Who Played Spiderman - part 3
Part 3
In the first two parts of this set of posts I wrote about the motivation and design a question answering system that can answer "who" queries like "who played Spiderman" or "who shot John Lennon?" It's not perfect. When doing the Spiderman query, chances are the desired answer will be at or near the top of the list of most frequently appearing names but so will "Peter Parker." How do you distinguish between a real answer and a fictional name? Likewise, if you ask "Who shot John Lennon?" You'll probably get Mark David (if you search for two word names) or if you're more inclusive, Mark David Chapman, but you'll also find John Lennon at or near the top of your results.
This system may not be perfect but then again, it can be written by one or a small group of students relatively new to CS in a few periods as opposed to a team of full time Google Software Engineers. I also truly like the fact that it isn't perfect. Too many student experiences involve perfect questions with predetermined perfect answers. I love the fact that they see false positives and true negatives in this. I love that they can get pretty good but not perfect results. This makes the experience more interesting, more fun, and more authentic.
This lesson can also easily be extended beyond "who" queries. "When" queries are also pretty straightforward. Things like "When was D-Day" or "When did Elvis die?" For a "when" query, you start the same way as for a "who" query. Use the search API to get a bunch of documents that mention the query terms. Then you look for dates. Dates can be even easier to search for than names. Particularly if your students know regular expressions. Dates could be things like:
- two digits / two digits / two digits
- Month name number, number
- etc.
or in regular expressions, something like:
- [0-9]{2}/[0-9]{2}/[0-9]{2}/
- (Jan|Feb|Mar…)[0-9]{2}, {0-9]{4}}
- etc.
Just as with "who" queries there are issues to deal with - searching for "when did Elvis die" might give you when Elvis was born instead but again, these are great motivators for further exploration and discussion.
So, there's the project. What level of students can do this? It depends. I first did this in my Software Devepment class. That class was offered to student who had completed APCS-A. The class covered using APIs and setting up a web application using Python and Flask so they were able to do the entire project, soup to nuts. An APCS class could also do the project without the web application just so long as they have access to the search API. In fact, an intro programming class could also have success with this project.
This was actually one of the lessons I've missed since I haven't been able to work it into my Hunter courses so I was really happy when the class discussion led me to call and audible and talk about it.
I hope some of you out there experiment with it and if you do, I'd love to hear about your tweaks, implementations and successes.