I used to be a regular at an online forum that was notable for the relatively high quality of the discourse. At least, that's how I remember things. I look back fondly at a time, five years ago, when that particular section of the internet was peopled by intelligent, erudite, witty philosophers who had informed opinions on just about every topic. Or so it seemed.
It could be just my rose-coloured recollection, but today I constantly grumble about the way it has become 'democratised'. It seems that the higher than average quality has attracted an increasing number of people who actually don't have a lot to contribute, and this leads to a lowering of the average standard.
Or maybe I'm just an old curmudgeon, constantly harking back to some mythical golden age? It's all just a matter of opinion. So, the web being all about data mining and 'measure' being the mantra of successful business, I was wondering if there was any way to analyse hundreds of thousands of posts and reach some meaningful objective conclusion. I came up with this idea after reading up on language testing and research....
The British National Corpus is a body of sample texts comprising 100,000,000 or so words. The samples come from different categories (recordings of conversations, business communications, newspaper stories, etc) and have been analysed individually and collectively to determine how often each word or phrase occurs in certain situations or on average. From this data a list can be compiled of all words in order of how commonly used they are. Presumably, if you were to run every post by a specific individual through the same software, you would be able to compare the vocabulary they use with this list and get a measure of their 'lexical range'.
Word counts are already available in php forums, so we would be able to say that Poster X writes an average of 30 words per time, and uses a limited vocabulary, while Poster Y writes 100 words and has an extremely wide lexical range. This probably wouldn't give you a meaningful measure of 'intelligence', but it would go some way towards quantifying their style of interaction.
You may find that you identify certain types of posters more clearly and can then tailor your site management in response to this. Another approach would be to analyse all posts in the order they were submitted, and plot a graph over time of the averages. Did average post lengths increase or decrease over time? Did vocabulary employed follow the same trend or did it do something else? How does this correspond to the number of posts on any given day/week?
You could also look for other key words or phrases and see how often they appeared, probably as a percentage of daily posts or total words. You might learn, for instance, that your supposedly intelligent initial group used the word 'fuck' more often than it is seen today, or vice versa. Who knows? It's an interesting idea, although I don't know how hard it would be to do.
Recommend this article...
Users' Comments (0) 
|
|
|
