My team and I have been working on a project that involves classifying different strings in a text, or more specifically, Wikipedia Infoboxes. For example, Barack Obama would be classified as a Personal Name, California as a Toponym, and 1914-1916 would be classified as a Time Interval.
However, Wikipedia data is not only vast, but it's also very messy and hard to work with, making our task very difficult.
The issue that we have come across is being able to classify compound strings or lists that contain more than one type of information. We don't want to classify each individual part of the string, but the string as a whole. For example, we have the following strings that we don't know how to classify as a whole:
- 2006, 2007, 2012
- Wouldn't classify this as date, date, date.
- Possibly classify as a list since this has the same type of information in a row.
- Rangana Herath (23)
- We wouldn't classify this as Personal Name, Age.
- Turkish (3.55%) (2002)
- Kazakhs reduced from 60% to 38% of the republic's population
We want to know if there are terms that we could use that would help us classify these compound strings that contain more than one type of information. If so, what are these terms or names that we could use to classify these compound strings? How should we classify the examples provided above?
