Earlier this month, I asked a question on LinkedIn Answers:
As a donor, if you had one wish for the nonprofit sector, what would it be?
After nearly two weeks, I received 18 answers. Beth Kanter asked me to do an analysis of the answers. She’s very good at this sort of thing so I didn’t want to let her down by doing a crappy job on it. This posed a neat little dilemma. What kind of analysis should I use? I decided to go the extra-geeky route (of course). I decided to do the following:
- Extract all the answers off the LinkedIn Answers page.
- Concatenate all the answers into one text stream.
- Use the Yahoo! Term Extraction API to create a list of keywords from the text stream I created by entering it into a form over at Blogoscoped. Those of you who don’t know anything about the Yahoo! Term Extraction API can think of it as a way for software to assess unstructured text and pull key terms out of it. It’s actually used by some blog plug-ins to help you auto-tag your blog entries. I gave it my LinkedIn answers and gave me a list of keywords.
- That allowed me to do a series of simple finds in Word to find the frequency of the word or phrase in the keywords list. The following is the results of that first pass over the list of words created by the Yahoo! Term Extraction API. I had to delete the key term “achieve financial” since Yahoo! mistook the text “achieve, financial” for an actual phrase. Think of this list as a kind of tagcloud for the LinkedIn answers I received.
Number of appearances in text Extracted Term 8 nonprofits 5 accountability 4 transparency 4 one wish 2 professionalism 2 charities 2 loyalty 1 pet peeve 1 business acumen 1 financial sustainability 1 glossy magazines 1 hot buttons 1 independent sector 1 valid concern 1 professional managers 1 waste resources 1 volunteerism 1 nonprofit organizations 1 endlessly 0 achieve financial
- I then dumped out things that referred to the question or only showed up once or obvious words like “nonprofits”. I also had to take a look at the keywords “professionalism” and “professional managers” and realized that it and the word “professional” were actually tied together and showed up four times instead of 2. That gave me the following list:
Number of appearances in text Extracted Term 5 accountability 4 transparency 4 professionalism
Clearly, a pattern is already being generated here. The question was asked of a mixed-crowd of nonprofit and private sector individuals. A large portion of them were asking for accountability, transparency and professionalism. What’s fun about this entire exercise is that we now have a way to extract good data from LinkedIn Answers without having to have Beth Kanter’s cyberbrain when it comes to recognizing patterns in data. And of course, it says a lot about the nonprofit sector and definitely makes me feel like I’m on the right track with my socialmarkets project. It’s also a neat demonstration of how to use Web 2.0 properties to actually derive knowledge from a mass of unstructured data.
I’d love to continue the experiment on a much larger basis and I hope there’s enough information in this posting for you to try out your own Yahoo!-fueled experiments with unstructured survey questions. If you wouldn’t mind using your LinkedIn network to ask the same question, I’d love to see if the responses change depending on who asks it and what kind of network they have. My suspicion is that it won’t simply because LinkedIn networks expand exponentially by the 3rd degree thus negating any unique aspects of a person’s network but I’d still love to confirm that suspicion.
Loading ...
