THIS RUBRIC WILL CHANGE SLIGHTLY BEFORE WEEK 6. PLEASE CHECK
BACK FOR THE FINAL VERSION
For a successful final project, you will be graded on the following
dimensions, each contributing in varying degrees to your final grade.
They are presented roughly in order of importance, and the four ‘Levels’
of mastery can be thought of as corresponding roughly to ‘95%’, ‘80%’,
‘60%’, and ‘0%’. Your overall grade for the project will be based
roughly on your overall level across the categories, adjusting for other
factors (unclarity, missing sections, exceptional or low effort,
etc).
Three Big Things
1: This is about language
The biggest thing to remember is that this is a class about
language and computers. Some students write very compelling
papers about human computer interaction, or interaction design, or
troubleshooting Python installations for NLP, or the difficulties of
designing tools to interface with a certain task, but completely fail to
demonstrate that they’ve paid any attention in LIGN 6, by neglecting the
important, language-focused elements of the virtual assistant or NLP
tools. A paper which is of the proper length, mentions all of the steps
of the interaction process (where applicable), and describes an
interesting tool can still be an “F” quality paper if it never engages
with natural language processing and linguistics. So, in each step of
your paper writing process, ask yourself ’How am I showing the
linguistic knowledge about natural language processing that I’ve gained
from LIGN 6?”
2: Effort Matters
Every year, I get one or two students who attempt to submit
exceptionally short, shoddy, or last-minute work, figuring ‘oh well,
I’ll probably get a C and that’ll be OK’. And every year, those students
are shocked when they realize that they’ve just earned a 15% on
something worth 30% of their grade. Papers which do not even come close
to meeting the length guidelines, just engage with the rubric in a few,
token measures, or show little understanding of course material are
really frustrating as an instructor, and are a sign that you simply
didn’t try very hard.
So, please, start early, put in effort, and make sure that you’re
meeting the standards in the rubric. And if you choose not to do these
things, please act with integrity and understanding when you fail the
assignment.
3: Self Grading
We’re experimenting with self-grading this quarter. So, in addition
to writing and completing your project, you’re going to grade it too,
according to the rubric below, giving yourself a score for each rubric
item as well as your honest appraisal of what you’ve earned on the
project.
This will be treated as the ‘default’ grade for your project, and as
we look at your project, the main question we’ll be asking is ‘Has this
student accurately assessed their own work?’. If we feel you were fair
to yourself and your classmates, your grade will stand. But if we feel
that your grade does not reflect your work and efforts, we’ll adjust the
grade down (or up!) accordingly.
To see what this looks like, see the self-grading
Guidelines with sample cover sheet
Project Proposal
Your project proposal should be around one single spaced page, plus
or minus, providing and discussing the following information:
- Your name
- Who you’re working with
- What you’re planning to do in your final project
- If you’re planning to do Option 1:
- What’s your system designed to do?
- What are some sample queries/commands? What answers should the
system give?
- What are a few of the complexities that make this different from
what existing assistants do?
- What do you think will be ‘the hard part(s)’ for your system?
- If you’re planning to do Option 2:
- What task do you want to work on?
- What tool are you interested in using?
- What kind of experience do you have with similar tools or languages?
(e.g. explain why this is possible)
- What kind of data do you want to use to test your tool?
- If you’re planning to do Option 3:
- What do you have planned?
- What is the final product?
- Why is this topic interesting to you and your group members?
I’ve uploaded a sample project
proposal which is basically perfect. It’s describing an ‘option 3’,
off the beaten path project, and it’s a bit extra, so you can get full
credit doing a bit less work than this, but this is exactly the kind of
work I’d love to see in your proposal.
Grading Rubric
A Masterful proposal will…
- Discuss all of the elements above, providing concrete examples and
deliverables
- Provide clear discussion, particularly in option 2 projects, of how
you plan to implement this
- Lay out the steps required to complete this project in a clear
outline
- Show evidence of careful consideration of the problems and
difficulties involved.
