Unit 5 project: A bot for a real job

The school office bot in Lesson 5.2 was a start. Today your group builds a bot for a job you choose — and tests it the way a real product is tested: with real questions from real people, not questions you made up.

Work in your project groups of 3–4. One period. Code in our_bot.py; the report in bot-report.odt.

Choose a job. Then ask five real people — students, teachers, the librarian, the lab assistant — what they would type to a bot doing it. Write down their exact words, at least 15 questions altogether. Collect only the questions: no names, no personal details.

Some jobs that work well:

• The library — timings, how to borrow, fines, where things are
• The computer lab — rules, how to log in, what to do when something breaks
• The exam timetable — which paper, which day, what to bring
• Typing practice — how to find your typing.com lesson, your goal, what WPM means

1. Start from school_bot.py. Replace its answers with at least eight keywords and answers for your job.
2. Normalise every question before matching, the way Lesson 5.3 did: lowercase, and strip ?, . and !.

3. Type every one of the real questions, exactly as they were given to you.
4. Mark each one: answered right, answered wrong, or didn't know.
5. Count: right out of total. That is your bot's score.

6. Make two changes that fix failures you found. Test again and record the new score.
7. In bot-report.odt, fill:
   • The job, and who it helps
   • Score before and after — out of the real questions
   • Three questions it still gets wrong, and why — use the stages of NLP: was it the words (lexical), the meaning (semantic), the connection to an earlier question (discourse), or what the person really meant (pragmatic)?
   • Its "I don't know" message — what should a person do next?

Each person writes one sentence for the person whose job the bot helps: what the bot will do for them, and one thing they should still expect to do themselves. Sign with your roll number. This part is marked for you alone.