Enter a search term such as “mobile analytics” or browse our content using the filters above.
That’s not only a poor Scrabble score but we also couldn’t find any results matching
Check your spelling or try broadening your search.
Sorry about this, there is a problem with our search at the moment.
Please try again later.
TwitterJobSearch is a natural language search engine that reviews all of the tweets about jobs to identify which ones are offers of employment, and lists them on its site.
It was developed by Workdigital Ltd, a vertical search company that also founded both workhound.co.uk. I've been talking to co-founder Bill Fischer about TwitterJobSearch...
How does TwitterJobSearch work?
We are a search technology company, and our speciality is natural language search. A year ago, we thought it would be interesting to run millions of tweets through our search engine to see if it would work for 140 characters and if we could distinguish between those which were about actual jobs, or which were just people talking about about jobs.
It's relatively easy to spider a full webpage with plenty of content, but more of a challenge when applied to a tweet.
With some fine tuning, we were able to distinguish the genuine job listings from the irrelevant tweets, so a tweet relating to a news story about job losses in the construction industry for instance, or someone bitching about their own job, could be distinguished from a tweet that is linked to a genuine job listing.
This is something a keyword search can't achieve, but natural language search can be very effective for. If something appears to be a job offer, we follow the links from that tweet and spider the page, grabbing the relevant data associated with it.
So, we can see a mention of a vacant position for a sous chef, then we can follow the links from the tweet to the restaurant's website, and list details of the job, the sort of details that wouldn't be featured in the original tweet.
Are many people listing jobs on Twitter?
People like to multi-post positions they are seeking to fill, and also, since it is very easy to post a tweet, they will often mention jobs on Twitter first, so we can sometimes get stuff that hasn't been advertised elsewhere. To give you an idea of scale, 240,000 unique job vacancies were tweeted on the last 30 days.
It is mainly US-focused, but we also show jobs and get some traffic from elsewhere, especially the UK, Canada, Australia and Germany. At the moment, we are sending about 1.4m clicks through to jobs every month, to other job listings boards, LinkedIn pages, company websites etc.
Do people advertise jobs with you directly?
We pick up the majority from searching Twitter, but companies can list their jobs directly with us. We are working with clients like Kellogs, Astro Zenica and others. Some will pay to post a single job, while we create branded job feeds and Twitter accounts for brands like Adidas.
How have you funded / monetised the service?
We took some Series A funding in August of last year. Our revenue comes from advertising; it is free to have your job indexed but we charge for featured listings on a CPC basis.
How much traffic are you getting?
User numbers have been growing, we've had 150,000 unique visitors over the last 30 days, as well as more than 100,000 followers on Twitter, which is our main source of traffic.
How does it work?
It's very easy to build a keyword search engine for Twitter, but for us in practice, the hard work of building up a library of words and phrases associates with job vacancies to scan for.
The system pools and analyses each of the phrases we find that may be job listings. If the system finds false positives, then we tweak it to improve the accuracy, and the vast majority of listing on our site are accurate.
I think people who use search engines are unforgiving, and perhaps don't realise the work that goes on to produce accurate results, and can be frustrated if they see any errors in the results.
What's next? Will you apply the same principles to other sectors?
Jobs is one of the biggest listings categories, but also one of the most difficult to search and index. We intend to expand the idea into other areas. We are currently doing tests for automotive and real estate listings. Automotive is easier to search for as there are fewer possible attributes associated with cars for sale, price, number of doors, mileage etc.