Word sense disambiguation - Revision history

Tristan Miller: /* See also / Word sense disambiguation resources

2014-12-12T11:21:51Z

← Older revision		Revision as of 11:21, 12 December 2014
Line 40:		Line 40:
	== See also ==		== See also ==
	* [[Word Sense Disambiguation (State of the art)]]		* [[Word Sense Disambiguation (State of the art)]]
			* [[Word sense disambiguation resources]]

	== External links ==		== External links ==

Pdturney: Reverted edits by Creek (talk) to last revision by Pdturney

2012-06-25T11:20:24Z

Reverted edits by Creek (talk) to last revision by Pdturney

← Older revision		Revision as of 11:20, 25 June 2012
Line 19:		Line 19:
	The information in these resources has been used in several ways, for example Wilks and Stevenson <ref>Y. Wilks and M. Stevenson. The Grammar of Sense: using part-of-speech tags as a first step in semantic disambiguation. To appear in Journal of Natural Language Engineering, 4(3).</ref> use large lexicons (generally machine readable dictionaries) and the information associated with the senses (such as part-of-speech tags, topical guides and selectional preferences) to indicate the correct sense. Another approach is to treat the text as an unordered bag of words where similarity measures are calculated by looking at the semantic similarity (as measured from the knowledge source) between all the words in the window regardless of their positions, as was used by Yarowsky <ref>D. Yarowsky. Word-sense disambiguation using statistical models of Roget's categories trained on large corpora. In Proceedings of the 14th International Conference on Computational Linguistics (COLING-92), pages 454-460, Nantes, France, 1992.</ref>.		The information in these resources has been used in several ways, for example Wilks and Stevenson <ref>Y. Wilks and M. Stevenson. The Grammar of Sense: using part-of-speech tags as a first step in semantic disambiguation. To appear in Journal of Natural Language Engineering, 4(3).</ref> use large lexicons (generally machine readable dictionaries) and the information associated with the senses (such as part-of-speech tags, topical guides and selectional preferences) to indicate the correct sense. Another approach is to treat the text as an unordered bag of words where similarity measures are calculated by looking at the semantic similarity (as measured from the knowledge source) between all the words in the window regardless of their positions, as was used by Yarowsky <ref>D. Yarowsky. Word-sense disambiguation using statistical models of Roget's categories trained on large corpora. In Proceedings of the 14th International Conference on Computational Linguistics (COLING-92), pages 454-460, Nantes, France, 1992.</ref>.
	===Corpus based===		===Corpus based===
	This approach attempts to disambiguate words using information which is gained by training on some corpus, rather that taking it directly from an explicit knowledge source. This training can be carried out on ~~[http://www.thai-sbobet.com sbobet]~~either a disambiguated or raw corpus, where a disambiguated corpus is one where the semantics of each polysemous lexical item is marked and a raw corpus one without such marking.		This approach attempts to disambiguate words using information which is gained by training on some corpus, rather that taking it directly from an explicit knowledge source. This training can be carried out on either a disambiguated or raw corpus, where a disambiguated corpus is one where the semantics of each polysemous lexical item is marked and a raw corpus one without such marking.
	====Disambiguated corpora====		====Disambiguated corpora====
	This set of techniques requires a training corpus which has already been disambiguated. In general a machine learning algorithm of some kind is applied to certain features extracted from the corpus and used to form a representation of each of the senses. This representation can then be applied to new instances in order to disambiguate them. Different researchers have made use of different sets of features, for example local collocates such as first noun to the left and right, second word to the left/right and so on. However, a more common feature set is to take all the words in a window of words around the ambiguous words, treating the context as an unordered bag of words.		This set of techniques requires a training corpus which has already been disambiguated. In general a machine learning algorithm of some kind is applied to certain features extracted from the corpus and used to form a representation of each of the senses. This representation can then be applied to new instances in order to disambiguate them. Different researchers have made use of different sets of features, for example local collocates such as first noun to the left and right, second word to the left/right and so on. However, a more common feature set is to take all the words in a window of words around the ambiguous words, treating the context as an unordered bag of words.
Line 31:		Line 31:
	An example of this is the dynamic matching technique<ref>Radford et al. (1996)</ref> which examines all instances of a given term in a corpus and compares the contexts in which they occur for common words and syntactic patterns. A similarity matrix is thus formed which is subject to cluster analysis to determine groups of semantically related instances of terms.		An example of this is the dynamic matching technique<ref>Radford et al. (1996)</ref> which examines all instances of a given term in a corpus and compares the contexts in which they occur for common words and syntactic patterns. A similarity matrix is thus formed which is subject to cluster analysis to determine groups of semantically related instances of terms.
	Another example is the work of Pedersen <ref>T. Pedersen and R. Bruce. Distinguishing word senses in untagged text. In Proceedings of the Second Conference on Empirical Methods in Natural Language Processing, Providence, RI, August 1997.</ref> who compared three different unsupervised learning algorithms on 13 different words. Each algorithm was trained on text with was tagged with either the WordNet or LDOCE sense for the word but the algorithm had no access to the truce senses. What it did have access to was the number of senses for each word and each algorithm split the instances of each word into the appropriate number of clusters. These clusters were then mapped onto the closest sense from the appropriate lexicon. Unfortunately the results are not very encouraging, Pedersen reports 65-66% correct disambiguation depending on the learning algorithm used. This result should be compared against that fact that, in the corpus he used, 73% of the instances could be correctly classified by simply choosing the most frequent sense.		Another example is the work of Pedersen <ref>T. Pedersen and R. Bruce. Distinguishing word senses in untagged text. In Proceedings of the Second Conference on Empirical Methods in Natural Language Processing, Providence, RI, August 1997.</ref> who compared three different unsupervised learning algorithms on 13 different words. Each algorithm was trained on text with was tagged with either the WordNet or LDOCE sense for the word but the algorithm had no access to the truce senses. What it did have access to was the number of senses for each word and each algorithm split the instances of each word into the appropriate number of clusters. These clusters were then mapped onto the closest sense from the appropriate lexicon. Unfortunately the results are not very encouraging, Pedersen reports 65-66% correct disambiguation depending on the learning algorithm used. This result should be compared against that fact that, in the corpus he used, 73% of the instances could be correctly classified by simply choosing the most frequent sense.

