateSubfieldDiscoveries (region-wide, per research area) in src/metrics.js, or write a new pure function alongside them if the shape is genuinely different. Follow the existing pattern: compare a current period against the equal-length prior period via buildPriorPeriodData, apply a minimum-evidence guard (e.g. "both periods need an adjusted count of at least 2") so a tiny denominator can't dominate the list, and return a ranked, capped (limit, default 5) array.
2. Render it. In src/discoveries.js, call the shared card(title, help, body, className) helper with your new data. card() automatically slugifies the title into a stable id (discovery-<slug>), wires the ⓘ tooltip from help, and adds a Copy Link button scoped to that card — you get a shareable, self-documenting card for free. Use schoolLink()/areaLink() to link names back into Search.
3. Add a unit test. Every existing Discovery has a test/data.test.js case with a small synthetic fixture that asserts the ranking and thresholds; follow that pattern rather than relying only on live data. test/e2e/core-flows.spec.js covers the URL/share/hash-scroll mechanics generically, so a new card doesn't need its own e2e test unless it adds new interactive behavior.

Keep the help text honest about the methodology (thresholds, what counts as "prior period," what's excluded) — it's the only methodology note most readers will see, and it's what distinguishes a Discovery from an unsupported claim.

📂 Project Structure

cspicks/
├── public/
│   ├── professor_history_openalex.json  # Historical affiliations
│   ├── nsf-awards.json                  # Synchronized US NSF funding data
│   ├── nsf-name-crosswalk.csv           # Roster vs publication-table name resolutions
│   ├── school-aliases.json              # OpenAlex → CSRankings name mapping
│   ├── manual_affiliations.csv          # Community corrections
│   ├── og-image.png                     # Site-wide social preview image
│   ├── sitemap.xml                      # Static pages + one deep link per university
│   └── robots.txt
├── csconfs/                         # Conference-schedule page and synchronized data
│   ├── data/                        # Locally maintained conference schedules
│   ├── main.js                      # Search/filter/URL controller
│   ├── schedule-data.js             # Date, grouping, filtering, and sorting rules
│   └── schedule-render.js           # Native CS Picks schedule cards
├── src/
│   ├── data.js                       # Compatibility entry for loading, filtering, and ranking
│   ├── data/                         # Institution aliases and conference-set definitions
│   ├── metrics.js                    # Compatibility exports for domain metrics
│   ├── metrics/                      # School, researcher, discovery, comparison, and ranking metrics
│   ├── analysis/                     # Analysis renderers grouped by tab and responsibility
│   ├── simulator/                    # Candidate resolution and result rendering
│   ├── styles/                       # Foundation, component, and page-specific styles
│   ├── filters.js                    # Shared region/year/venue/history filter bar
│   ├── charts.js                     # Chart.js defaults, redraw, and theme handling
│   ├── main.js                       # Search page controller (also drives the Discoveries view)
│   ├── search-results.js             # Search result sections
│   ├── search-cards.js               # Professor/school card rendering
│   ├── suggestion-box.js             # Shared autocomplete menu (Search and Funding)
│   ├── search-suggestions.js         # Search autocomplete rows
│   ├── comparison.js                 # `A vs B` head-to-head mode (schools, professors, or areas)
│   ├── compare-view.js               # Comparison chart and summary rendering
│   ├── discoveries.js                # Discoveries view (rendered on the Search page): insight cards
│   ├── simulator.js                  # Simulator page UI and orchestration
│   ├── simulation.js                 # Pure matching and rank-impact logic
│   ├── analysis.js                   # Integrated analysis and data-health logic
│   ├── funding.js                    # NSF funding search page
│   ├── nsf.js                        # Funding attribution and rendering
│   ├── seo.js                        # Dynamic <title>/description/canonical/OG per view
│   ├── share.js                      # Reusable Copy Link / Web Share control
│   ├── analytics.js                  # Opt-in, no-op-by-default usage tracking hooks
│   ├── tooltip-position.js           # Positions ⓘ tooltip panels beside their trigger
│   └── style.css                     # CSS styles
├── scripts/
│   ├── build-openalex-history.js     # Generates historical affiliations
│   ├── sync-nsf-awards.mjs           # Resumable scoped/all-US NSF synchronization
│   ├── sync-nsf-roster-names.mjs     # Re-resolves NSF names against the CSRankings roster
│   ├── build-school-aliases.js       # Generates school-aliases.json
│   ├── generate-og-image.mjs         # Renders og-card-template.html to public/og-image.png
│   └── generate-sitemap.mjs          # Generates public/sitemap.xml from the CSRankings roster
├── index.html                        # Search, results, integrated analysis, and the Discoveries view
├── csconfs.html                      # CS conference schedule
├── csconfs-submit.html               # Conference submission/correction form
├── nsf.html                          # Nationwide NSF funding explorer
├── grants.html                       # CS research awards, fellowships & grants explorer
├── grants-submit.html                # Award/grant submission and edit form
├── simulator.html                    # Ranking simulator page
└── README.md                         # GitHub-hosted FAQ, methods, and data documentation

