How to Write a Thesis
Introduction
- This guide is in English as it was originally thought for a course is taught in English.
- The principles we cover, however, apply equally well to theses written in Italian or in any other language.
- The material in this guide reflects:
- First-hand experience as a thesis writer
- First-hand experience as a thesis co-supervisor
- First-hand experience as a reviewer for international journals and conferences
Disclaimer
Every thesis has its own context, and these guidelines should be adapted depending on:
- The candidate (background, skills, project trajectory, writing style, etc.)
- The supervisor (expectations, working style, etc.)
- The type of thesis (experimental, literature-based, etc.)
The Most Important Rule
Make your thesis clean, correct, and polished.
- Minimize typos, grammar mistakes, and formatting inconsistencies.
- A well-written thesis saves your supervisor and reviewers from spending time on corrections that should not be necessary.
- Clear, correct writing signals:
- Professionalism
- Respect for the reader’s time
- Attention to detail
- Care for your own work
- A thesis that is technically sound and well written leaves a strong positive impression from the very beginning.
Thesis Types
At the University of Udine, I commonly encounter three main types of theses: Experimental, Internship Report, and Literature Review. In the following, we detail related aspects.
Experimental Thesis
- IT: Tesi Sperimentale
- Develops or tests new ideas, methods, or research questions.
- Often involves designing experiments, running analyses, and discussing results.
- In stronger cases, it may lead to publishable research.
Internship Report
- IT: Tesi di Tirocinio
- Based on a collaboration with a company, lab, or institution.
- Focuses on describing the project context, the work carried out, and the technical or methodological contribution.
- The emphasis is typically on applied problem-solving, not on generating new scientific knowledge.
Literature Review
- IT: Tesi compilativa
- Synthesizes and critically analyses existing scientific literature on a given topic.
- Requires building a coherent narrative, identifying gaps, and showing understanding and comparison of prior work.
- The contribution is conceptual rather than experimental.
Thesis Structure
We now detail how a thesis should be structured. First, by considering its general structure and then considering each of its subcomponents.
General structure
A general structure of a thesis is as follows:
Abstract
Table of Contents
List of Figures (*)
List of Tables (*)
Chapter 1 - Introduction
...
Chapter n - Conclusions
Bibliography
Appendix A (*)
Some notes on this general structure:
- Items marked with (*) are optional depending on formatting rules and the number of figures/tables.
- You might also want to add additional lists (e.g., List of Algorithms) if the number of algorithms reported in pseudo-code is high.
All in all, atypical structure for a thesis may look like the following
Abstract
Table of Contents
List of Figures (*)
List of Tables (*)
Chapter 1 — Introduction
Chapter 2 — Related Work
Chapter 3 — Methodology
Chapter 4 — Experimental Analysis
Chapter 5 — Conclusions
Bibliography
Appendix A — Extended Results
Some notes on this complete structure:
- Items marked with (*) are optional depending on formatting rules and the number of figures/tables.
- The core chapters (2–5) follow a logical flow: define the problem → position it in the literature → present your approach → evaluate it
- Appendices are ideal for material that is important for completeness but unsuitable for the main narrative (e.g., full result tables, parameter files, large figures). Still, they are not always necessary.
Abstract
- A short summary of the entire thesis (typically 150–250 words).
- Should be self-contained: a reader must understand the topic, approach, and key findings without reading further.
- Covers four essential elements:
- Context: What is the thesis about? Why does it matter?
- Objective: What research question or goal does the thesis address?
- Approach: What methods, models, or experiments were used?
- Results & contribution: What did you find? What is the key takeaway?
- Avoid details (e.g., parameters, long lists, citations).
Write it last, but polish it carefully: it is often the first and sometimes the only part people read.
Introduction
- Present the research problem and explain why it matters.
- Motivate the work so to give the reader enough context to follow the rest of the thesis.
- Define the objective and, when applicable, the research questions.
- Provide a brief roadmap of the thesis: what each chapter contains and how the story unfolds
- If the thesis was carried out in collaboration with a company (or a research center outside the University of Udine), include a concise description of the partner organization.
- Keep it factual, technical, and relevant to the thesis.
- Avoid promotional or marketing-style language.
A typical structure for the introduction is as follows
Chapter 1 - Introduction
# Opening paragraphs introducing the problem:
# context, real-world relevance, scientific motivation.
1.1 Objectives and Research Questions
- What the thesis aims to achieve.
- Specific, clearly stated research questions (if relevant).
1.2 Contributions of the Thesis
- What is new or valuable in the work.
1.3 Thesis Structure
- Short overview of each chapter.
Methodology
As for the Related Work, also in the case of the Methodology chapter, the content and structure change based on the type of thesis one is writing.
Experimental / Internship Report
This chapter describes how you approached the problem and why you made specific design choices. Include:
- The overall method or algorithmic framework.
- Assumptions and simplifications.
- Implementation details that matter (architecture, libraries, parameter choices).
- Experimental design: datasets, instances, metrics, reproducibility considerations.
- Justification of the chosen approach (not just a description).
Keep the focus on clarity: another person should be able to reproduce your work based on this chapter.
Literature Review
This chapter explains how you conducted the review, i.e., your search strategy and selection process. Include:
- Databases consulted (e.g., Scopus, ACM DL, IEEE Xplore, Web of Science).
- Keywords, search strings, inclusion/exclusion criteria.
- Time window and scope of the review.
- Screening process, deduplication and selection steps (PRISMA or other frameworks).
- How articles were categorized, compared, and synthesized.
This gives your review rigour and transparency, showing that your conclusions stem from a systematic process.
Experimental Analysis
- Present the findings clearly using tables, graphs, and figures.
- Organize results into sections or subsections when multiple experiments or related topics are addressed.
- For each major experiment, provide a dedicated section with:
- a brief description (recall why/what)
- the results
- a discussion explaining what the numbers mean
- Compare your findings with the existing literature, highlighting similarities, differences, and improvements.
- Avoid overcrowded tables (if necessary, put them in the Appendix)
Ideally, each section should tackle one of your research questions
Conclusions
- Provide a concise summary of the main contributions and key findings of the thesis.
- Reinforce the significance of the work, i.e.m why the results matter and what they add.
- Outline possible applications, extensions, or follow-up studies that could build on your work.
- Keep the tone reflective but focused: the conclusion should close the narrative, not introduce new material.
A typical structure for the Conclusions is as follows
Chapter 5 - Conclusions
# Opening paragraphs
5.1 Key Findings
- Summary of the main results.
- What the thesis has shown or demonstrated.
5.2 Future Work
- Possible extensions and improvements.
- Ideas for further research.
Bibliography
- Choose one citation style (e.g., APA, IEEE, ACM) and use it consistently throughout the thesis.
- For websites, always include the date of access, since online content can change over time.
- Use reliable and authoritative sources:
- Prefer peer-reviewed articles, books, and established conference proceedings.
- Be cautious with preprints (e.g., arXiv): they are useful, but not peer-reviewed—cite them knowingly.
- Avoid citing predatory journals or conferences.
- Check reputation (e.g., Elsevier, Springer, IEEE, ACM are generally safe).
- Be cautious with outlets from unknown publishers or those listed on Beall’s List or similar warning lists.
Managing Bibliography Efficiently
- Use a reference manager
- Zotero and Mendeley are perfectly fine.
- If you prefer a lightweight approach, an Excel sheet plus a well-maintained .bib file also works.
- Prefer BibTeX/BibLaTeX when working in LaTeX (more on this later)
- Keep your .bib entries clean, consistent, and free of duplicates.
- Always check metadata accuracy:
- Titles, authors, venues, DOIs, page numbers.
- Reference managers often import incomplete or incorrect data; fix errors early.
- Keep track of preprints vs. published versions and cite the official publication whenever possible.
Generalities on Chapter, Sections, and Subsections
- Organize your text hierarchically: chapters → sections → subsections.
- Use subsections only when they improve clarity and structure.
- If you introduce Section 1.1, it must be followed by Section 1.2 (at least).
- Avoid creating single, isolated subsections (e.g., “1.1” without “1.2”).
- The same rule applies to subsections (e.g., don’t create “1.1.1” if there is no “1.1.2”).
- Each chapter should open with a brief introduction and should end with a short summary
- The Introduction and the Conclusions Chapters are exceptions.
- Maintain coherence throughout:
- Logical flow from chapter to chapter.
- Consistent terminology and structure.
- Ensure consistent titling style: stick to one convention, e.g. use always Title Like This (Capitalized Words) or Title like this (Sentence case)
Figures and Tables
Figures
- Label your axes
- Include axis names and units when relevant (e.g., Time (s), Objective Value).
- Avoid cryptic abbreviations unless they are defined in the text.