	===Hybrid Approaches===		===Hybrid Approaches===
	These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.		These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.

Creek: /* Corpus based */

2012-06-25T10:31:46Z

Corpus based

← Older revision		Revision as of 10:31, 25 June 2012
Line 19:		Line 19:
	The information in these resources has been used in several ways, for example Wilks and Stevenson <ref>Y. Wilks and M. Stevenson. The Grammar of Sense: using part-of-speech tags as a first step in semantic disambiguation. To appear in Journal of Natural Language Engineering, 4(3).</ref> use large lexicons (generally machine readable dictionaries) and the information associated with the senses (such as part-of-speech tags, topical guides and selectional preferences) to indicate the correct sense. Another approach is to treat the text as an unordered bag of words where similarity measures are calculated by looking at the semantic similarity (as measured from the knowledge source) between all the words in the window regardless of their positions, as was used by Yarowsky <ref>D. Yarowsky. Word-sense disambiguation using statistical models of Roget's categories trained on large corpora. In Proceedings of the 14th International Conference on Computational Linguistics (COLING-92), pages 454-460, Nantes, France, 1992.</ref>.		The information in these resources has been used in several ways, for example Wilks and Stevenson <ref>Y. Wilks and M. Stevenson. The Grammar of Sense: using part-of-speech tags as a first step in semantic disambiguation. To appear in Journal of Natural Language Engineering, 4(3).</ref> use large lexicons (generally machine readable dictionaries) and the information associated with the senses (such as part-of-speech tags, topical guides and selectional preferences) to indicate the correct sense. Another approach is to treat the text as an unordered bag of words where similarity measures are calculated by looking at the semantic similarity (as measured from the knowledge source) between all the words in the window regardless of their positions, as was used by Yarowsky <ref>D. Yarowsky. Word-sense disambiguation using statistical models of Roget's categories trained on large corpora. In Proceedings of the 14th International Conference on Computational Linguistics (COLING-92), pages 454-460, Nantes, France, 1992.</ref>.
	===Corpus based===		===Corpus based===
	This approach attempts to disambiguate words using information which is gained by training on some corpus, rather that taking it directly from an explicit knowledge source. This training can be carried out on either a disambiguated or raw corpus, where a disambiguated corpus is one where the semantics of each polysemous lexical item is marked and a raw corpus one without such marking.		This approach attempts to disambiguate words using information which is gained by training on some corpus, rather that taking it directly from an explicit knowledge source. This training can be carried out on [http://www.thai-sbobet.com sbobet]either a disambiguated or raw corpus, where a disambiguated corpus is one where the semantics of each polysemous lexical item is marked and a raw corpus one without such marking.
	====Disambiguated corpora====		====Disambiguated corpora====
	This set of techniques requires a training corpus which has already been disambiguated. In general a machine learning algorithm of some kind is applied to certain features extracted from the corpus and used to form a representation of each of the senses. This representation can then be applied to new instances in order to disambiguate them. Different researchers have made use of different sets of features, for example local collocates such as first noun to the left and right, second word to the left/right and so on. However, a more common feature set is to take all the words in a window of words around the ambiguous words, treating the context as an unordered bag of words.		This set of techniques requires a training corpus which has already been disambiguated. In general a machine learning algorithm of some kind is applied to certain features extracted from the corpus and used to form a representation of each of the senses. This representation can then be applied to new instances in order to disambiguate them. Different researchers have made use of different sets of features, for example local collocates such as first noun to the left and right, second word to the left/right and so on. However, a more common feature set is to take all the words in a window of words around the ambiguous words, treating the context as an unordered bag of words.