- Show clear understanding of the challenges, particularly for option
2, and discuss the how the student(s) can meet them
An Acceptable proposal will…
- Discuss all of the elements above, but with fewer examples or direct
deliverable items
- Provide discussion of how you plan to implement the project, but
with some areas of vagueness or large missing steps
- Focus more on deliverables than the process of getting there, with
little evidence of planning
- Feel a bit more rushed, or like a rough gesture in the direction,
rather than a clear set of plans
- Show understanding of the challenges involved, but doesn’t show how
the student(s) are equipped to meet them
A Novice proposal will…
- Discuss most or all of the elements above, but perhaps omitting some
examples.
- Provide only vague information about the planned direction
- Promise an outcome with no discussion of the process
- Feel last-minute, under-planned, or show little effort.
- Be substantially shorter than the one page target length
A Way Off proposal will…
- Fail to discuss many or all of the elements above.
- Describe only a vague plan, with little differentiating
information
- Give no details about the planned process
- Feel like little effort was spent in creating the proposal
- Be substantially shorter than the one page target length
Rough Draft
You are encouraged (but not required) to submit a rough draft to me
at any point prior to the start of finals week. This should be a largely
structurally complete paper, although leaving some schematic areas,
pending a question for me (e.g. “Should I go into [subtopic] here?”) is
very much acceptable. This is a great opportunity to make sure you’re on
the right track, and to check in to make sure you’re executing your
topic well. Rough drafts will get general and specific feedback, but of
course, there may be issues which I don’t ‘catch’ during that initial
read or new issues slipping in, so you’re not guaranteed to be 100%
problem free even if you address all my feedback. Most importantly, this
is a great way to avoid being ‘surprised’ by a poor final grade on the
paper, if you’ve missed a major component of the rubric or of the
complexity of your system.
Given that you have the opportunity to submit a draft, as well as to
discuss your paper with me at any point, your final paper grades are
final and non-negotiable. Even if you were ‘way off’ and
misunderstood the prompt, rubric, or goals of the assignment, you had
ample opportunity to get feedback or submit a draft, so you will not be
offered opportunities for a ‘do over’.
Required Cover Page and
Self-Assessment
Project Option 1: Design a
System
Cover Page and Self-Grading
All projects are required to include a cover page and self-grading.
To see what this looks like, see the self-grading
Guidelines with sample cover sheet
Projects which do not include a cover page with self
assessment will have 30% deducted
- Masterful: Completes and includes a cover page
including at least:
- Project Title
- Name of the Project Host (both the person and the organization)
- Name of the Faculty Advisor (if different)
- Name of all participants
- List of any supplementary files or urls (e.g. source code links,
website link, etc)
- Whether you want detailed feedback, or just a final grade
- Your self-assessment, including a rubric leve
- Your suggested project grade
- Way Off: Any element of the cover page is
missing.
Scope of Writeup
- Masterful: Discusses, indirectly or directly, the
‘Things to consider’ for all elements in the Virtual
Assistant Interaction process guide in sufficient depth to
demonstrate understanding of the difficulties (or simplicities) of their
particular system, or, when a given step doesn’t apply, explains why. If
additional steps are required, they are described in detail, and the
ethical concerns in the process guide are addressed as well.
- Acceptable: Discusses all of the elements in the
process, but in one or two areas shows lacks nuance or fails to capture
or discuss important complexities faced by their particular system.
- Novice: Fails to thoroughly discuss some of the
elements of the interaction process (without justification of why that
step doesn’t apply), or shows repeated failures of understanding of the
process and complexities.
- Way Off: Writeup focuses exclusively on one or two
elements of the process without prior approval, does not address the NLP
complexities of multiple steps, or does not demonstrate understanding of
the many intermediate steps in the NLP chain.
Demonstration of Knowledge
- Masterful: Student paper shows considerable and
nuanced knowledge and attention to the material from the course, making
regular reference to concepts from class and homeworks, fully engaging
with the material taught in class, and potentially, goes beyond it into
independent research. Additionally, the student shows some realistic
understanding of the ease or difficulty of designing the system as
described.
- Acceptable: The paper shows attention and
engagement with the class materials, but occasionally neglects some
important facts or discussions from the course, or occasionally shows
confusion with the nature, functioning or limitations of some of the NLP
processes discussed.
- Novice: The paper makes occasional reference to the
concepts discussed in class, but repeatedly shows failures of
understanding of the concepts discussed or their applicability to the
system in design.
