Skip to main content

Task Selection

A topic is a theme you and your team are interested in; a task is something you can actually build and evaluate. Where Research Topics asked you to commit to a topic, task selection asks your team to pick a well-defined problem within it, one with a clear input and output, a dataset or domain to work with, and a way to measure success. This section guides you to narrow a topic into a manageable task, identify gaps in current knowledge, and clearly articulate research ideas.

Exercise​

Work in your project team through the following steps to narrow your committed topic into a specific, executable task. Your goal is to have a concrete task with research objectives and contributions, as well as data and methods necessary to achieve them.

  1. As a team, survey the literature freely within your committed topic (e.g., Google Scholar, ACL Anthology, Papers with Code, arXiv, recent proceedings) until you can confidently answer the items below:
    tip

    This is a fast, exploratory pass to commit to a task. A deeper, systematic literature review comes later in Related Work.

    • What specific tasks are people tackling within this topic, and which one(s) seem most promising for you?
    • What data or domains are commonly used, and which are underexplored or newly available?
    • What methods currently perform best, and where do they still fall short?
    • What would a baseline for your project look like, and is it feasible to experiment within a semester?
  2. Based on your survey, select a specific task for your team and describe the types of data and/or domains you plan to focus on (10 min).
  3. Outline the primary research objectives of your project, which may include:
    tip

    The novelty of your research will be assessed based on these objectives. As you write the Intellectual Merit and Approach of your Framework later, make sure each objective is framed as a direct response to a specific challenge you raise there.

    • Devising an innovative methodology that outperforms previous approaches on specific or relevant datasets.
    • Constructing a novel dataset that facilitates task adaptation to new domains or languages.
    • Conducting a comprehensive analysis of state-of-the-art models, focusing on comparative aspects to gain insights into their strengths and weaknesses.
    • Developing a practical application employing existing resources to advance translational research.
  4. Create a concise slide presentation summarizing the above activities (one per group) as a preparation for your checkpoint presentation the following week.