Line 31:		Line 31:
	An example of this is the dynamic matching technique<ref>Radford et al. (1996)</ref> which examines all instances of a given term in a corpus and compares the contexts in which they occur for common words and syntactic patterns. A similarity matrix is thus formed which is subject to cluster analysis to determine groups of semantically related instances of terms.		An example of this is the dynamic matching technique<ref>Radford et al. (1996)</ref> which examines all instances of a given term in a corpus and compares the contexts in which they occur for common words and syntactic patterns. A similarity matrix is thus formed which is subject to cluster analysis to determine groups of semantically related instances of terms.
	Another example is the work of Pedersen <ref>T. Pedersen and R. Bruce. Distinguishing word senses in untagged text. In Proceedings of the Second Conference on Empirical Methods in Natural Language Processing, Providence, RI, August 1997.</ref> who compared three different unsupervised learning algorithms on 13 different words. Each algorithm was trained on text with was tagged with either the WordNet or LDOCE sense for the word but the algorithm had no access to the truce senses. What it did have access to was the number of senses for each word and each algorithm split the instances of each word into the appropriate number of clusters. These clusters were then mapped onto the closest sense from the appropriate lexicon. Unfortunately the results are not very encouraging, Pedersen reports 65-66% correct disambiguation depending on the learning algorithm used. This result should be compared against that fact that, in the corpus he used, 73% of the instances could be correctly classified by simply choosing the most frequent sense.		Another example is the work of Pedersen <ref>T. Pedersen and R. Bruce. Distinguishing word senses in untagged text. In Proceedings of the Second Conference on Empirical Methods in Natural Language Processing, Providence, RI, August 1997.</ref> who compared three different unsupervised learning algorithms on 13 different words. Each algorithm was trained on text with was tagged with either the WordNet or LDOCE sense for the word but the algorithm had no access to the truce senses. What it did have access to was the number of senses for each word and each algorithm split the instances of each word into the appropriate number of clusters. These clusters were then mapped onto the closest sense from the appropriate lexicon. Unfortunately the results are not very encouraging, Pedersen reports 65-66% correct disambiguation depending on the learning algorithm used. This result should be compared against that fact that, in the corpus he used, 73% of the instances could be correctly classified by simply choosing the most frequent sense.

	===Hybrid Approaches===		===Hybrid Approaches===
	These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.		These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.

Pdturney: /* History */

2011-01-04T18:55:18Z

History

← Older revision		Revision as of 18:55, 4 January 2011
Line 12:		Line 12:

	==History==		==History==
	The problem of WSD was first introduced by Warren Weaver in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>		The problem of WSD was first introduced by Warren Weaver in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>. In 1975 Kelly and Stone <ref>E.F. Kelly and P.J. Stone. 1975. Computer Recognition of English Word Senses, Amsterdam: North-Holland.</ref> published a book explicitly listing their rules for disambiguation of word senses. As large-scale lexical resources became available in the 1980s, the automatic extraction of lexical knowledge became possible, disambiguation was still knowledge- or dictionary - based though. With the rise of statistical methods in CL in the 1990s, WSD became one of the main focus' of supervised learning techniques.
	In 1975 Kelly and Stone <ref>E.F. Kelly and P.J. Stone, 1975, Computer Recognition of English Word Senses, Amsterdam: North-Holland.</ref> published a book explicitly listing their rules for disambiguation of word senses.
	As large-scale lexical resources became available in the 1980s, the automatic extraction of lexical knowledge became possible, disambiguation was still knowledge- or dictionary - based though.
	With the rise of statistical methods in CL in the 1990s, WSD became one of the main focus' of supervised learning techniques.

