Inverted Index Project

I haven't spoken much about the class I've been teaching this semester. It's an intro CS course - a programming heavy intro. I decided to use Python with a transition at the end to C++. The transition is to mirror Hunter's normal first CS course that ends with a C++ intro to prepare the students for next semester's CS course which is a more intense OOP class using C++ - the language we use in our core courses.

Throughout the semester I've tried to use a variety of interesting application areas so as to try to give the students some idea of the possibilities that studying CS will open up for them.

After covering Python dictionaries and lists I thought we'd play by building an inverted Index.

The basic idea is to map a set of words back to source files. For example, given the following four one line files:

files     contents
file.01     if you prick us do we not bleed
file.02     if you tickle us do we not laugh
file.03     if you poison us do we not die and
file.04     if you wrong us shall we not revenge

You could build a data structure mapping each word back to the file(s) that contain it (partially shown here),

Word   Files containing It's
if   file.01 file.02 file.03 file.04
you   file.01 file.02 file.03 file.04
prick   file.01
us   file.01 file.02 file.03 file.04
do   file.01 file.02 file.03

You can, of course, store more information - how many times a word appears in a file, where it appears, etc.

This is a fairly easy structure to build. A dictionary where the keys are the words in the file and the values are lists of the documents containing the words.

  inverted_index = {
      'if' : ['file.01','file.02','file.03','file.04'],
      'you' : ['file.01','file.02','file.03','file.04'],
      'prick' : ['file.01'],
      'us' : ['file.01','file.02','file.03','file.04'],
      'do' : ['file.01','file.02','file.03'],
      ...
}

In addition to letting us work with dictionaries and lists, we can also review file access and even the python CSV module if we want.

We can immediately write simple queries – "what document(s) contain the word 'prick,' but things get more interesting if you write functions to perform and and or queries - "what document(s) contain the words 'prick' or 'do'" for instance.

Why are we building this (besides as a data structure and programming exercise)? I've seen a number of references to using an inverted index when building a web search engine. In fact, I think that's something you do early on in the Udacity Mooc. I just wanted to play with information retrieval.

I remembered that there was a collection of information, including last statements from executed offenders in Texas. Someone conveniently converted it into a Google Spreadsheet. The format's a little different from our simple four file example but then there's more data. It's straightforward enough to download the spreadsheet as a CSV file and then read it with a Python program that builds it into an inverted index.

Now we have some interesting data to play with.

How many offenders used words like "sorry" or "apologize?" How about references to religion? We can do all sorts of and and or queries.

We just played with this a bit but I could see all sorts of explorations. What about taking some great work of literature and turning it into an inverted index by chapter. You could query characters or certain words and see where and when they appear in the book. A new and different way of exploring literature.

So, there you have it - an interesting little project we played with this past semester. We did it in an intro Python course but I could see it as an interesting project in AP CS A using hashmaps and lists.

Comments

Comments powered by Disqus



Enter your email address:

Delivered by FeedBurner

Google Analytics Alternative