Q1. Build a Chatbot (13 Marks)
You are an RPA consultant specialising in the design of Conversational Agents, you have been engaged by Tourism Australia to consult on potential automation opportunities. One of your tasks is to design and build a chatbot application which can act as a virtual companion to tourists in Australia.
- Describe 5 functionalities related to tourism that can be fulfilled through a chatbot. (2 marks)
- Tourism Australia will be looking to collect data from the interaction of the tourists and the chatbot, and is seeking your advice in terms of the ethical considerations, what is some advice you will offer? (3 marks)
- Choose one of the functionalities that you have elaborated in a) above and design a chatbot flow using Dialogflow CX to fulfil this; outline your thought processes and the conversation design.
Demonstrate techniques that you have learnt in class. (5 marks)
- Record a demonstration introducing the bot and the sample flow that you have built. (less than 2 minutes) (3 marks)
You can upload the video on Microsoft Stream add the link in the report. Please make sure you give access to your teaching team.
Q2. Recommending Cloud Computing Systems (5 Marks)
Cloud computing has become an indispensable in a modern solopreneurs toolkit. Consider the following types of solopreneurs, choose any two and map out cloud computing-based systems that would address their specific requirements. Provide appropriate rationale and commentary for each system you recommend.
| Independent Real Estate Agent | • • | Managing and tracking leads Listing property |
| • | Virtual tour for high-quality property showcase | |
| • | Contract management with electronic signature capabilities | |
| • | Scheduling capabilities with integration to common calendar apps for ease of use
|
|
| Independent
Fitness Trainer |
•
• |
Customizable workout and nutrition planning
Video conferencing with high-quality video and audio for virtual training |
| • | Client management with progress tracking and reporting features | |
| • | Secure and simple payment processing | |
| • | Scheduling with reminders and appointment confirmation features
|
|
| Freelance
Software Developer |
• •
• |
Code editing and debugging tools
Source control management, Continuous integration and deployment tools Project management and issue tracking systems |
| • | Cloud computing platforms for development and testing | |
| • | Time tracking and invoicing
|
|
| Independent Photographer | •
• |
High-quality photo editing
Storage with large capacity for high-resolution images |
| • | Client gallery for showcasing and selling work | |
| • | Scheduling for booking shoots | |
| • | Capabilities to manage client relationships and follow-ups
|
|
| Independent
Content Creator (Blogger/Vlogger) |
• •
• |
Content management system with SEO optimisation features
Social media management and scheduling tools Video and image editing |
| • | Analytics and audience engagement tracking | |
| • | Monetization and advertisement management tools
|
Q3. Using off the Shelf Models (5 Marks)
You are developing a news aggregator platform and have been provided with a dataset of 200 news publications. Your task is to use a zero-shot classification model from Hugging Face to categorise these publications into appropriate sections of your platform utilising Google Colab and the transformers library. The sections are World, Sports, Business, Science and Technology, Entertainment, Lifestyle. You may choose any other sections you might find relevant as well. Critically discuss any insights you have gained from your categorisation process. (5 Marks)
Q4. Using Large Language Models (15 Marks)
You are a Data Analyst at a popular department store, the following is a prompt template which has been proposed for analysis and automation task based on customer reviews (referred from https://www.databricks.com/blog/actioning–customer–reviews–scale–databricks–sql–ai–functions).
“A customer left a review. We follow up with anyone who appears unhappy. Extract all entities mentioned. For each entity:
| • | classify sentiment as [“POSITIVE”, “NEUTRAL”, “NEGATIVE”] |
| • | whether customer requires a follow-up: Y or N |
| • | reason for requiring follow-up |
Return JSON ONLY. No other text outside the JSON. JSON format:
{
entities: [{“entity_name”: <entity name>,
“entity_type”: <entity type>,
“entity_sentiment”: <entity sentiment>,
“followup”: <Y or N for follow-up>,
“followup_reason”: <reason for follow-up> }]
}
Review: <review text>”
You have been asked to evaluate this proposed solution for use in the department store’s after sales automation process. Using at least three major freely available large language models, choose 10 reviews to work as your testing sample and analyse them.
Your analysis should include:
- The reasoning behind the choice of test sample. (1 mark)
- The results obtained from each model. (4 marks)
- An analysis of how each model’s output accurately reflects the sentiment expressed in the review and an appropriate response. (5 marks)
- Any notable differences in the results between the models, along with a critical analysis and your recommendations. (5 marks)
DELIVERABLES
- A professionally written report with the answer scripts to the 4 tasks. Include any links to recordings.
- The report should be submitted as a pdf, font size 11.
| Criteria D | C | B | A | |
| Build a
Chatbot
(13 marks) |
A minimal attempt at the Chatbot design and implementation. | A basic attempt at the Chatbot design and
implementation .
|
A good at the Chatbot design and implementation. | A complete attempt at the Chatbot design and implementation. |
| Recommend
ing Cloud Computing Capabilities
(5 marks)
|
A minimal attempt at a suitable recommendation with justifications. | A basic attempt at a suitable
recommendatio n with justifications. |
A good attempt at a suitable
recommendation with justifications. |
A complete attempt at a suitable
recommendation with justifications. |
| Analysis using off the
Shelf Models
(5 marks) |
A minimal attempt at the analysis and implementation. | A basic attempt at the analysis and implementation
. |
A good attempt at the analysis and implementation. | A complete attempt at the analysis and implementation. Provides detailed and critical insights from the analysis.
|
| Using Large
Language Models for Analysis
(15 marks)
|
Minimal understanding of the concepts. Incorrect or minimal use of large language models. Limited comparison between models and minimal critical insights. Unclear and unstructured analysis. | Basic understanding of the concepts. Some use of large language models with basic comparison and limited depth in insights.
Analysis is basic with some clarity issues.
|
Good
understanding of the concepts with application of large language models. Clear comparison with some depth and good insights. Mostly clear and structured analysis with minor issues. |
A complete understanding of concepts with proficient use of large language models. Detailed and thorough comparison with critical insights. Well-structured, clear, and articulate Analysis. |