	==Approaches==		==Approaches==

Pdturney: /* History */

2011-01-04T18:54:24Z

History

← Older revision		Revision as of 18:54, 4 January 2011
Line 12:		Line 12:

	==History==		==History==
	The problem of WSD was first introduced by Warren ~~weaver~~ in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>		The problem of WSD was first introduced by Warren Weaver in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>
	In 1975 Kelly and Stone published a book explicitly listing their rules for disambiguation of word senses.		In 1975 Kelly and Stone <ref>E.F. Kelly and P.J. Stone, 1975, Computer Recognition of English Word Senses, Amsterdam: North-Holland.</ref> published a book explicitly listing their rules for disambiguation of word senses.
	As large-scale lexical resources became available in the 1980s, the automatic extraction of lexical knowledge became possible, disambiguation was still knowledge- or dictionary - based though.		As large-scale lexical resources became available in the 1980s, the automatic extraction of lexical knowledge became possible, disambiguation was still knowledge- or dictionary - based though.
	With the rise of statistical methods in CL in the 1990s, WSD became one of the main focus' of supervised learning techniques.		With the rise of statistical methods in CL in the 1990s, WSD became one of the main focus' of supervised learning techniques.

Pdturney: /* Introduction */

2011-01-04T18:49:53Z

Introduction

← Older revision		Revision as of 18:49, 4 January 2011
Line 8:		Line 8:
	#The boy leapt from the bank into the cold water.		#The boy leapt from the bank into the cold water.
	#The van pulled up outside the bank and three masked men got out.		#The van pulled up outside the bank and three masked men got out.
	We immediately recognise that in the first sentence bank refers to the edge of a river and in the second to a building. However, the task has proved to be difficult for computer and some have believed that it would never be solved. An early sceptic was Bar-Hillel who famously proclaimed that "sense ambiguity could not be resolved by electronic computer either current or imaginable''. <ref>Y. Bar-Hillel. Language and Information. Addison-Wesley, 1964.</ref>		We immediately recognise that in the first sentence bank refers to the edge of a river and in the second to a building. However, the task has proved to be difficult for computer and some have believed that it would never be solved. An early sceptic was Bar-Hillel who famously proclaimed that "sense ambiguity could not be resolved by electronic computer either current or imaginable". <ref>Y. Bar-Hillel. Language and Information. Addison-Wesley, 1964.</ref>
	However, the situation is not as bad as Bar-Hillel feared, there have been several advances in word sense disambiguation and it is now at a stage where lexical ambiguity in text can be resolved with a reasonable degree of accuracy.		However, the situation is not as bad as Bar-Hillel feared, there have been several advances in word sense disambiguation and it is now at a stage where lexical ambiguity in text can be resolved with a reasonable degree of accuracy.

	==History==		==History==
	The problem of WSD was first introduced by Warren weaver in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>		The problem of WSD was first introduced by Warren weaver in 1949 <ref>W. Weaver. 1949. [http://www.hutchinsweb.me.uk/MTNI-22-1999.pdf Translation]. In Machine Translation of Languages: Fourteen Essays, ed. by Locke, W.N. and Booth, A.D. Cambridge, MA: MIT Press.</ref>

Pdturney: /* Hybrid Approaches */

2011-01-04T18:46:59Z

Hybrid Approaches

← Older revision		Revision as of 18:46, 4 January 2011
Line 35:		Line 35:
	===Hybrid Approaches===		===Hybrid Approaches===
	These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.		These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.
	Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the ``obvious'' cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.		Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the "obvious" cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.

	==Discussions==		==Discussions==
	Word Sense Disambiguation has several debates within the field as to whether the senses offered in existing dictionaries are adequate to distinguish the subtle meanings used in text contexts and how to evaluate the overall performance of a WSD system. For example, does it make sense to describe an overall percentage accuracy for a WSD system or does evaluation require specific comparison of system performance on a word by word basis.		Word Sense Disambiguation has several debates within the field as to whether the senses offered in existing dictionaries are adequate to distinguish the subtle meanings used in text contexts and how to evaluate the overall performance of a WSD system. For example, does it make sense to describe an overall percentage accuracy for a WSD system or does evaluation require specific comparison of system performance on a word by word basis.

