Psych 245A:
Computational Models of Language Learning

Discussion questions

Participation in class discussion

Presentations

Writing assignment

Everyone is expected to contribute to the class discussion. To prepare for this, you should come up with at least two questions or observations about the assigned readings. This will ensure that you have something you are comfortable discussing in class. Note: Discussion points should not be something that you can easily google the answer for.

Before coming to class, it will help if you can summarize the key points of the reading selection(s) to be discussed that day. You should be able to follow the flow of logic. Keeping these questions in mind will help direct your reading (and help you come up with discussion points):

• What is the theoretical or empirical issue being addressed? If this is a review article, what are the main questions that are being addressed in this area?
• For theoretical sections, what are the key assumptions and data used as the foundation of the argument? Do you agree with them?
• For experimental and modeling studies, what is the specific method and how does it work? Do you understand the experimental/modeling design? Do you understand the figures in the paper? What do the figures represent? Are the overall conclusions/interpretations supported by the data in the results?
• What are the main results & implications of the paper? For empirical studies, keep in mind at least one set of data points or one figure that illustrates the main gist of the paper. For review papers, keep in mind the current successful approaches that are being pursued, what the current findings are for the big questions, and what big questions remain.
• How does this relate to other selections read that week, or in previous weeks? How does it relate to the reference readings? Most selections are chosen precisely because they are on a similar topic as others during that week. All of them aim to answer something about computational models of language, so keep in mind what they're trying to tackle. Can you think of any extensions that might logically follow from the current studies?
• Important: Don't get hung up on every little wrinkle. A given article may have terminology you're unfamiliar with and background assumptions you don't share. Don't panic: This is what class discussion is for.

Preparing two discussion points is meant to help you be more comfortable joining the discussion in class. You should attempt to say something at least once during the session.

You are responsible for leading the discussion for at least one reading selection covered during the course. For each class discussion you lead, your grade will be based on how well you lead the discussion - you certainly don't need to know everything, but you need to have thought about what the article's about so we can all try to figure out any problem areas together.

In addition, you will need to review at least some of the reference readings (aim for two, if available) to understand how they relate to the main reading. Select some that seem interesting to you. Please note that you are not responsible for all the reference readings.

You should prepare a presentation that covers the main points of the paper, including:

• the theoretical or empirical issues being addressed and/or the main questions
• any key assumptions or data that the main issues rest on
• experimental/model design (or different approaches used if this is a review paper)
• key figures or examples discussed
• main findings and/or relationship between reported findings and broader questions in the area
• relationship to selected reference readings

Because students will come to the class with different levels of background in the area, there are three options for what the final writing assignment can be:

• A literature review
• A detailed project proposal
• An implemented study + project description

No matter which option is selected, 30% of your writing assignment grade will be determined as follows:

• Peer review session participation: 10%
  This will occur during the final class session, and you receive credit by (i) having a draft of your writing assignment that other students can read (5%) and (ii) reading and commenting on the drafts of other students (5%). A peer review should aim to have the following components, which professional reviews typically have:
  
  • A sentence or two summarizing what the assignment is about.
  • A section for major comments (general issues/observations, suggestions for organization, etc.)
  • A section for minor comments (specific questions/observations about one part, typos, etc.)
  Comments on your assigned papers must be uploaded by Friday 3/17/17 at 3:30pm so that the writers can utilize these comments in their final submissions.
  Please note that comments should be in text format, and entered into the peer review area of the Canvas EEE assignment.
  
• Final presentation of writing assignment: 10%
  This is a 5 minute overview of what you covered in your writing assignment. You should prepare a few summary slides to aid your discussion of your work. All students will present during the designated session during finals week.
  Please note that your slides should be in pdf form.
• Overall writing style: 5%
  This is credit given for a writing assignment that is comprehensible, well-structured, and has proper spelling and punctuation.
• Citations and reference section: 5%
  This is credit given for (i) including relevant citations appropriately throughout the writing assignment and (ii) having a reference section at the end listing the citations in their full form. Look to the articles we read and refer to in class for a sense of how to do both these things appropriately.

The remaining 70% of your writing assignment grade will be determined by the grading scale specific to the writing assignment you choose to do.

Literature review

The literature review should aim to be concise (ideally about 4 pages, single-spaced + references afterwards).

The set of articles reviewed can consist of one of the following:

• only specific research articles (number required: at least four)
• only review articles (number required: at least two)
• a combination of review articles and specific research articles (number required: at least one review article and at least two specific research articles)

The goal of the literature review is to synthesize the ideas being presented, rather than simply summarizing what each article says. So, for any combination of articles being reviewed, the following must be included:

• Main questions/issues of each article, and how they relate to each other (Are they overlapping in any way? What's similar vs. what's distinct?)
• Specific methodologies used, and how they relate to each other (Again, do they overlap? Are there similar approaches or are they very distinct?)
• Discussion of results, and how they relate to each other (Do they agree or are there discrepancies? Are there higher level similarities? Do the results build on each other to create a larger, coherent whole?)
• Discussion of open questions, and how they relate to each other (Is there overlap? What's similar vs. what's distinct?)

Remember: You're aiming to describe the current state of research that these articles collectively represent. Someone should be able to read your literature review and have a good idea about what the main approaches are to the question(s) being investigated, what the interesting results are, and what questions remain, as represented by the articles in your review.

Detailed project proposal

The detailed proposal should aim to be concise (ideally about 6 pages, single-spaced + references afterwards).

You must include the following in your proposal:

• The question you are trying to address
• Relevant previous work on this question. Include whatever detail is relevant for your current proposal -- you don't need to include everything they previously did.
• The methodology you would use (the specific algorithm, computational framework, example stimuli, potential relevant data sets, etc.)
• Possible results, and what implications each set of results would have for the question you are trying to address. Since you are proposing a study rather than carrying it out, you obviously will not have results yet. However, your study was probably motivated by what you would expect to happen - so your discussion of possible results is very important. What are the implications for the different possible outcomes in your study?
• How the results would fit into the "big picture". What question would it answer, would it answer it better than previous studies on this topic or address shortfalls of previous studies, etc.

Remember: You're aiming to create a proposal that's detailed enough for you (or someone else) to carry out the project described. Whoever's carrying it out should be able to read your proposal and know exactly why it's interesting to carry out the project described.

Implemented study + project description

You may choose to implement a study, particularly if it's in your area of research (and so you've been working on it already). In this case, you'll write a more targeted paper similar in structure to the detailed proposal. However, because a segment of your final project is the implementation of the study itself, you can provide an abbreviated version of the aspects required for the detailed project proposal. Your paper should focus on these aspects:

• The question you are trying to address and why this is a sensible study to answer that question
• The specific experimental paradigm/model/analysis you used and the data you used
• Detailed results (graph or table, etc.)
• Discussion of what the results mean and how to interpret them with respect to the question you're trying to address

The paper should probably be no shorter than 3 pages single-spaced and no longer than 6 pages single-spaced, not including references.

Remember: You're aiming to describe your implemented study so that someone understands why it was created, how exactly it was implemented, and what the results mean.