Message Boards Message Boards

Semantic Networks

Posted 9 years ago

What I want to do is construct the semantic network of a book (pdf format) and then a semantic network from various scientific articles (pdf format) and then compare the two to see if the book is scientifically correct and what new concepts does it add to the existing scientific research. Can anyone help me with the code. Is it even possible to do something like this in Mathematica? I am new to Mathematica and I am not really a good programmer. (Just some simple C++ programming experience)

6 Replies

Assuming the below image is correct then question/computation would be the book/articles

I apologize. I cannot understand what that means.

The "symbolic representation" part is very broad. It's not just a semantic network. Some of tools used in Wolfram|Alpha for symbolic representation are available in Mathematica. See the documentation on Entity for more information. Some tools used in Wolfram|Alpha are not available. Much of the algorithmic linguistic understanding is very domain specific or manually curated in some degree.

Can't this be done once for the book and once for the scientific articles and then check if there are matches, like genes in a DNA strand?

You mean like sequence alignment? It would be graph matching, which is harder, but I guess it's the same idea. And yes, that sounds really cool, but it's just not feasible. Who right now has semantic representations of documents that are so good it would make sense to this? No one.

I will stop beating around the bush. This is not how NLP and semantic computing works. Semantic networks aren't as good as you imagine them or as easy to make. If you're interested in this, work at becoming a better programmer and take some NLP/machine learning courses. These will you a good idea of what kinds of things are realisitic and possible.

POSTED BY: Sean Clarke

Assuming the below image is correct then question/computation would be the book/articles and symbolic representation would be the semantic network. Can't this be done once for the book and once for the scientific articles and then check if there are matches, like genes in a DNA strand? Does the linguistic understanding must be made in engines like Apache UIMA?

enter image description here

For discussion and sample code in areas overlapping with #2 in your list, see Bill Turkel's new open source textbook on using Mathematica for digital humanities research: http://williamjturkel.net/digital-research-methods-with-mathematica/

POSTED BY: Arno Bosse

Some notes. Of course you understand that any program that could do this flawlessly might as well be capable of passing a turing test as well...

  1. I'm not sure what this means. Do you want the program to "read" the articles? If so, we need to be more clear on what this means. Are you just looking to import the articles into Mathematica? That can mostly handled with the Import command

  2. There are a lot of functions do textual analysis like WordCounts (http://reference.wolfram.com/language/ref/WordCounts.html) that you might use. You can filter out common words using (DeleteStopWords) http://reference.wolfram.com/language/ref/DeleteStopwords.html. This would give a simple but effective way of finding the keywords of an article.

  3. This is also very vauge. What would it mean to "read" the equation? What does it mean for a program to "read" an equation?

  4. Semantic Network is a very vauge term. Are you trying to create something like a RDF database where there are entities linked by properties? Generating anything like this from text is huge problem actively being researched.

I would recommend becoming more familiar with this subject area. Learn how RDF, sparql and related technologies work. Read a book on NLP so you can learn what the practical limitations of the field are.

POSTED BY: Sean Clarke

This is probably too big of a question to address on a forum.

What do you mean by a "semantic network of a book"? That is what kind of semantic network did you have in mind? What features of the book are captured in the semantic network?

This doesn't sound like a project for a beginner. What is your experience in semantic computing? If you are new to this, you should just first try some simpler project to gain an understanding of the field and its capabilities.

POSTED BY: Sean Clarke
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract