Practical guides

Practical, no-nonsense guides for students — written to be handed out, printed, and worked through rather than read once.

Everything is LaTeX; the compiled PDFs are committed, so you can download a PDF directly or edit the source and rebuild.

---

Starting a PhD in Machine Learning or NLP

A starter kit for new PhD students, in three formats that share the same content at three levels of detail.

	File	Format	Use it for
📘	phd/phd-ml-nlp-starter-guide.pdf	20 pages, A4	The full guide — read it through in week one
📄	phd/phd-ml-nlp-onepager.pdf	1 page, 3 columns	Print it and pin it above the desk
🖥️	phd/phd-ml-nlp-slides.pdf	46 slides, 16:9	A lab induction talk or a first-year seminar

What they cover

• The publication landscape — ML venues (NeurIPS, ICML, ICLR, AISTATS, AAAI, IJCAI, COLM …), NLP venues (ACL, EMNLP, NAACL/EACL/AACL, COLING/LREC, TACL, Findings …), adjacent venues and journals, how the ACL Rolling Review submit-then-commit cycle actually works, and how to judge a venue (Scholar h5, CORE, CSRankings) — plus how to spot a predatory one.
• Deadlines — the two countdown boards worth bookmarking, the AoE trap, and a calendar workflow that backward-plans from camera-ready.
• Writing and sharing — Overleaf (now with its own metered AI assistant), a local git + latexmk workflow, one-sentence-per-line diffs, \citet vs \citep, latexdiff, arxiv-latex-cleaner, anonymity periods, releasing code and checkpoints, and making the code citable with CITATION.cff and a Zenodo DOI.
• Zotero — the connector, Better BibTeX with pinned keys and auto-export into your paper repo, group libraries, and the habits that keep a library trustworthy. Includes the quickest way to get an entry into your .bib (Google Scholar's BibTeX links), what to check in it, and cleaner sources — the ACL Anthology, DBLP, or any DOI straight from the terminal. Also the three BibTeX gotchas (protected capitals, author format, arXiv fields), bibtex-tidy, Zotero's retraction alerts, and a pre-submission audit that checks every DOI resolves.
• A focused survey as the first project — a 10-step protocol, inclusion criteria, forward and backward snowballing, ASReview for screening large piles, an extraction schema, and finishing with three gaps stated as testable questions.
• Searching the literature — Google Scholar as an instrument (operators, search within citing articles, alerts, profile), plus Semantic Scholar, OpenAlex, the ACL Anthology, OpenReview, DBLP, Connected Papers, Scite and Unpaywall, and Scholar Labs for question-level search.
• Reading the signals — citation velocity over volume, and what GitHub stars, commit history, issues and licences really tell you about a baseline.
• Setting up an AI assistant — project memory, MCP connectors, saved prompts, Google's Deep Review as a free first review of the final PDF, and five non-negotiable guardrails (starting with: never cite a paper you have not opened).
• Your personal site — GitHub Pages plus the AcademicPages template, step by step, with https://y3nk0.github.io as a worked example.
• Infrastructure and habits — reproducible experiments, seeds and variance, experiment tracking, the weekly advisor meeting, and checklists for week 1, month 1, months 2–3, year 1, and before every submission.

Key links from the guides

• AI conference deadlines: https://paperswithcode.co/ai-deadlines and http://aideadlines.org/?sub=ML,CV,NLP
• Zotero https://www.zotero.org · Better BibTeX https://retorque.re/zotero-better-bibtex/ · bibtex-tidy https://github.com/FlamingTempura/bibtex-tidy
• Overleaf https://www.overleaf.com · Prism (closes 1 Jan 2027) https://prism.openai.com
• ACL Anthology https://aclanthology.org · OpenReview https://openreview.net · DBLP https://dblp.org
• AcademicPages https://github.com/academicpages/academicpages.github.io

---

Doing a Master's Thesis in Machine Learning or NLP

For the project between the two: a contribution, but a bounded one.

	File	Format	Use it for
🧭	masters-thesis/masters-thesis-ml-nlp-guide.pdf	12 pages, A4	The full guide — agree it at the first meeting
📄	masters-thesis/masters-thesis-onepager.pdf	1 page, 3 columns	Print it and pin it above the desk
🖥️	masters-thesis/masters-thesis-slides.pdf	22 slides, 16:9	A kickoff talk for a master's cohort

What it covers

• The bar, against the other two degrees — what changes from an undergraduate project (a contribution, a positioned related-work chapter, significance where claimed) and what a PhD asks that you do not owe.
• The one-claim test — the sentence your thesis exists to defend, and what an empty slot in it tells you to work on next.
• Five routes to a contribution — extend one paper, reproduce and interrogate, a shared task with an angle, an applied or industry thesis, or analysis and theory — plus the six places open questions are written down.
• The two-page proposal, and the scope test that catches the way theses actually fail.
• Industry and NDA theses — the six questions to settle in writing before you start, and how to write around confidential data.
• Evidence that survives an examiner — baselines first, seeds and spread, paired bootstrap or permutation tests, reporting compute, and the mistakes that cost a grade.
• Supervision — what to decide yourself, what to bring to the meeting, what to escalate the same day.
• Writing, the defence, and turning the thesis into a paper — what to cut, where to send it, authorship, and timing against the thesis deadline.

