Dirk Mischke
April, 16th 2015
This is what happens when you let a “Weird” nerd develope a “Word” prediction algorithm: Wordo [wɜrdoʊ]
In the developement process I was contemplating how to make the new prediction tool different, and (whats even more important!) better known to my potential customers.
Who wants to read about a new word prediction algorithm, its code, its math? Nobody!
People want to use the tool and experience the application of the algorithm!
That's why I made “Wordo”.
The algorithm was developed using three large text files, one containing US-news, one with US-blog posts and one with US-twitter posts.
After cleaning the raw data and removing punctuation, whitespaces and profain words, I created n-gram representations of the combined three files, showing combinations of 2, 3 and 4 words.
With these n-grams I was able to write a prediction function, which shows the most likely next word, given the input phrase.
The game can be played with as many people as desired. You establish an order, clockwise or counterclockwise and which person is the first to type in a phrase with one word missing. The person who picks the phrase can't get points in this round. Everybody else has to write down (on paper, phone or tablet) the word they guess is the missing word. After everybody picked a word, the 'predict' button on the screen can be clicked to show the top 4 results. Points are awarded for each word that is shown as a result, 4 points for the top word, 3 points for the second, 2 points for the third and 1 point for the fourth. You record the points and the next person starts the next round with choosing a new phrase. The game can be played for a predetermined number of rounds. Of course, the rules can be changed anytime!
The word prediction algorithm is a work in progress and not perfect in its current state.
Presenting the tool as a game has several benefits:
With distribution of the game for free through social media outlets, we can gain quickly a large number of users.
To bring the project to the next level I suggest
Try out for yourself and see if you too will say “That's Wordo!” https://dirkm2577.shinyapps.io/Wordo-DataScienceCapstone/