|This article is of interest to the following WikiProjects:|
'We shall deal with the deep approaches first'? Is this a transcription of a university professor's lecture, or what. Aardark 17:04, 11 June 2006 (UTC)
Homographs, vs. other WSD tasks
The "bass" example is a homograph, while sense disambiguation often deals with finer distinctions (like the people sitting at a table verses the actual table).
A song could be about fish, the first example isn't explicit enough to claim all people recognize it as the musical type of bass.
It's not that hard to disambiguate "The dog barks at the tree", if you are using a tagger in the disambiguation process. 'bark' is a verb and as such can only be of the sense 'dog noise'. The examples given in the article don't adequately describe the challenges of WSD. -- 126.96.36.199 (talk) 04:41, 10 March 2008 (UTC)
There are other algos, and some of the most interesting and seemingly most successful, have been proposed by Radu Mihalcea. Article should mention, if not review these. I'm also thinking that the current algo description is over-long, wandering off into excess detail for what should be a more general article? linas (talk) 23:38, 9 May 2008 (UTC) 1
Disambiguation ("remove uncertainty of meaning from (an ambiguous sentence, phrase, or other linguistic unit)." in Oxford American Dictionary) is a real word. DEddy (talk) 20:07, 14 September 2008 (UTC)
- It could still be a neologism. I wonder if it existed before W'pedia. It is pretty crazy. What on earth does it mean? Is there an *ambiguation? May one *ambiguate? Sure, there's ambiguity but what is this? Above we have 'It's not that hard to disambiguate' but does anyone ever say "disambiguate" out loud? Have you ever heard anyone say it with a straight face?
- If it were a neologism, dictionaries would say so. English is highly irregular. For example, someone can be 'inept'. Can they be 'ept'? Does the illness 'distemper' imply a medical meaning for 'temper'? 'Disadvantaged' is used far more often than 'advantaged' (although that may be because of the disparity between the rich and poor). I have heard others say 'disambiguation' and 'disambiguate' many times, particularly in computer science classes. I myself say 'disambiguate' whenever it is the best word to use; I believe my face is always straight when I say it. Our use of words depends on many factors, such as the extent of our education (including whether we paid attention in class and did our homework or not), the speech patterns of the people around us during childhood as well as adulthood (adultery?). David Spector 01:01, 23 March 2010 (UTC)
Relation to polysemy
Deleted the reference to polysemy 6/16 because technically speaking, the term refers only to groups of meanings that are distinct but related in some relevant way (cf the Wiki Polysemy entry). Thus 'mole' the rodent and 'mole' the scientific unit are not classified as polysemous but as non-polysemous homonyms. Yet the two are grist for the disambiguation mill like any other potential ambiguities at the word level. Experts please verify, and please excuse gaffes (this is my very first Wiki edit). Astigmatist (talk) 21:41, 16 June 2012 (UTC)
Recent revert of section blanking
When there is no discussion about removing sections of the article, when there are no edit summaries over the numerous changes by a single editor, there I must quickly revert (at least temporarily) without question. Discuss?? — CpiralCpiral 17:47, 20 September 2012 (UTC)
Sorry for the blanking out of the section without an explanation. But there is a valid explanation. The previous version of the WSD evaluation section contained outdated evaluation descriptions and links that was >10 year old and most techniques in evaluating WSD has changed drastically depending on which variety of WSD task. So i figured it is only appropriate to discuss about WSD evaluation specific to each variant of WSD. Most of the original content of WSD evaluation section had now been moved to Classic Monolingual WSD. It's because the monolingual WSD tasks is commonly referred to as "classic" in the recent evaluation workshops, see http://www.cs.york.ac.uk/semeval-2013/task12/index.php?id=task-description. Alvations (talk)
Simplifying the language
I think the language needs to be simplified here. It is not only because the language is a bit flowery, but it is also hard to understand. For instance, what do you mean exactly by POS-tagging and WSD making constraints to each other? I am not a specialist in WSD, so I cannot do a lot of editing without risk of introducing errors. — Preceding unsigned comment added by Srchvrs (talk • contribs) 00:19, 28 October 2013 (UTC)
Deep vs Shallow approaches
It is clearly a very subjective division. Especially if you claim that deep approaches require a comprehensive knowledge base. For example, the deep QA architecture of IBM Watson does not rely on such knowledge. Yet, it is considered deep enough. At the very least, you cannot generalize and say such a distinction is universal in NLP (using knowledge base as the single criterion). Going down the rabbit-hole, you can use only surface forms and basic co-occurrence statistics. But, you can also do some basic semantic parsing. Or syntax parsing. How deep would it be? Some knowledge is there, it is not exactly shallow.
A better classification is given here: http://aclweb.org/aclwiki/index.php?title=Word_sense_disambiguation