The planned Ontario Garlic Tours is an online travel planner for tourists. Harnessing AI and LLM, it is intended to offer personalized recommendations and route planning, connecting tourists with farms, restaurants, and attractions, thereby boosting economic opportunities for Ontario's agriculture and tourism sectors, with potential applications beyond. The pilot will harness artificial intelligence and large language models (LLM) in the promotion of the tourism and agriculture industries. This initiative, with the pilot to launch August 2024, offers an opportunity for students to apply their skills in a real-world setting, addressing limitations in current online travel planners. Joining us in this endeavour are developers Reza Zeinali and Fariborz Lesani. Reza and Fariborz bring experience in software development, innovative technology applications, and a strategic approach to driving innovation across various sectors. The pilot project will benefit significantly from their expertise and visionary leadership.

Timeline: 

Current plan is to develop a beta model to launch August 2024, with a full-scale tool launching August 2025.

Ontario Garlic Tours: Benefits for All

• Tourists: The planner unlocks hidden gems for adventurous eaters. Instead of generic tours, it helps tourists curate agri-tourism trips full of local flavors. Farm visits, culinary events, and niche farm products offer personalized 'beyond the plate' experiences.

• Farmers: The planner forges direct connections with consumers, fostering brand awareness and boosting profits. It's a showcase for garlic and other farm offerings, increasing crop visibility and sales potential.

• Rural Economic Development: This tool reshapes 'peak season', making a region attractive during off-peak season. It drives traffic not just to farms, but to surrounding businesses – multiplying the economic benefit. The critical mass created by clustering offerings from multiple communities makes destinations more enticing, lengthens stays across communities, diversifies experiences, and contributes to regional brands. Data from the planner can guide future development decisions.

• Food & Beverage: Restaurants gain visibility among a foodie crowd, with hyperlocal promotion tapping into the 'farm-to-table' trend. Garlic acts as a hook, attracting diners eager for destination dining. The planner streamlines reservations and orders, boosting conversion without adding to a restaurant's workload. Participation showcases a commitment to local sourcing and positions establishments as part of a thriving culinary ecosystem.

The concept of single-food tourism celebrations and collaborative tourism –where multiple businesses and community partners participate in something greater than the sum of their parts–is not new. Other very successful single-food events include Maine Lobster Fest, Georgia Peach Fest, Bala Cranberry Festival, Annapolis Valley wine tours, and Gilroy Garlic Festival, generating millions of dollars in revenue. These events and others will serve as a partial template for the development and implementation of the Ontario Garlic Tour planner.

Why Garlic:

• Garlic is universally revered and connects with diverse interests and activities. This includes:
• Garlic’s prominence as the most widely used ingredient in virtually every cuisine, after water and salt.
• Garlic’s hidden culinary properties, including its use in dessert and alcoholic beverages.
• Garlic’s essential use in every culture for all kinds of medicinal purposes
• Garlic’s connection to a wide range of popular and esoteric interests including health and wellness, cultural identity, gardening and farming, climate change and space travel.
• Garlic featured in popular culture, including film, vampire lore and literature and music.
• Garlic as a bridge between cultures, contributing to shared garlic-themed experiences among people from all ethnic and religious communities.

We are seeking students with a keen interest in LLMs, RAG (Retrieval-Augmented Generation) techniques, data preprocessing, and experimentation. Skills in scripting and automation, coupled with a basic understanding of the tourism industry, will be invaluable. This project offers computer science and tourism students the chance to work on refining the application of LLMs in travel planning, enhancing personalization, and overcoming the current weaknesses of current online travel planners. Task description, provided by Fariborz Lesani, includes:

• Familiarity with Large Language Models (LLMs): Understanding of the architecture, capabilities, and limitations of large language models like GPT, BERT, or similar pre-trained models.

• RAG and Data Retrieval Techniques: Knowledge of Retrieval-Augmented Generation (RAG) techniques and experience in utilizing them to enhance text generation tasks. Understanding of methods for retrieving relevant data and incorporating it into the generation process.

• Data Preprocessing and Formatting: Proficiency in preprocessing and formatting custom data to be compatible with the existing LLM and RAG framework. This includes text cleaning, tokenization, and data augmentation if necessary.

• Experimentation and Evaluation: Ability to design experiments to evaluate the effectiveness of RAG-based approaches for generating tourism-related content. Familiarity with evaluation metrics specific to text generation tasks.

• Scripting and Automation: Experience with scripting languages like Python for automating data retrieval, preprocessing, and model evaluation tasks. Knowledge of relevant libraries and frameworks for efficient workflow automation.

• Understanding of Tourism Domain: Basic knowledge of the tourism industry and common terminology used in travel-related content. This understanding will facilitate the selection and customization of data for fine-tuning the LLM.

• Collaboration and Communication: Strong communication skills to collaborate effectively with team members and stakeholders. Ability to articulate ideas, discuss findings, and propose improvements to the existing LLM and RAG pipeline.

While deep expertise in NLP, ML, and model fine-tuning may not be as crucial in this context, having a foundational understanding of these concepts can still be beneficial for comprehending the underlying principles behind LLMs and RAG techniques. Additionally, practical experience with scripting, data manipulation, and domain-specific knowledge will be invaluable for successfully executing the tasks associated with fine-tuning an existing LLM for the tourism industry.