We've been going through tens of thousands of search results and testing several relevancy algorithms and we're not getting any closer. As thesocialCV.com is building more than 600 million profiles, it's kind of important. So we're going to let the customer decide.
Let us know, what do you prefer?
Advanced Boolean Queries
Semantic Skills Extraction and Matching
Social Authority Rankings
Job Title Match
Click-Thru Data from Live Searches
User Generated Ratings
Career Path Match
or other approaches, we're happy to make them all available but we need your help.
Here's an example of one relevancy system we've developed, TalentRank.
We know there must be other, maybe better, approaches. The key is to turn that gut instinct a recruiter has when they scan a CV (their blink response) and turn it into an algorithm that can automatically find the best candidates. What's more important: Job Title or Skills relevancy? The quality of the company or the quality of the education? Professional engagement or are they an engaging professional? There are hundreds of dimensions that can be used to evaluate a CV.
We need to find a way to turn the analog gut feel to binary.
William - interesting proposal you have there in regards to what your company offers. You've not mentioned anything about pricing... what do you have in mind? I'm assuming it's not going to be free? Now if it were - you'd be on to something massive... if not... well, the profiles you are putting together can all be found on the web anyway right? You don't have their direct contact numbers, or email addresses - you simply find their twitter account or company contact details right?
I may have some bits wrong - but please take this as nothing more than constructive critisism! I just want to try and understand you service a bit better so we can make a call on whether it's something we as a group (the business I run), would be interested in.
Luke: We think expert sites with user-generated rankings will be critical in helping us to rank candidates - particularly those with common skills. We are currently using API data from several of the expert sites and have some of those rankings as part of our relevancy algorithm. Identifying the right proxies and weighting is really a challenge and we're constantly user testing different algorithms - here's a smart take from foursquare at some approaches to collaborative filtering and data challenges that we are grappling with: http://engineering.foursquare.com/2011/03/22/building-a-recommendat... - Sounds like you've given this matter serious thought and it would be great to have a quick call to talk proxy data and algorithms. bill at thesocialcv.com
Sean: You're right that it won't be free and our challenge is to deliver a great ROI. If there's a trick, for example, to find 67,000 software developers with PHP and .net experience, have boolean and meta-data to sort/filter them, and half a dozen ways to connect with them in less than a minute, then.... Be happy to set up a web demo.
Henning: Agreed that skills matching is very important. Our challenge is that we're building a system with 600 million profiles. We perform the semantic skills extraction but routinely have thousands of search results that match the skills requirement (nobody likes going through hundreds of pages of search results) Simplicity is important but I think we need to build even more sophisticated tools to allow recruitment professionals find the best candidates quickly. That being said, we allow for complex boolean queries and have faceted search to also help whittle the search results.