---

Doing an Undergraduate Thesis in Machine Learning or NLP

A guide for final-year students, written to be handed out at the first supervision meeting.

	File	Format	Use it for
🎓	undergrad-thesis/undergrad-thesis-ml-nlp-guide.pdf	15 pages, A4	The full guide — hand it out at the first supervision meeting
📄	undergrad-thesis/undergrad-thesis-onepager.pdf	1 page, 3 columns	Print it and pin it above the desk
🖥️	undergrad-thesis/undergrad-thesis-slides.pdf	31 slides, 16:9	A thesis-kickoff talk for a whole cohort

What it covers

• What a thesis is and is not — what examiners actually look for, and a three-question test for whether your scope is realistic.
• Finding a topic — check the supervisor's advertised topics first and do a small piece of research before the first meeting; if you are free to choose, three routes: a shared task or challenge (SemEval, workshop tasks at *ACL venues, WMT, CoNLL, BabyLM, CLEF labs, TREC tracks, Kaggle and the competition platforms), an application thesis (a working method taken carefully to a new language, domain or deployment constraint), or an analysis / theoretical thesis (reproduction, ablation, robustness, probing, efficiency, or a critique of the metric itself).
• Six places where gaps are written down explicitly — limitations sections, future work, survey taxonomies, workshop calls for papers, task overview papers, and leaderboards that have plateaued.
• The one-page proposal to agree with your supervisor before writing code.
• The toolkit — Zotero, LaTeX and Overleaf, git, free GPUs, experiment tracking, data licensing and GDPR — plus the quickest way to get a reference into your .bib, how to check it, bibtex-tidy, Unpaywall, the right citation commands, and Scholar Labs for question-level search.
• Using an AI assistant honestly — five rules, starting with the one that decides the defence: you must be able to explain every line you submit, plus Google's Deep Review as a free first review of the draft.
• Doing the work — baselines first, the mistakes that cost marks (test-set leakage, one seed, unfair comparisons), and why negative results are results.
• Writing and defending — a chapter-by-chapter structure with sizes, writing advice, likely defence questions, a six-month timetable, and checklists for week 1, before freezing results, and before submission.

---

Layout

phd/                 the PhD starter kit: full guide, one-pager, slides
undergrad-thesis/    the thesis guide: full guide, one-pager, slides

Each directory holds the .tex sources and the built PDFs side by side.

Building

Needs a TeX distribution (TeX Live, MacTeX or TinyTeX) with latexmk. The deck uses beamer and tikz; everything else is standard.

make            # build any PDF whose source changed
make check      # build, then verify the one-pagers still fit on one page
make purge      # after a push: ask the CDN to refetch the PDFs right away
make clean      # remove build artefacts (the PDFs are kept)

make check fails if a one-pager has grown to two pages, and warns about any line that runs past the margin. The website at https://y3nk0.github.io/guides/ serves these PDFs from this repository through jsDelivr, which refreshes a branch within about 12 hours; after pushing a rebuilt PDF, run make purge to make it pick up the new build at once.

Builds are reproducible — the PDF timestamp is pinned (SOURCE_DATE_EPOCH in the Makefile), so the same sources give byte-identical PDFs on any machine and after any fresh clone, and git sees no change after a rebuild.

To build a single document by hand: latexmk -cd -pdf phd/<file>.tex.

Note on the one-pagers: both are tuned to fill exactly one A4 page. If you add material, drop the base font size at the top of the document body (\fontsize{6.35}{7.45} for the PhD sheet, \fontsize{6.05}{7.15} for the thesis sheet) by a tenth of a point or two and recompile.

---

Author

Konstantinos Skianis — Assistant Professor of Natural Language Processing, Department of Computer Science and Engineering, University of Ioannina · https://y3nk0.github.io

Corrections and additions are welcome — venues, deadlines and tools move fast, so open an issue or a pull request if something here has gone stale.

A note on how these were written

The content, structure and editorial decisions in these guides are mine. I wrote them with the help of an AI assistant (Claude), working under my specific directions — drafting from my outlines, typesetting the LaTeX, and checking that links and references resolve. I reviewed and edited everything that appears here, and any errors are my own.