Natural Language Processing and Constituency Forest

Well, if we could put Visual information (Visual Processing and Vector Calculus) into a hierarchical arrangement,
It should work the same to language.

In fact, in my personal opinion, the first intelligence machine will probably be a chatbot.

Again, here I only intend to discuss the arrangement of language.
The recognition is still unreachable :woozy_face:

A constituency tree is something looks like this.
I will recommend people google a real one :joy:

Then we add a time sequence into it.

This is a set of a constituency tree.

a constituency forest will require to merge same parts.

Then we get something nice and hierarchical like the one in Visual representation. :smiley:

Language has a hierarchical structure because reality has a hierarchical structure. Things are made up of other things, which are made of other things. Objects are hierarchical. You don’t necessarily need a hierarchical model to process hierarchical data, you just need recursion.

@Bear I’ll keep pushing you to read our latest papers. If nothing else, watch the last three episodes of HTM School. You are talking about the same type of things we talk about here on this forum, but we think about them in very different ways.


Yeah If there is a no such necessity to associate it with visual information :wink:

You are traveling down another very well traveled path.
If you wish to avoid writing this from scratch you can use Parsey McParseface today.

I would strongly advise you to do as Matt suggests and learn what is out there before you waste a lot of your obvious creative energy reinventing what already exists; we would love to see that creative energy used to come up with new things rather than reinventing old things.

Then only way to do this is to put in the effort and learn what has already been done. As you can see - we are more than happy to point you at the resources you need for this but you will have to spend the time to read through them.


Thanks for being so patient. :blush:

I am still at a stage don’t even understand what is new or what is old.
I believe that people already have all the materials,
and we need to integrate them to create a more meaningful system.

To understand what is “apple.”
Not only know its location in syntax but to touch it, to see the shape, and to taste the flavor.

To define meaning via experience.
To disambiguate through sensational association.

1 Like

Again, it should come as no surprise that many very talented people have determined that this level of understanding is important and have applied a great effort in addressing these questions.

I suggest that at this level of your learning the nature of your questions should be in the form: I am interested in (subject xyz) - can you point me in some introductory material and an overview of the state of the art. This will be the path to learning what is out there now and what questions researchers in the field are exploring.

I don’t know if you are behind the great firewall but I find that google is a very good tool to start with myself. It is amazing how fast you can “get the lay of the land” on a new topic that way.

If you don’t mind me asking; where is your home?


I am from Wuhan which is located at the center of China.
But I am in America now as an international student :smiley:

Yeah, and I have troubles of reading English articles.
They are really hard :joy:

1 Like

I have spent about 6 months in Mandarin speaking places since 1992 and can read and speak some Mandarin - but my tones are terrible. I totally get how hard it is to absorb technical material in a different language.

We can focus on visual material and video as these may be easier for you now.

Matt’s videos should be a great place to start.

Pro tip: when you do a google search, click on show as images or video. A picture may be worth more than a thousand words.