I11 · Intermediate Capstone: An End-to-End AI Project

Writing a Project Abstract

A project abstract is a short, self-contained summary, usually 150 to 250 words, that states the problem and why it matters, the approach and data used, the main results with numbers, and the significance or next steps. It is written in plain, precise language so that a reader who sees nothing else understands what was done and found. For AI projects it should mention the data source, the model type, the evaluation metric and how the result compares with a baseline.

10 min Intermediate · Sublevel Advanced Jump to the activity

What you need: The earlier topics of this module.

Definition

Judges at a science fair may spend 30 seconds deciding whether to stop at your stall. Reviewers of a research paper read the abstract before anything else. A 200-word abstract can decide whether anyone reads the rest of your work.

A project abstract is a short, self-contained summary, usually 150 to 250 words, that states the problem and why it matters, the approach and data used, the main results with numbers, and the significance or next steps. It is written in plain, precise language so that a reader who sees nothing else understands what was done and found. For AI projects it should mention the data source, the model type, the evaluation metric and how the result compares with a baseline.

In simple words

An abstract is like the trailer of a film: in a minute it tells you what the story is about, why it is interesting and how it ends, so you can decide whether to watch the whole thing.

Overview

  • A project abstract is a short summary, usually 150–250 words, of what you did and found.
  • Readers use it to decide whether to read the full report.
  • A standard structure is problem, method, data, results and conclusion.
  • Good abstracts include specific numbers, such as accuracy or error reduced.
  • They avoid jargon, abbreviations without definitions and claims the results do not support.
  • Write the abstract last, after the results are final, and revise it several times.

Why it matters

Abstracts are required for competitions, conferences, journals, internship applications and project reports. Writing one also tests your own understanding: if you cannot summarise the project in 200 words, its goal or result is probably not clear yet.

Key terms

Abstract
A short, complete summary of a project's problem, method, results and significance.
Background
One or two sentences on why the problem matters.
Method
What data and approach were used.
Quantitative result
A specific number, such as "mean absolute error of 14 µg/m³, 30% lower than the baseline".
Significance
Why the result matters and what could happen next.
Keywords
Four to six terms that help others find the work.

Types

An abstract has several types of part, and abstracts themselves come in a few types.

1. Problem and motivation

One or two sentences on the problem and why it matters.

  • How it works: State who is affected and the cost of the problem, ideally with a number.
  • Example: "Small shops lose an estimated 8% of sales each month because popular items run out."
  • When to use it: Use it to make the reader care in the first lines.

2. Method and data

What you built and what data you used.

  • How it works: Name the model type, the main features and the dataset size and source in plain words.
  • Example: "We trained a random-forest regressor on 18 months of sales from 40 shops."
  • When to use it: Use it to show the work is credible and repeatable.

3. Results

The main findings with numbers and a comparison with a baseline.

  • How it works: Report the key metric, the baseline and the improvement; mention any important limitation.
  • Example: "The model cut forecast error by 35% compared with a four-week average."
  • When to use it: Use it to give readers the most important facts even if they read nothing else.

4. Conclusion and impact

What the results mean and what could happen next.

  • How it works: Connect results back to the users, state the main takeaway and suggest one next step.
  • Example: "A simple dashboard could help shopkeepers restock more accurately; we next plan a pilot in 10 shops."
  • When to use it: Use it to end with purpose rather than a list of numbers.

5. Types of abstract: informative, descriptive and graphical

Different abstract styles for different audiences.

  • How it works: An informative abstract gives results and numbers; a descriptive one only says what the report covers; a graphical abstract summarises the work in one picture.
  • Example: A science-fair poster using a graphical abstract with a pipeline diagram and the key result.
  • When to use it: Use informative abstracts for project reports and competitions; add a graphical one for posters.

Explained in detail

The structure of a strong abstract

PartPurposeApproximate length
Background and problemWhy the topic matters and what gap exists1 to 2 sentences
ObjectiveWhat the project set out to do1 sentence
Data and methodData source, size, model and evaluation approach2 to 3 sentences
ResultsMain numbers, compared with a baseline1 to 2 sentences
Significance and limitsWhat the result means, key limitation and next step1 to 2 sentences

An example abstract for our project

"Fine particulate pollution (PM2.5) frequently reaches unhealthy levels in Indian cities during winter, yet most residents learn of bad air only on the day itself. This project aimed to forecast the next day's 24-hour average PM2.5 for one city so that schools could receive alerts a day in advance. In this illustrative abstract, using three years of simulated daily station readings and weather data, we compared a persistence baseline (tomorrow equals today) with linear regression and random-forest models, evaluating on the most recent year held out by date. A linear model using tomorrow's wind forecast reduced the mean absolute error by about 25 per cent compared with the baseline, and with an alert threshold of 80 µg/m³ it detected 99 per cent of days classed as Poor or worse; the random forest did not improve on it. Accuracy was lowest during sudden episodes such as festival fireworks. Whether schools would act on a message by 6 p.m. remains a question for the pilot. Future work will add more monitoring stations and test the alerts in a pilot."

Tips for clear writing

  • Write the abstract last, after the results are known, but draft a version early to clarify your goal.
  • Use specific numbers and comparisons; "good accuracy" tells the reader nothing.
  • Avoid jargon and undefined abbreviations; spell out terms such as PM2.5 the first time.
  • Be honest about limitations; reviewers trust work that admits its weaknesses.
  • Stay within the word limit, and ask someone outside the project to read it.

Different audiences need slightly different abstracts. A science-fair abstract can open with a vivid fact about the problem; a conference abstract follows the field's conventions closely; an internship application may add one sentence on what you personally built. Keep a master version and adapt it, always keeping the numbers identical across versions.

How it works

  1. State the problem and why it matters in one or two sentences.
  2. Give the objective in one sentence.
  3. Describe data, model and evaluation briefly.
  4. Report the main results with numbers and a baseline comparison.
  5. End with significance, a key limitation and the next step; check the word count.

Try it yourself

Match the sentence to the abstract part

Which part of the abstract is each sentence?

  • "PM2.5 often reaches unhealthy levels in winter."
  • "We aimed to forecast next-day PM2.5 to alert schools."
  • "We compared a persistence baseline with a random forest."
  • "The error fell by 25 per cent compared with persistence."
  • "Future work will test alerts in a pilot."

Good abstract sentence or not?

Decide for each sentence.

  • "Accuracy was lowest during festival fireworks."

  • "Our model got really good results."

  • "Mean absolute error was 8.3 µg/m³, versus 11.7 for the baseline."

  • "We used AI and ML with DL and NLP techniques."

Common mistakes

  • Describing the method in detail but giving no results.
  • Using vague words such as "good" or "improved" without numbers.
  • Exceeding the word limit or leaving abbreviations unexplained.

Check yourself

Question 1

1. What is a typical length for a project abstract?

Question 2

2. Which sentence best reports a result?

Question 3

3. Why mention limitations in an abstract?

Summary

  • An abstract summarises problem, objective, method, results and significance in about 150 to 250 words.
  • Use specific numbers, compare with a baseline and explain terms plainly.
  • State limitations and next steps honestly, and keep within the word limit.

Finished reading?

Sources: University of Wisconsin Writing Center: writing an abstract · Central Pollution Control Board: National Air Quality Index