- Way Off: Writeup largely fails to demonstrate
attention to or knowledge of the material from the course, instead
focusing on non-NLP related elements or giving explanations so
superficial as to be available even to those who haven’t taken LIGN 6.
Note that papers which make little reference to concepts learned
in class will be penalized disproportionately, as you had one
job!
Richness
- Masterful: Project describes a system with
considerable interactional depth, discusses 5-10 different commands,
involving back-and-forth interaction with the human, creation of a
discourse context, and complexity of commands at or beyond current
commercial offerings (e.g. Siri or Alexa).
- Acceptable: Project describes a system with some
richness, but which primarily processes single complex commands and
gives output, without considerable interaction or discourse. Or, the
student focuses exclusively on one or two commands, without prior
approval.
- Novice: Student describes a simple
call-and-response system with low NLP complexity, where pre-specified
commands are responded to with canned phrases or actions. (e.g. “Alexa,
what is your name?” “My name is Alexa, thanks for asking.” or “What is
the best menu choice for a vegetarian?” “Pasta Primavera”)
- Way Off: The system described makes only limited
use of natural language processing, or the system’s principal described
complexities are outside of the NLP world (e.g. for a pizza ordering
system, the bulk of the paper is spent discussing the robotic creation
of the physical pizzas).
To do this assignment well, expect to write 4000-5000 words. Note
that students with strong command of the material might be able to excel
in a bit less, and students who are struggling could easily provide
10,000 words without showing their knowledge. Note that papers submitted
by groups will be expected to have a greater scope and detail than
individual papers, so a 5000 word paper from five people would be
considered ‘suspiciously short’, but it’d be fine from a single person.
Additionally, feel free to use APA or MLA formatting as a baseline,
particularly for citations, and you should use hierarchical formatting
(e.g. labeled sections, subsections, subsubsections), but I will not be
grading on deviations from an arbitrary set of formatting laws, because
we both have better things to do with our lives.
- Masterful: The paper is sufficiently long to
demonstrate knowledge, richness of the system, and show proper scope,
but without dragging or feeling like words are being added for the sake
of talking. Formatting is reasonable, readable, and enhances the text.
Citations, where given, are reasonably formatted according to an
accepted standard, and contain the information needed to be
followable.
- Acceptable: Paper length is sufficient, although
perhaps a bit too brief or too wordy. Formatting is acceptable, if
occasionally distracting or adding difficulties.
- Novice: Paper is ‘a bit too short’ to accomplish
the needed demonstration of knowledge, or buries lack of knowledge in a
mountain of text. This could also apply if the formatting is
distracting, problematic, or hurtful to the argumentation.
- Way Off: The paper is so short as to be unable to
address the appear ‘last minute’ or ‘low effort’, and not given enough
space to demonstrate knowledge. Or is typeset using ‘Comic Sans MS’ or
‘Papyrus’ fonts, is presented without formatting or sectioning,
citations are missing crucial information
Structure and Organization
Please structure your paper with sections, subsections, and
subsubsections (where needed), to make things easier both in reference
and readability. Don’t worry about ‘transitions’, just as long as the
organization is OK.
- Masterful: The paper is well organized, with the
introduction discussing the structure of the paper to guide the reader.
Sections are numbered and cross-referenced throughout the paper where
needed. Transitions between sections are clear, and the sections make
sense. (Hint: The Interaction Process guide is a pretty reasonable
outline!)
- Acceptable: Structure of the paper is acceptable,
if somewhat difficult to understand or poorly explained. The reader has
a rough understanding what’s going on, even if it’s never
explained.
- Novice: The paper’s structure is counterintuitive
and not understandable, with the structure (or lack thereof) hurting the
argumentation considerably.
- Way Off: The paper has no discernable structure nor
organization.
Language and Argumentation
- Masterful: Student language use is consistently
clear and understandable, with a reasonable level of formality for
academic writing. Additionally, the paper was clearly proofread, with
typoes and English language grammar issues relatively rare and not
considerably affecting the argumentation.
- Acceptable: Student language use is largely
understandable, but in places unclear, or with a sufficient number of
typoes or English language grammar issues as to start to hinder my
understanding of the argumentation.
