Enhancing Customer

The initial phase involved thorough research and data gathering. We collected official menus from the client's website and also compiled a comprehensive database of alcoholic ingredients, including common spirits and less alcohol-containing components like certain extracts and flavorings. This database served as a crucial reference point throughout the annotation process.

We appointed a subject matter expert in the food and beverage industry to lead the project. The SME guided the team in research and verification methods, especially those with unfamiliar ingredients to identify whether the item contained alcohol or not. Our SME crafted detailed annotation guidelines, outlining clear criteria for classification, including:

• Explicit alcoholic beverages (e.g., cocktails, wine, beer)
• Food items prepared with alcohol (e.g., Tiramisu, Vodka penne, etc.)
• Items with trace amounts of alcohol (e.g., some desserts, sauces)
• Non-alcoholic drinks (e.g., mocktails)

Our team used text classification as the primary annotation technique for categorizing each menu item into alcoholic or non-alcoholic categories. MS Excel was used as the annotation tool to facilitate efficient data entry and categorization, with clear labeling for easy review and updates. The classification process involved:

• Identifying critical ingredients in the item description
• Determining the presence of alcohol based on ingredients and preparation methods
• Assigning appropriate labels: "alcoholic," "non-alcoholic," or "trace alcohol"
• Flagging items for secondary review for any uncertainties

The project involved a rigorous quality assurance phase. Each item initially labeled as unclear was reviewed by another annotator or subject matter expert, which helped us catch any discrepancies or potential errors. Before project delivery, the final dataset was reviewed by the QA resource, ensuring that all items were correctly categorized.