This isnt really a full paper so much as published notes, but it might have some interest.

http://ai.r-i-software.com/PAPERS/10-27-2012_1.htm

VLG

> http://en.wikipedia.org/wiki/Category:Tasks_of_Natural_language_processing

Wikipedia has a Category for “Tasks of Natural Language Processing”.

> http://en.wikipedia.org/wiki/Category:Word-sense_disambiguation

“Tasks of Natural Language Processing” has a subcategory for “Word-sense disambiguation”.

> http://www.meta-guide.com/home/bibliography/google-scholar/word-sense-disambiguation-dialog-systems-2011

I’ve got a Meta Guide webpage for “Word-sense Disambiguation & Dialog Systems 2011”, a quick and dirty literature review for 2011.

> http://www.meta-guide.com/home/bibliography/google-scholar/sentence-boundary-disambiguation-dialog-systems

I’ve also got a webpage for “Sentence Boundary Disambiguation & Dialog Systems”.

> http://www.meta-guide.com/home/about/best-of-the-best-videos/best-word-sense-disambiguation-videos

Okay, I just added a new quick and dirty video page for “Best Word-sense Disambiguation Videos” (8x), in which WordNet seems to factor prominently.

http://code.google.com/p/mindforth/wiki/DisAmbiguation

is the Mentifex AI take on disambiguation in AI Minds.

http://www.scn.org/~mentifex/dkpj.html

is the latest work on Wotan Supercomputer AI.

Cool stuff ! My little ‘pseudo’ paper was dealing specifically with the idea of topical disambiguation, but certainly the concept of (disambiguity) has many overlapping elements. I can see where the same technique could be employed as a failsafe when and if word sense algorithms have failed.

For instance If your algorithm has failed to produce a definitive selection, you could fall back on “Did you mean [top possibility]? or [second top possibility]?” Or “I take it you meant [best guess]”

With regards to how RICH is resolving topical ambiguity, it certainly could be regarded as a “trick” rather than real deep parsing, but as such it is something that I have myself employed in conversation, either when Ive been caught not paying attention or as a teaching device, so it feels fairly natural in most cases although certainly not always. (we use Steve Jobs quite a bit in testing LOL)

User: Hows Steve?
RICH: Steve….who?
User: Jobs
RICH: What about ...Jobs?

is a little stiff, but eventually it produces the unambiguous

User: How is Steve Jobs?
RICH: Not good hes dead.

Which could be a step up from the ambiguous

User: How is [anyone]
Bot: Fine as far as I know

Even though my AIML programming skills are weak to non-existent, I think that the technique could be adopted by AIML or other AI language users if anyone finds it useful

VLG

Vincent Gilbert - Nov 27, 2012:

User: How is Steve Jobs?
RICH: Not good hes dead.

Which one of the definition of “hes” applies :

Web definitions

  Hamlet Evaluation System. An evaluation system devised and run by Americans in Saigon which required monthly computerized reports from all the DSAs in the country.

www2.iath.virginia.edu/sixties/HTML_docs/Resources/Glossary/…

  Refers to the Health and Exercise Science department and HES-10, a general education requirement for all students.

http://www.enotes.com/furman-university-guide/furman-slang

  health examination survey; hematoxylin-eosin stain; human embryonic skin; human embryonic spleen; hydroxyethyl starch; hypereosinophilic syndrome; hyperprostaglandin E syndrome

http://www.health.am/acronyms/h3/

  (German) the note ‘B flat’ (more usually called ‘B’)

http://www.sheetmusicdigital.com/musicdictionaryletterH.asp

  (Acronym) HCO Executive Secretary

internationalfreezone.net/definitions.shtml

  Hurricane Evacuation Study.

http://www.femalaw.com/glossary.php

  a spouted vessel, water jar.

what-when-how.com/archaeology-of-ancient-egypt/glossary-arc…

  a common abbreviation for human embryonic stem cell (q.v.); see also hESC

proteomics.awardspace.info/reference/MSCglossary.html

OR , do you mean “hes” == “he’s” (‘he is’) ? If so, perhaps update your system to put that : apostrophe in there

And here you all thought that ~I~ was the “grammar gestapo”.

Hi, Victor! Nice to see you drop in.

BTW, I would have also mentioned a slight lack of proper punctuation, in the form of at least a comma (but preferably a semi-colon), but that’s just me.

Great points! It truly highlights some of the problems associated with the natural language processing.

In the conversation above, “hes” is irrelevant as it is not part of the interrogative but part of the reply. As part of the generalized Turing [requirements]*, RICH replies very often leave off punctuation or skips capitalization. https://sitn.hms.harvard.edu/sitnflash_wp/2012/07/ai/

In topical disambiguation, really what we are trying to do is get from

Richard [ambiguous]
to
Richard Nixon [unambiguous]
Richard III [reduced ambiguity]

in those cases where attempts at deep parsing have failed to produce or track a specific topic.

Grammatical disambiguation [word sense, etc. ] is an entirely different headache! However, my opinion is that reliance on the users correct use of grammatical modifiers would ultimately prove futile and it fails entirely when moving into “speech to text” systems . (RICH is based on the assumption that ultimately input will be audio\visual) Simply put there is no phoneme for apostrophe. (both would be rendered as ‘heez’ I believe) 

We can look at referential topical disambiguation by looking at your list of possible definitions.  When I as a human look at RICH’s reply to my question, would I know from the conversation that “hes” (even with the missing apostrophe) refers to the topic [Steve Jobs] or would I have to decide whether or not the reply referred to German Sheet music!  In fact even If the reply translated to text as “heez ded”, I would have known that (1) “heez” was actually [he is] , that “ded” was actually [deceased] and (2) that it referred to the topic [Steve Jobs] . Whats even more impressive [about human beings] is that if we changed the topic, then came back to it, I would still be able to track it.  Humans have a miraculous ability to discern and track topics.

Great stuff!

VLG

* In this case [requirements] is taken to mean the general thoughts on the subject rather than an actual specification.