- Novice: The argumentation is regularly unclear, and
my ability to understand the content of the paper is considerably
hindered by typoes, difficult to understand argumentation, or English
language grammar issues.
- Way Off: The paper is not written in English, is
largely incomprehensible, or shows evidence of automatic machine
translation.
Academic Integrity and
Citation
Note that plagiarism or other
academic integrity issues will result in an automatic ‘0’ on the
paper.
- Masterful: All direct quotes or external references
are cited, with the source given clearly in a ‘Works Cited’ section, in
the format of your choice (APA is a fine default) including the
necessary information to track down the resource. All other words are
your own.
- Acceptable: All direct quotes or external
references are cited, but the citations do not include enough
information for the reader to locate the original resource.
- Way Off: Sources are not cited, or there is
evidence of academic dishonesty. Be careful to cite all your sources and
quotes (outside of the class material).
Project Option 2: Implement
some NLP
Cover Page and Self-Grading
The rubric here is identical to that for Option 1.
Scope of Writeup
- Masterful: Discusses, indirectly or directly, all
ten elements of a successful project option 2 writeup described in the course syllabus
guide in sufficient depth to demonstrate their engagement with that
particular system, or, when a given step doesn’t apply, explain why. If
additional steps are required, they are described in detail.
- Acceptable: Discusses all of the elements, but in
some areas shows lack of depth, missing elements, failure to capture or
discuss important complexities faced by their particular system.
- Novice: Fails to discuss some of the elements of
the interaction process (without justification of why that step doesn’t
apply), or shows repeated failures of understanding of the process and
complexities.
- Way Off: Writeup focuses exclusively on one or two
elements of the process without prior approval, or does not demonstrate
understanding of the many intermediate steps in the NLP chain.
Demonstration of knowledge
- Masterful: The student has successfully installed
the tool locally, and has demonstrated the use of the tool on real
natural language data.
- Acceptable: The student shows that they’ve managed
to make the tool work, and demonstrates learning in that process.
- Novice: The student has not managed to successfully
implement the tool, but presents deep discussion as to the
difficulties of this above and beyond ‘computers are hard’, engaging
with the NLP aspects of the process.
- Way Off: The student fails to install the tool
without discussion, or demonstrates the ability to install or implement
the tool, but does not engage with the NLP elements of it, going little
further than ‘Hello World’. Or, student has used an online
implementation of this tool, e.g. a web interface.
Richness
- Masterful: Student paper shows considerable and
nuanced engagement with natural language processing and the tool(s) of
their choice, showing that they’ve not just ‘managed to install it’, but
have begun to demonstrate skill and facility with it, and have applied
it in ways that are revealing and demonstrative for NLP. Student also
provides nuanced discussions of the failings and stengths of the system,
making concrete reference to their work.
- Acceptable: The student has implemented the tool,
but without discussion of the implications for NLP and for the
implementation of this tool in production settings. Discussions of the
failings and stengths of the system are general, or fail to engage in
consistent depth.
- Novice: Results from the tool’s run are presented
with major problems, or without discussion and analysis.
- Way Off: The tool is not implemented, or results
are only on a test file.
The rubric here is the same as for Option 1, but the expected
length will be shorter (5-7 single spaced pages) because more
of the work is happening ‘outside the paper’.
Structure and Organization
Option 2 projects, in particular, lend themselves well to a
section-by-section write-up, with large chunks of code/output
interspersed. The rubric for this portion is the same as for option
1.
Language and Argumentation
The rubric here is identical to that for Option 1.
Academic Integrity and
Citation
The rubric here is identical to that for Option 1.
Option 3: Build-Your-Own
With a build-your-own project, it’s difficult to describe the exact
expectations, as they vary from paper to paper. However, you can always
email me to get clarifications if you’re unsure what I’m looking for.
These do tend to be a bit longer, closer to 10-15 pages single spaced,
particularly if they’re more research-based papers (e.g. “The history of
TTS”).
However, the even for ‘Build your own’ papers, you’ll still grade
based on the ‘Formatting and Length’, ‘Structure and Organization’,
‘Language and Argumentation’, and ‘Academic Integrity and Citation’
rubrics for option 1.