g
ame, g
a
me. repea
t
.

image

When I started work on DishFu.com (one of my code playgrounds…its something like Yelp) a couple years ago I wanted to support natural language search so that, instead of clicking lots of checkboxes to search for the best burrito in my zip code, I could just ask “what is the best burrito near me” and get the answer. At the time (and even now) I didn’t find a ruby library that made this easy to do, so I thought I’d take a stab at it.

Natural is the result of the aforementioned stab and its available on github. Checkout Fantastical and Siri for inspiration and two good examples of why natural language search is neat and how it enhances usability. Natural works by parsing the input text, looking for every possible match in its vocabulary, scoring all the possible combinations of matches, and choosing the highest scoring combination. One can achieve a similar effect (and faster) by enumerating a bunch of regular expressions and trying to match against the most complicated regular expression first and working your way down until you find a match, but I wanted a library that could also glean partial meanings and choose between different potential matches in a more flexible way. Suggestions and pull requests welcome.