📊 Data Sources

• CSRankings - Faculty and publication data
• DBLP - Publication metadata and author profiles
• OpenAlex - Historical affiliation data
• NSF Award Search - NSF awards, investigators, program managers, programs, dates, and intended amounts
• CS Research Awards & Fellowships (public/grants.json) - Curated database of 85+ major CS research awards, faculty fellowships, student grants, and industry RFPs (NSF, DARPA, DOE, DoD, tech industry, and foundations)

❓ FAQ, Methods, and Data

How are rankings calculated?

CS Picks follows the CSRankings approach and uses the geometric mean of adjusted publication counts across research areas. This rewards breadth across computer science rather than dominance in only one area.

Score = (product of (adjusted count + 1)) ^ (1 / number of areas)

Each eligible paper contributes 1.0 in total, divided equally among its authors. For example, each author of a four-author paper receives an adjusted publication count of 0.25. CS Picks always uses this fractional-author credit.

What do the university statistics mean?

• Rank movement compares the selected period with the immediately preceding period of equal length.
• Momentum is the percentage change in adjusted publication count between those periods.
• Median per faculty is the median adjusted publication count among active faculty.
• Top-three concentration is the share produced by the university’s three highest-output faculty.
• Breadth counts active and sustained research areas.
• Team-size proxy divides raw publication count by adjusted publication count. It describes coauthor intensity, but cannot distinguish internal from cross-university collaboration.

What does History mode do?

By default, every eligible paper is credited to its author’s current CSRankings institution. History mode instead attempts to credit a paper to the institution where the author was affiliated in the publication year. Historical affiliation records are estimates and can be incomplete or incorrect, especially for older years, visiting positions, and renamed institutions.

What does the Rankings checkbox do?

Off by default, result lists read as plain lists of universities and people. Turning it on shows university overall and per-area ranks for the selected region, years, and conference set, and ranks people by adjusted publication count over the same selection. Equal values share a rank.

Can I compare two universities or professors?

Yes. On Search and NSF Funding, type both names separated by vs, such as Carnegie Mellon University vs Univ. of Illinois at Urbana-Champaign. Both search boxes autocomplete, and the second target must be the same kind of target as the first. Search compares publication output by research area; NSF Funding compares awards, attributed funding, and matched CS faculty.

What do the conference-set options mean?

• CSRankings (Default) uses the primary CSRankings venue set and excludes optional next-tier venues.
• CSRankings (All) includes both the primary and extended CSRankings venues.
• CORE A* includes only conferences mapped to the CORE A* tier.
• CORE A*/A includes conferences mapped to either the CORE A* or A tier.

Conference definitions can change upstream. CS Picks synchronizes its venue rules from CSRankings when possible and keeps a bundled fallback.

How does the ranking simulator work?

Choose a target university and one or more researchers. Current faculty are modeled as removals, faculty at another ranked university as transfers, and external DBLP researchers as additions. The simulator applies their eligible publication records to the selected period and conference set, then recalculates overall and per-area ranks. Results are hypothetical and should not be interpreted as predictions, hiring recommendations, or evaluations of individuals.

Where does the data come from?

