A Presentation on Self-Organizing Learning of Semantics

In this presentation (PPT and PDF) I summarize the two self-organizing approaches to semantic learning covered on Tuesday and Wednesday. In the presentation, I propose the "problem of learning language," and discuss the constraints that have been used to make the language problem tractable. In particular, I differentiate between contraints (or assumptions) that are theoretically grounded, and those that are merely a matter of implementation.

In the process, I explain singular value decomposition by way of an analogy to eigenfaces, as well as analyze the similarities of hebbian learning mechanisms with both the math underlying latent semantic analysis as well as the math in the "grounded bootstrapping" experiments of Steels & Kaplan.


Anonymous Anonymous said...

Nice post! I'm enjoying a lot your posts! Btw, if you want to have a look to another perspective of cognitive linguistics you should check the Neural Theory of Language (http://www.icsi.berkeley.edu/NTL/).

Finally I got some spare time to update my site and link your great blog. Keep the good work man.

9/07/2006 02:18:00 PM  
Blogger Chris Chatham said...

Great, thanks! I will definitely check out the NTL link you provide. Looks interesting.

9/07/2006 02:42:00 PM  
Anonymous Anonymous said...

The "PPT" and "PDF" links both link to the powerpoint version.

9/07/2006 04:38:00 PM  
Blogger Roti said...

That was very nice =)

I got an idea on something nearing the end though. i hope you don't mind me saying it. it's something that i struck me after reading it.

The language forms start to generalise after the language games get more complexed. Perhaps we can say that, language just like other things that human beings use, are interrelated? meaning that they use the same logic?

Let's say logic is uniform. Take the philosophies for example. In my opinion, the philosophies that arise are all from two processes but came from one basic logic. Proposedly the first philosopher came up with his own philosphy but it was according to his bias. But then another person with sufficient intellect managed to grasp that "basic logic" in his words and thus use the "basic logic" according to his own bias to create another philosophy. The process involved in making his own bias, from the first philosopher's idea would be protagonist or antagonist. Though these two are used at the same time, it's the ratio of the use that counts. Example: he has similar ideas here but other places he may have different ideas.

Perhaps language came up the same way? That language evolved from some kind of "basic structure" and it evolved into a more complexed, more specific orientated community-serving bias? As you mentioned in the PDF, although there might not be contact among the two users of different language, though the physical items related to the language making are still largely similar. Thus maybe those similar objects and i think features may also become similar in the end product because everything strives into specialisation (thinkers think deeper and make up more names for each individual trait of an item that they have found), are the "basic structure" that we use for language learning. Thus with this "basic structure", one can use his or her own opinion to form a new language.

Though, perhaps babies have a great speed of learning because they have yet to form any opinions, thus no biases. They just absorb everything, starting from physical items (that are concrete), then slowly associate them to more subjective items.

That's my opinion on it. Do you think it feasable?

9/07/2006 06:45:00 PM  
Blogger Chris Chatham said...

Thanks Anon - I fixed the PDF link.

Moonshadow: I like the direction you seem to be going, and I think you're right about the evolution of language on a developmental time-scale from concrete to abstract. Thanks for sharing!

9/07/2006 07:35:00 PM  

Post a Comment

<< Home