Pdturney: Reverted edits by Armin (Talk); changed back to last version by Lhausmann

2011-01-04T18:42:03Z

Reverted edits by Armin (Talk); changed back to last version by Lhausmann

← Older revision		Revision as of 18:42, 4 January 2011
Line 36:		Line 36:
	These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.		These approaches can be neither properly classified as knowledge or corpus based but use part of both approaches. A good example of this is Luk's system <ref>A. Luk. Statistical sense disambiguation with relatively small corpora using dictionary definitions. In Proceedings of the 33rd Meetings of the Association for Computational Linguistics (ACL-95), pages 181-188, Cambridge, M.A., 1995.</ref> which uses the textual definitions of senses from a machine readable dictionary (LDOCE) to identify relations between senses. He then uses a corpus to calculate mutual information scores between these related senses in order to discover the most useful. This allowed Luk to produce a system which used the information in lexical resources as a way of reducing the amount of text needed in the training corpus.
	Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the ``obvious'' cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.		Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the ``obvious'' cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.
	~~== WSD Evaluation ==~~
	~~Concise Information source: [http://en.wikipedia.org/wiki/Word_sense_disambiguation#Evaluation WSD Evaluation on Wikipedia]~~
	~~<BR>~~
	~~A 2010 Paper with new ideas: [http://www.informaworld.com/smpp/content~db=all~content=a929740345~frm=abslink Significance of Novel WSD Algorithms]~~

	==Discussions==		==Discussions==
	Word Sense Disambiguation has several debates within the field as to whether the senses offered in existing dictionaries are adequate to distinguish the subtle meanings used in text contexts and how to evaluate the overall performance of a WSD system. For example, does it make sense to describe an overall percentage accuracy for a WSD system or does evaluation require specific comparison of system performance on a word by word basis.		Word Sense Disambiguation has several debates within the field as to whether the senses offered in existing dictionaries are adequate to distinguish the subtle meanings used in text contexts and how to evaluate the overall performance of a WSD system. For example, does it make sense to describe an overall percentage accuracy for a WSD system or does evaluation require specific comparison of system performance on a word by word basis.

Armin: /* WSD Evaluation */

2011-01-04T14:01:35Z

WSD Evaluation

← Older revision		Revision as of 14:01, 4 January 2011
Line 39:		Line 39:
	Concise Information source: [http://en.wikipedia.org/wiki/Word_sense_disambiguation#Evaluation WSD Evaluation on Wikipedia]		Concise Information source: [http://en.wikipedia.org/wiki/Word_sense_disambiguation#Evaluation WSD Evaluation on Wikipedia]
	<BR>		<BR>
	Paper: [http://www.informaworld.com/smpp/content~db=all~content=a929740345~frm=abslink Significance of Novel WSD Algorithms]		A 2010 Paper with new ideas: [http://www.informaworld.com/smpp/content~db=all~content=a929740345~frm=abslink Significance of Novel WSD Algorithms]

	==Discussions==		==Discussions==

Armin: /* WSD Evaluation */

2011-01-04T14:00:43Z

WSD Evaluation

← Older revision		Revision as of 14:00, 4 January 2011
Line 37:		Line 37:
	Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the ``obvious'' cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.		Another example of this approach is the unsupervised algorithm of Yarowsky <ref>D. Yarowsky. Unsupervised word-sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Lainguistics (ACL '95), pages 189-196, Cambridge, MA, 1995.</ref>. This takes a small number of seed definitions of the senses of some word (the seeds could be WordNet synsets or definitions from some lexicon) and uses these to classify the ``obvious'' cases in a corpus. Decision lists <ref>R. Rivest. Learning decision lists. Machine Learning, 2(3):229-246, 1987</ref> are then used to make generalisations based on the corpus instances classified so far and these lists are then re-applied to the corpus to classify more instances. The learning proceeds in this way until all corpus instances are classified. Yarowsky reports that the system correctly classifies senses 96% of the time.
	== WSD Evaluation ==		== WSD Evaluation ==
	Concise Information source: [http://en.wikipedia.org/wiki/Word_sense_disambiguation#Evaluation WSD Evaluation on Wikipedia]		Concise Information source: [http://en.wikipedia.org/wiki/Word_sense_disambiguation#Evaluation WSD Evaluation on Wikipedia]
			<BR>
	Paper: [http://www.informaworld.com/smpp/content~db=all~content=a929740345~frm=abslink Significance of Novel WSD Algorithms]		Paper: [http://www.informaworld.com/smpp/content~db=all~content=a929740345~frm=abslink Significance of Novel WSD Algorithms]

Word sense disambiguation - Revision history

Tristan Miller: /* See also */ * Word sense disambiguation resources

Pdturney: Reverted edits by Creek (talk) to last revision by Pdturney

Creek: /* Corpus based */

Pdturney: /* History */

Pdturney: /* History */

Pdturney: /* Introduction */

Pdturney: /* Hybrid Approaches */

Pdturney: Reverted edits by Armin (Talk); changed back to last version by Lhausmann

Armin: /* WSD Evaluation */

Armin: /* WSD Evaluation */

Tristan Miller: /* See also / Word sense disambiguation resources