Three ideas.
One direction.

Deliverables and deadlines

The brief gives teaching weeks, not exact calendar dates. Confirm dates through the lecturer or eLearn rather than deriving them from the planning document's title.

Assignment 1 format and content

The report must be handwritten in pen on A4 paper, with two content pages excluding the cover, references, and appendix. Every member must handwrite at least one section and put their name at its start.

It must cover application overview, evidence of value and impact, comparison with the pre-AI approach, environmental and social/economic sustainability, at least one relevant UN SDG, technical limitations, ethics, and a justified reflection about adoption.

The poster must be one A3 sheet, hand-drawn and hand-coloured. Include the problem, a simple AI workflow, benefits, sustainability impact, one key risk, and a recommendation for a general audience.

The report cover and poster reverse must contain the programme, academic session, course code/name, group number/name, and a numbered list of members with names, IDs, and tutorial groups. Submit both physical items in class and upload separate, clear scanned PDFs to eLearn.

Required filenames:

Group[Number]_[GroupName]_A1_Report.pdf
Group[Number]_[GroupName]_A1_Poster.pdf
Group[Number]_[GroupName]_A2_Prototype.ipynb

Declare generative AI tools, purpose, and how outputs were checked in an appendix. Record reused code and sources honestly. This reference is AI-assisted planning material; group members should write their final Assignment 1 analysis themselves and draw their own poster.

The brief lists deductions for incorrect filenames, poor or illegible report handwriting, typed/printed reports, and posters containing printed, typed, or digitally created elements. Its later tables specify -1% for filename errors, -1% for report handwriting, and -5% for the stated report/poster format violations; the opening page uses marks instead. Follow the physical-format rules strictly and clarify the deduction units.

Marking priorities

The rubric converts each criterion's 0-5 score using (score / 5) * weight. A working feature is only part of the assessment; evidence, explanation, sustainability, ethics, and individual understanding also matter.

Details to clarify with the lecturer

• Report allocation differs: the criteria give reflection 1% and appendix 1%; the rubric gives reflection 2% without an appendix row. Include both components.
• The brief requires AI-use disclosure but also penalises AI-generated text/images in Assignment 1. Disclosure does not automatically make copying or hand-copying AI content permissible.
• The opening instructions describe deductions in marks, while later tables use percentages. Do not assume the units are interchangeable.
• Late work within one week is capped at a stated "10%"; work later than that is zero. The basis of the cap needs clarification.

Keep the prototype small and show intermediate values. Use a few named functions rather than hiding all processing in one large block. Every diagram and claim in the presentation must match what the notebook actually does.

For a four-person group, primary responsibilities can be data/research, first technique, second technique, and evaluation/presentation. A fifth person can coordinate reproducibility and documentation. Responsibilities divide the work, not the understanding: each member must review and explain the whole submitted pipeline.

After each stage, have a member who did not write it trace one input through the code and make a small change. Rehearse questions about technique choice, malformed inputs, API failure, limitations, sustainability, ethics, and each member's contribution.

The excellent-level rubric asks for justified choices against alternatives, systematic testing, and evidence of the value of integration. Do not claim superiority simply because two techniques are present; use actual comparisons and report limitations honestly.

Assignment and original plan

• Official assignment brief Author-local attachment · request a shared copy
• Original handwritten project plan reviewed on 10 October 2026 Author-local attachment · request a shared copy

Local attachment links may only work on the author's computer. For teammates, distribute accessible copies of the supplied PDFs through the group's agreed channel.

Repository evidence

Source review on 10 October 2026 used the default main branches: Yinwei commit 297fadc0aebd831d6e3ba7f63977552ad67d96be and Class Interpreter commit 86362a3757e2cb419a68cacbd397f786e0bdbfa2. The links below pin the reviewed versions. Neither reviewed tree contained an .ipynb file. Other branches, unlocated/private projects, builds, model execution, audio acceptance, and prototype benchmarks were not verified. The repository search did not locate a matching claim-checking project among accessible/public results.

• Yinwei player overview: older phase checklist
• Yinwei audio/UI design contract: distinguish specification from current code
• Yinwei root React application: parent-meeting summary
• Yinwei native Flutter bridge
• Yinwei native/mock backend selection
• Yinwei Rust playback and WAV export code
• Yinwei Flutter listening/transfer limitations
• Class Interpreter overview and existing local API route
• Class Interpreter Python scoring, summaries, entry IDs and storage
• Class Interpreter browser capture, concept display and note export

Technical starting points

• scikit-learn: Text feature extraction and TF-IDF
• scikit-learn: Cosine similarity

AI assistance: Codex helped review the assignment, interpret the plan, inspect repository documentation/source, and draft this reference. The group should verify facts against the supplied brief and cited sources, independently understand any reused code, and test its eventual notebook. This file does not demonstrate that a prototype has been implemented or passed testing.