• CSRankings supplies the faculty roster, publication data, venue taxonomy, and institution information.
• The locally maintained schedule under csconfs/data/ supplies conference dates and submission timelines. Estimated entries should be confirmed on the linked conference website. Deadline countdowns use Anywhere on Earth (UTC−12), and schedule year filters refer to the conference year.
• Schedule records are maintained through research of official conference and sponsoring-society websites plus contributor corrections. Historical acceptance totals came from emeryberger/csconferences.
• DBLP supplies author-search and publication metadata used by the simulator.
• OpenAlex and manually reviewed corrections supply estimated historical affiliations.
• NSF Award Search supplies the public award records used by the NSF funding beta.

How does the NSF Funding beta work?

The synchronizer searches current CSRankings faculty as primary investigators and retains an award only when its NSF recipient matches the faculty member’s current CSRankings institution. For awards with multiple investigators, NSF’s estimated total award amount is divided equally among all PIs and co-PIs; university totals sum the shares assigned to matched current faculty. This is a matched-faculty statistic, not a university’s complete NSF portfolio.

Institution names are normalized and matched to the most specific institution so branch campuses are not absorbed by a flagship. The synchronizer records roster and publication-table name variants in public/nsf-name-crosswalk.csv. It also distinguishes confirmed transfers and enriches collaborative projects with exact-title sibling awards while keeping each university’s local attribution unchanged.

What are the data limitations?

• Historical affiliations are assembled from automated sources and manual corrections; coverage is uneven.
• Current-roster mode assigns past work to current institutions and is not a historical department ranking.
• Publication, author, conference, and eligibility records can change when upstream sources update.
• The collaboration statistic is a coauthor-intensity proxy, not a measurement of cross-institution collaboration.
• NSF investigator matching can miss name variants and deliberately excludes awards made to another institution, even when they may belong to the same faculty member’s earlier career.
• NSF dollar totals use estimated total award amounts grouped by award year; they are not annual expenditures or fiscal-year obligations.
• A passing Data Health audit means the calculation is internally consistent with loaded inputs, not that every deployed upstream page is identical at that moment.

Privacy, corrections, and acknowledgments

CS Picks fetches public scholarly metadata on demand and does not store names entered into Search or the simulator as user-submitted personal data. To report an affiliation correction or other discrepancy, open a GitHub issue with the name, correction, supporting source, and applicable years. CS Picks is inspired by CSRankings; historical affiliation estimates use OpenAlex plus manually reviewed corrections.

📝 License

Original CS Picks source code and project material are available under the MIT License. The repository also contains third-party data, metadata, dependencies, and assets that are not relicensed by MIT. See the source-specific notes below before redistributing those materials.

Data and asset licensing

The MIT license applies to original CS Picks source code and other original project material to the extent that Roars Lab owns the rights to it. It does not relicense material obtained from external sources.

Material	Source	Reuse notes
Faculty, publication, institution, and venue data	CSRankings	Follow the current CSRankings repository license and attribution requirements.
Publication and author metadata used by the simulator	DBLP	DBLP’s dataset is released under CC0, with attribution/source-link guidance. Follow DBLP’s current terms and conditions.
Historical affiliation data and institution mappings	OpenAlex	OpenAlex describes its data as CC0/public domain. This applies to OpenAlex metadata, not to the underlying scholarly works or full text, which retain their own rights.
Award records and investigator metadata	NSF Award Search	These are public government records, but the snapshot and matching process are project-specific. Verify current NSF terms and source attribution before redistribution.
Conference schedule records	csconfs/data/conferences.json	Curated project data assembled from the official conference and sponsoring-organization pages linked in each record. Verify those source terms before reuse.
Fonts and JavaScript dependencies	Their respective upstream projects	Each dependency keeps its own license; see package-lock.json and the upstream project for details.

The files under public/ are deployable snapshots or project-generated transformations. The MIT license should not be interpreted as permission to reuse third-party contents without complying with the applicable source terms. When in doubt, use the source links above and retain the source attribution included in the files and application documentation. Source terms can change, so downstream redistributors should verify them at the time of reuse.