[Freddy]

I'm messing around with some bot stuff and would like to get my bot to be able to define a word.

I'm using VB, but can't find any examples of how to use Wordnet.  I can't even download it, as the links are dead at Princeton...

http://wordnet.princeton.edu/wordnet/download/current-version/#win

I'm looking for something like COM or a DLL to search for words.  Not getting very far though.

Ideas anyone ?

[Bragi]

http://wordnet.princeton.edu/wordnet/download/current-version/#win
this is where you can download the database files and app for windows. Best to download version 3 and replace the db files of the windows app so that it works with the latest version.


http://wordnet.princeton.edu/wordnet/related-projects/#.NET
here you can download .net and com libraries that provide the data in your own app.

[Freddy]

Thanks. Yes I found those pages but yesterday the links did not work, today they are

[Bragi]

strange

[Freddy]

Got it going finally. 

[spydaz]

To really use the word net data in my app I used ultra hal as a plugin for my vb app.
Then I extracted the nouns and verbs from the text .dat files
Then I used the list to extract the hypernyms and other nyms to a database table in Ms access ...
Then I wrote my own program to used the database that was created .... Removing the wordnet and ultra Hal components ...,
The concept net open mind common sense module is easier to do .... And the same type of table structure can be created as the word net but the common sense data is not clean many many mistakes ... But the conceptual model is fine and you can extend the predicates to many more than they have used ...,,

The lists of nouns can be used to create learning patterns , which can be used to create predicates which intern will detect new nouns ....

Mad little. Circle ... But with a list of nouns you can go a long long way ....

[spydaz]

Understanding data laws , you may find that after data has been through transformations and restructuring / extract / transform : load then the data is really new data .... Therefore it's yours to do with what you will ...
As long as it's not personal details ....
Often newspapers or data providers will resell data after it has been segmented or transformed to businesses looking for specific information ... Without citation ... To the original sources ... A fine line in data protection ...

Yet ; Facebook sell their traffic data etc .... And then often you will find a site selling Facebook and Twitter data combined neatly into new formats ....

[spydaz]

All transformers

[ivan.moony]

Not sure if this fits, but here is a short list of ontologies with definitions of different notions:

http://dbpedia.org/About
http://babelnet.org/
http://www.freebase.com/
http://people.mpi-inf.mpg.de/~gdemelo/yagosumo/

Some of them are of >500MB size, so I'm not sure about their usability.

[spydaz]

I've also used a couple of these resources and their usability is exactly the problem ... Plus the data once you get in is also badly formatted ...
To open you can use a cloud computing service (processing takes place in the cloud) that's when the formatting hits you in the face ... I think they also know .... Or they would have subsets much smaller to download and try ...

Data is no good if you have to write a program to extract the information again ...

[spydaz]

Especially dbpedia

[ivan.moony]

dbpedia can alternatively be downloaded in json or csv formats also. Processing them should be simple thing to do, as those are widely accepted textual formats.

[ivan.moony]

I'm having a few YAGO-SUMO files in tptp format right here in text editor. The most interesting are:

• "wordnet-classes.tptp" (enumeration of classes and their properties as in object oriented programming language)
• "means.tptp" (mappings from text representations to symbols used in other files)
• "type.tptp" (says which specific objects belong to which classes)

Other files are just a bunch of very narrow selection of properties of objects with their values, I think not very usable for general knowledge base, but who would know it...

To read tptp files you need a parser for tptp. If you don't have one (I didn't find any at a quick look), you can build your own parser for tptp with a variety of *parser generators* for different languages (including c#), but you have to provide tptp grammar syntax that can be found on TPTP home page.

Once you can parse tptp files of interest, I guess the next step is to construct and upload usable records to mySQL database on server because one cannot expect users to download at least 3GB of data each time she/he boots up an interface engine of a smartbot. When the data is online, a nice solution would be a partial download of a bunch of records related to current theme of which user want to talk about (some crowdsourced statistical analysis of record importance would apply here).

I am satisfied with YAGO-SUMO. It has what I expect and can be used as a starting knowledge representation for a smartbot of next generation.

P.S.
As of class description, there is class-subclass enumeration, but you have to clue up by yourself which properties belong to which class. There is just enumeration of possible values that belong to specific properties, not tied to any class...

[infurl]

I've been working with different versions of Yago over the years and can handle Yago2 pretty easily with the software that I've written for it. However Yago3 is available now and it is 91GB uncompressed. Have any of you tried working with Yago3 yet?

[Freddy]

I had not heard of Yago, it looks very useful. Currently do you have to download everything or is anyone running it as an online service ?