Categorization
Training Data
Overview
- Accuracy, Precision, Recall Calculation Fix
- Training Data Format
- Knowledge Graph Use
Improving Results
On the same data set as last week.
Calculations on Training Data
New input format
Calculations on Training Data
Frequency counter tool automatically gives notated data now
Calculations on Training Data
Gathering Other Seed Data
Decided on other categories for seed data
- Travel (Vacation Cities)
- Sports (Teams)
- 3C Products (Brands)
- Politics (US Politicians)
- Music (Song Titles)
- Clothing / Adornments (Brands)
- Arts & Culture (Video Games)
- TV & Entertainment (TV Shows)
Using Knowledge Graph
- To be done after gathering the data for verification
- Can be checked whether a query is a restaurant or not.
- Offers many types - Standardized "Schema"
Query -->
Result
|
V
Focus on the @type variety
Consider Restaurants
- We consider Restaurants to be the blanket terms which will encompass the lists.
- In reality, Schema (schema.org) considers it as a specific subsection.
Multiple Inheritance
Each entity derives its properties from different ancestors, and the properties of a particular category are the net sum of its ancestors + any of its own
Inherited Properties
Another Example
