How to create the best autosuggest using machine learning

Question

I am building an autosuggest functionality.

I have a set of data of about 23k entries, in this data is words, and combinations of multiple words, i.e. "London College of Arts" and "London".

I want to write an autocomplete search function that takes into account, the letters that have already been typed in, say "LO" but also some other data points I have:

• location,

• day of year,

• number of previous searches,

• scored items - say London College of arts has 20,000 students which makes it rank higher than "London school of economics" which only has 5,000 etc.

I am guessing from preliminary research that I need to start learning some kind of Machine learning technique, but exactly which would be best applied to this problem is still hard for me to see, SVN's, NN etc. I am finding it hard to find any practical examples of these algorithms being used on problems like this, any help or pointers to good resources is much appreciated.

Answer 1

Look up Apache Solr (it might be overkill for your needs for auto-suggest only), but offers a pretty decent auto-suggest OOTB, without you having to create from scratch. Since it offers a RESTful interface you can integrate into most any language of choice.

You can also weight by contextual variables such as location, date etc. Solr is a powerful tool that does a lot more than auto-suggest, but might serve your needs.