- Use a limited, coherent color palette
- Avoid using too many colors or overly saturated tones.
- Choose colors that remain readable in grayscale when printed.
- Ensure readability
- Text, tick labels, and legends must be large enough to read comfortably.
- Avoid overcrowded plots; simplify when possible.
- Use consistent styling across all figures
- Same fonts, same color scheme, same line thickness.
Tables
- Use a reasonable number of significant digits and keep precision consistent across columns.
- Ensure readability
- Font size should not be too small.
- Keep tables compact and avoid unnecessary borders or clutter.
- Align numbers properly: numerical columns should be right-aligned or decimal-aligned for easier comparison.
- Text columns can be left-aligned.
- Prefer clear column labels with units where needed (e.g., Time (s)).
- Highlight key values logically (best, median), but avoid distracting colour or formatting overload.
- Keep large or auxiliary tables in the Appendix
Referencing
- Every table and figure must be referenced in the text.
- Avoid leaving standalone elements without explanation or context.
- Connect your narrative to relevant sections and subsections when needed, but avoid excessive cross-referencing that interrupts the flow.
- When writing in LaTeX, use the cleveref package to manage references consistently.
- It automatically formats references (e.g., Figure 3, Table 2, Section 4.1).
- Use the correct terminology
- E.g., it is Figure, not “image,” “plot,” or “graph.”
- Keep capitalization consistent throughout your thesis (e.g., Figure 1, Table 3).
Captions
- A caption should explain what the figure/table shows without requiring the reader to guess.
- Begin with a clear, informative title.
- Example: Figure 3 — Solution quality over time for Algorithm A.
- Follow with a concise explanation: What data is shown? What the reader should notice?
- Avoid repeating the exact sentence used in the main text: captions should complement, not duplicate.
- If symbols or abbreviations appear in the figure, define them in the caption unless they are introduced elsewhere.
- Keep formatting consistent across all captions: Similar structure, similar level of detail.
- Do not include citations in captions unless strictly necessary (e.g., when reproducing a figure from another source).
LaTeX or Word?
Both are acceptable, but each has its strengths. Choose based on your project, your comfort level, and your thesis requirements.
LaTeX
- Recommended for technical theses
- Ideal for mathematics, algorithms, and cross-referencing
- Produces consistent and professional typesetting
- Better for large documents with many chapters
- Steeper learning curve, but worth it for scientific theses (especially in CS/AI/CO)
LaTeX Templates
- This guide comes with my own latex template, but feel free to pick the one you like the most (the thesis is yours, afterall). A few other suggestions (for the University of Udine)
- I cannot help with Word (the last official document I wrote in Word was back in 2019)
Word
- Easier to start with, familiar to most students
- Suitable for shorter or less technical theses
- Formatting may become fragile in longer documents
- Handling equations, references, and large bibliographies can become cumbersome
- Version control and consistency are harder to maintain
Other Useful Suggestions
Writing Style
- Avoid large code blocks
- Include only short, illustrative snippets (only if really necessary).
- Put full code in a repository (GitHub, Zenodo, OSF).
- Write clearly and concisely
- Use short, well-structured paragraphs.
- Prioritize clarity over complexity.
- Avoid long, convoluted sentences.
- Prefere active voice
- Use precise terminology
- Replace vague claims with measurable statements.
- Provide evidence for all assertions.
- Avoid redundancy
- Do not repeat the same idea in different words.
- Maintain cohesion and flow
- Use transitions: “Building on these results…”, “As shown in the previous section…”
- Define technical terms and acronyms
- Define them upon first use and use them consistently.
- Keep capitalization consistent (BERT, not Bert sometimes).
- Avoid pompous or overly complex language
- Aim for professional, straightforward writing.
- Justify claims
- Never write “our method is very effective” without evidence.
- Avoid vague expressions
- Replace some, many, kind of with exact values or explanations.
- Read other theses
- Look at well-written theses from your department to understand structure, tone, and expectations.
- Plan your writing early
- Don’t wait until all experiments are done.
- Start drafting chapters such as Introduction and Related Work while your work is still ongoing.
- Write regularly
- Short, consistent writing sessions are more effective than last-minute marathons.
- Back up your work often
- Use GitHub, cloud storage, or version control to avoid losing drafts or results.
Useful Material
Declarations
- Some parts of this guide were reviewed and improved with the assistance of large language models (LLMs). All content has been verified, curated, and approved by the author before inclusion.