# Classify feedback with an uncertainty bucket Open Practice / Measured machine learning / curriculum 1.0 Estimated guided practice: 4h estimated. This estimate has not been learner-tested. Original practice material · CC BY 4.0 · Non-degree and non-accredited. ## What you will build Ship an annotation and error guide with inputs, tests and a reproducible handoff. ## Before you start - Python onramp; inspect the worked fixture before beginning. - A text editor/spreadsheet for the manual route. Hosted Python/Jupyter for code exercises; applicable dependencies are in requirements.txt. - For actual AI calls, a legitimate available cloud service and authorised learner credentials; no particular provider or country is mandatory. ## Learning outcomes - Make review an allowed outcome and distinguish requests from issues. - Show exactly which tokens produce each label. - Keep tokenisation and decision rules separate. - Record failures rather than hiding them behind a high average score. - Keep a reproducible result and disclose safety, rights and regional access limits. ## Activity 1: Define labels with examples Estimated time: 35 minutes. ### Outcomes - Make review an allowed outcome and distinguish requests from issues. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Make review an allowed outcome and distinguish requests from issues. Inspect the supplied input before using any tool. Write the decision rules in your own words and calculate one small result independently. The reference output is an answer key for this deliberately bounded fixture, not proof of general capability. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. For learners across regions underserved or excluded by mainstream providers: verify country eligibility, language, account/identity requirements, data terms, cost, quota and connectivity before choosing a cloud service. Use only legitimate available alternatives; never bypass restrictions or invent a country/account. No service access? Complete the labelled manual/fixture route without claiming a live AI run. ### Worked example ```text SOURCE (authored fixture): { "texts": [ "please refund my order", "the item is broken", "refund this broken item", "hello there" ] } INDEPENDENT FIXTURE TARGET: { "labels": [ "request", "issue", "review", "review" ], "review_count": 2 } ``` ### Guided activity 1. Make review an allowed outcome and distinguish requests from issues. 2. Open input.json and expected.json from the practice pack; keep an unmodified original. 3. Record a manual prediction and the relevant rule before consulting the answer key. 4. Perform one small change and predict its effect before rerunning. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Define labels with examples — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Not broken still contains the token broken: document the baseline failure. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## Activity 2: Trace a keyword baseline Estimated time: 35 minutes. ### Outcomes - Show exactly which tokens produce each label. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Show exactly which tokens produce each label. Establish a baseline before changing prompts or code. Keep the original input immutable. Record expected values, observed values and whether each comparison passed. If the expected file conflicts with a justified policy, investigate rather than forcing a convenient match. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. ### Worked example ```text MANUAL BASELINE / compare every named field { "labels": [ "request", "issue", "review", "review" ], "review_count": 2 } Keep a separate observed-result column; the answer key is not a model response. ``` ### Guided activity 1. Show exactly which tokens produce each label. 2. Open input.json and expected.json from the practice pack; keep an unmodified original. 3. Record expected and observed values separately; explain every mismatch. 4. Perform one small change and predict its effect before rerunning. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Trace a keyword baseline — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Add punctuation: tokenisation should preserve the intended word boundary. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## Activity 3: Build a testable text pipeline Estimated time: 35 minutes. ### Outcomes - Keep tokenisation and decision rules separate. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Keep tokenisation and decision rules separate. Build one inspectable transformation or bounded AI prompt. Keep calculation and validation deterministic where possible. If you make a permitted hosted call, save the actual response separately from the authored examples and verify its material claims; the supplied reference never simulates a live model. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. ### Worked example ```text PROJECT CONTRACT: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. HOSTED REFERENCE: open hosted-lab.ipynb or run reference.py with all fixture files together. EXPECTED CONSTRUCTED RESULT: { "labels": [ "request", "issue", "review", "review" ], "review_count": 2 } ``` ### Guided activity 1. Keep tokenisation and decision rules separate. 2. Open hosted-lab.ipynb and the reference code from the practice pack; keep an unmodified original. 3. Record expected and observed values separately; explain every mismatch. 4. Perform Try another language: abstain and obtain appropriate evaluation before use. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Build a testable text pipeline — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Try another language: abstain and obtain appropriate evaluation before use. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## Activity 4: Attack ambiguity and negation Estimated time: 35 minutes. ### Outcomes - Record failures rather than hiding them behind a high average score. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Record failures rather than hiding them behind a high average score. Change one assumption at a time and predict the result first. Retain the failure, the revision and the rerun evidence. A failure is useful information: hiding a bad output or editing a test target after seeing the answer destroys the value of the experiment. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. ### Worked example ```text CHANGE TESTS — predict before running: 1. Not broken still contains the token broken: document the baseline failure. 2. Add punctuation: tokenisation should preserve the intended word boundary. 3. Try another language: abstain and obtain appropriate evaluation before use. Keep the original expected.json unchanged. ``` ### Guided activity 1. Record failures rather than hiding them behind a high average score. 2. Open failure-tests.md from the practice pack; keep an unmodified original. 3. Record expected and observed values separately; explain every mismatch. 4. Perform Not broken still contains the token broken: document the baseline failure. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Attack ambiguity and negation — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Not broken still contains the token broken: document the baseline failure. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## Activity 5: Compare a hosted model carefully Estimated time: 35 minutes. ### Outcomes - Use the same human-labelled set and do not infer emotion from a label. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Use the same human-labelled set and do not infer emotion from a label. Review what the fixture cannot establish. Explain who owns the decision, where human approval is needed and which access or professional constraints remain. Do not describe a small educational success as production readiness, certification, legal compliance or a guarantee. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. For learners across regions underserved or excluded by mainstream providers: verify country eligibility, language, account/identity requirements, data terms, cost, quota and connectivity before choosing a cloud service. Use only legitimate available alternatives; never bypass restrictions or invent a country/account. No service access? Complete the labelled manual/fixture route without claiming a live AI run. ### Worked example ```text REVIEW BOUNDARY: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. ACCESS / NOT TESTED: unavailable provider, unsupported language and absent external feedback must remain explicit. ``` ### Guided activity 1. Use the same human-labelled set and do not infer emotion from a label. 2. Open README.md and your evidence log from the practice pack; keep an unmodified original. 3. Record the human decision owner, limitations and any untested service access. 4. Perform one small change and predict its effect before rerunning. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Compare a hosted model carefully — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Add punctuation: tokenisation should preserve the intended word boundary. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## Activity 6: Ship an annotation and error guide Estimated time: 65 minutes. ### Outcomes - Deliver the label definitions, confusion examples and review policy. - Explain and retain a testable result and its limits. Text labels encode a task definition, not access to a person’s inner state. A transparent keyword baseline reveals ambiguities before complex models are introduced. Preserve negation and route conflicting evidence to review rather than forcing a convenient class. Deliver the label definitions, confusion examples and review policy. Deliver a reproducible bundle rather than a screenshot alone. Include inputs, rules, output, failure tests, a short README and residual limitations. Ask another person to follow the instructions if possible; self-review and actual external review must be labelled separately. Policy for this project: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. ### Worked example ```text HANDOFF MANIFEST: input.json — immutable source expected.json — fixture answer key actual.json — result actually observed evidence-log.csv — baseline plus three changes README.md — reproduction and access route limitations.md — unresolved failures, rights, approvals and scope Never bundle a credential. ``` ### Guided activity 1. Deliver the label definitions, confusion examples and review policy. 2. Open README.md and your evidence log from the practice pack; keep an unmodified original. 3. Record the human decision owner, limitations and any untested service access. 4. Perform one small change and predict its effect before rerunning. 5. Keep the result and your reasoning in the lesson notes; do not mark a simulated run as live. ### Expected evidence - Ship an annotation and error guide — a saved, inspectable artefact tied to the supplied rule. - A baseline or change-test record with expected/observed results and a limitation. ### Common mistakes - Try another language: abstain and obtain appropriate evaluation before use. - Treating a supplied answer key, an authored example or self-review as independent field validation. ### Check your understanding Which exact input, rule and expected value would let another learner falsify your conclusion in this activity, and what remains untested? ### Offline route Use the supplied fictional JSON, expected results and paper/text evidence log without a provider account. Manual or constructed completion is labelled; it is not a live cloud integration or trained-model run. ## End-of-path review These are self-check questions, not verified examinations. ### Question 1 For the supplied fixture, what is the expected labels? 1. ["request", "issue", "review", "review"] 2. It cannot be checked against the supplied input. 3. Whatever value the model reports most confidently. Review answer: 1. The independent answer key only validates this constructed fixture. Preserve the source, rerun counterexamples, and disclose what was not tested. ### Question 2 How should you handle the project rule: Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. 1. Silently change the rule to fit an attractive output. 2. State the rule, test it, and record exceptions or unresolved failures. 3. Treat model confidence as proof. Review answer: 2. Rules, source provenance and counterexamples make an output inspectable; an attractive result does not override policy. ### Question 3 If mainstream providers do not serve your region, what is the honest route? 1. Use a false country or borrowed account. 2. Claim the supplied fixture was a real API run. 3. Check legitimate available alternatives, or label the manual/fixture route and its limits. Review answer: 3. Access includes country, identity, language, cost and data terms. Do not bypass restrictions or invent a live result. ## Capstone review rubric Score each criterion 0 (missing), 1 (partial), or 2 (demonstrated). A self-review pass needs at least 8/10 and 2/2 on the critical safety/limits criterion. This is learner practice feedback, not verified assessment. ### Defined task and preserved input Original input and the project policy are retained; the actual question is clear. ### Checked baseline The supplied baseline reconciles field by field; unsupported live-run claims are absent. ### Three failure/change tests All three specified changes have predictions, observed results and explanations. ### Repeatable handoff Another person can follow the README; supporting practice evidence and unresolved limits are included. ### Safety, rights and access honesty [critical] Token match refund => request and broken => issue. Both or neither => review. This brittle baseline is educational, not a reliable sentiment or emotion detector. Regional access route is legitimate; credentials and private data are absent; manual/constructed work is clearly labelled. ## References - [Scikit-learn text feature extraction](https://scikit-learn.org/stable/modules/feature_extraction.html#text-feature-extraction) ## Editorial boundary Original authored practice material; deterministic fixtures build-tested. Not independently expert-reviewed or learner-validated. Time estimates unvalidated. Use fictional/redacted data and your own authorised provider credentials. Never include credentials in this pack or public work.