A way to create knowledge is to set up filters, channels which makes raw information to become usable, part of a larger context. In WordPress channeling and filtering is done by categories and tags. Lately I’ve been sensing difficulties in extracting knowledge from this blog, and that is due to inconsistent tagging and categorizing. Let’s see now if AI can help us with this task.
I have a project running since 2006 where I’m trying to capture knowledge via tags. You can check it yourself if it’s working or not.
For me it’s unsure, in the same way like the tagging in this blog. The problem is that when you add a new post you can’t immediately figure out the best tags to use.
When you write an article your mind is focused on getting together your thoughts into a whole ― while tagging is exactly the opposite, to oversee your entire blog / knowledge-base and pick those few tags which are relevant on the long term.
My mind is not able to switch from compressing things ― from writing ― into stretching to comprehend the whole knowledge I’ve created during the years ― to tagging.
Putting it simple, when you write a post you can’t remember the whole list of tags to choose the best ones. When you re-read your posts you can see that some tags are not working and you can change them. But you can’t re-read all those hundreds of posts at once to align all the tags.
If you re-read and re-tag your posts in a consistent way during a few sessions spanning a few days then you can achieve some consistency. But if you add a post later, after a few weeks, or months ― again the tagging might broke, you can face the same problem like at the beginning.
That’s why I feel creating knowledge on long term with manual tagging is not perfect.
In the same time when I’ve started Gust I’ve started to collect (electronic) music. I mean listening music on Last.fm, Deezer, Apple Music etc. and organizing them into playlists. Until last year I’ve had around 30 playlists, some of them over 1,000 songs.
It was the same exercise in filtering now in music instead of text.
I’ve failed so hard organizing music that I had to quit after ten years. Now I’m listening only mixes on Soundcloud ― short playlists with 10-20 songs made by professional DJs.
The advantage is that I’m getting the same joy and flow with 1% of the effort. Before Soundcloud I’ve spent 2-4 hours to listening to music to add 3-4 songs to a playlist. Now with a like I can save an entire mix.
The drawback is that those mixes are not mine. Sometimes they are much better than my old Deezer playlists but sometimes they are really far from them. And, they are not unique, they are not special ― anyone can have them discovered.
If a human can’t create knowledge on long term ― because it’s mood might be different when touching the knowledge-base from time to time ― perhaps and AI can do it because the AI is cold, is a machine, and displays the same mood all the time ― being objective anytime touching the base.
I’m planning to find a tool to extract keywords from a text ― call it text mining, natural language processing, be it a WordPress plugin, or an external service on Algorithmia.
To see if the tool works I’ll compare it’s results with my own tagging. Let’s see if AI is capable to do a real job here.