This chapter provides guidance on selecting a research topic and task, as well as how to compose the Introduction section to effectively motivate the readers and offer an overview of your work.
Learn how to structure the Introduction section.
Find your team members.
Select 3-5 papers related to your task.
For each paper, excerpt writings in the introduction section for the following categories:
Broad Impacts
Intellectual Merit
Approach
Findings
Contributions
Identify parts of the writings that do not fit into any of the categories above and propose appropriate categories for them.
Make slides summarizing the above analysis for a short presentation (one per group) and submit them to [Assignments → Exercises → 2.4. Exercise] in Canvas.
Homework 2: Introduction
Write the Introduction section in your individual overleaf project.
Recommended length: 50 - 100 lines.
Submit the PDF version of your draft, including the Introduction section.
Broad Impacts (1 point): Is the significance of the task well-justified?
Intellectual Merit (1.5 points): Are the challenges of the task effectively outlined?
Approach (1 point): Is the main approach coherently stated?
Findings (0.5 points): Are the (expected) findings clearly indicated?
Contributions (0.5 points): Are the contributions of the work concisely summarized?
References (0.5 points): Are the relevant works and sections appropriately referenced?
The scores for findings and contributions are lower because they cannot be completely figured out at this time of the writing, not because they are less critical.
Select a specific task and determine the research objectives of your project.
Examine the Areas and Topics of Interest below, and form a team comprising up to 3 members using [People → 2.1. Task Selection*] in Canvas.
Choose a particular task for your team and discuss the types of data you wish to concentrate on (e.g., open relation extraction on discussion forums).
Outline the primary research objectives (which may be more than one) of your project:
Devise an innovative methodology that outperforms previous approaches on specific or relevant datasets.
Construct a novel dataset that facilitates task adaptation to new domains or languages.
Conduct a comprehensive analysis of state-of-the-art models, focusing on comparative aspects to gain insights into their strengths and weaknesses.
Develop a practical application employing existing resources to advance translational research.
Make slides summarizing the above outline for a short presentation (one per group) and submit them to [Assignments → Exercises → 2.1. Task Section] in Canvas.
Today's team does not need to be the one for your final project; however, use this opportunity to explore the interests of your colleagues.
The novelty of your research will be assessed based on these objectives. Ensure that your objectives are designed to address the challenges outlined in the motivation.
Describe the motivation behind your selected task.
If the selected task has been popular in recent years, highlight its prevalence and reference recent seminal works related to the study:
TASK has recently gained lots of interests because REASONS (CITATIONS).
You may want to make separate citations per reason if appropriate e.g., REASON 1 (CITATIONS), REASON 2 (CITATIONS)
If the selected task is new or has not been widely recognized in recent times, explain its significance and cite relevant works that provide evidence of its importance:
TASK is important because REASONS (CITATIONS), yet it has been underexplored.
What are the advantages and disadvantages of opting for novel tasks compared to popular ones? Should you ever choose tasks that are neither novel nor popular?
Identify the challenges present in this task that you aim to enhance, and provide citations to previous works that have addressed these challenges:
Despite the great advancement, CHALLENGE is challenging because REASONS (CITATIONS).
Concisely describe the major approaches that have been developed to tackle these challenges and cite the corresponding works:
Several stuides have proposed APPROACH to overcome this issue (CITATIONS).
It should serve as an overarching description of previous works that have employed similar approaches.
Explain the limitations associated with those approaches.:
However, it is still limited by REASONS.
If your work addresses multiple challenges, you may consider dedicating a paragraph to each challenge, where each paragraph follows all the steps outlined in this section.
Provide the overview of your project.
In this paper, we present APPROACH to improve CHALLENGE (Section #).
You may need to describe multiple approaches if your work serves multiple objectives.
Explain the hypothesis underlying your approach by specifying which methods are expected to address particular issues:
We hypothesize that METHOD can resolve ISSUE because REASONS.
Our approach uses METHOD to resolve ISSUE because REASONS.
What distinguishes "approach" from "method"?
How do we differentiate between "challenge" and "issue"?
Our approach is evaluated on DATA with MODELS (Section #).
You may not want to mention the exact names of the datasets or models, but instead refer to their types to emphasize the generalizability of your work.
Our experiments show that FINDINGS (Section #).
At this stage, you do not know the outcomes of the experiments. Thus, write your expected findings here.
List specific contributions of your work:
This work makes THREE main contributions as follows:
We create a new dataset for TASK.
We introduce a new model for TASK.
MODEL achieves the state-state-of-the-art performance on TASK.
A real system is built for TASK.
Indicate the primary contribution of your work (if any):
To the best of our knowledge, this is the first work that does APPROACH for TASK.
This work will facilitate the development of ...
The scope of references here should encompass a broader range than the ones cited in the section.
Finding the most relevant works at this stage is strongly advised, especially considering the limited timeframe to complete the section, which should be used to understand the work rather than to explore additional sources.
Elaborate on your approach and reference the section:
Specify the types of data and/or models used to evaluate your approach, explain the reasons for using them, and reference the section for detailed information on this aspect:
Provide a high-level description of the primary findings from the analysis that offer supporting evidence for your hypothesis, and reference the section where a comprehensive breakdown of the results can be found: