Sanderson, M. (1994) Word Sense Disambiguation and Information Retrieval. Proceedings of the 17th ACM SIGIR Conference, 142-151.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....However, machine translation of the documents outperforms all other methods with long queries. An important problem in the translation of short queries is the lack of context for diambiguation of words that have more than one meaning and therefore may correspond to more than one translation [13] [24]. Therefore, in the case of short queries mostly all translations are considered instead of trying to disambiguate between them. Alternatively, the user is asked to select the appropriate translation in an interactive setting [7] Ambiguity is also of importance to interlingua approaches to CLIR ....

Sanderson, M. 1994. Word Sense Disambiguation and Information Retrieval. In: Croft, B. and K. van Rijsbergen. (Eds.), In: Proc. of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval. pp. 142-151. Springer-Verlag.

....the previous chapter. 4.2.2 Word Sense Disambiguation A classical approach to dealing with polysemy is to perform word sense disambiguation on the ambiguous terms. Many different approaches have been tried, including the use of dictionaries, morphological analysis, bilingual corpera, and Wordnet [70]. While disambiguation of individual terms can often be accomplished with a high degree of precision [84] it has rarely led to improved performance for an IR system and has often produced a significant drop in performance [80] Sanderson [70] runs a simulation experiment where he replaces groups ....

.... morphological analysis, bilingual corpera, and Wordnet [70] While disambiguation of individual terms can often be accomplished with a high degree of precision [84] it has rarely led to improved performance for an IR system and has often produced a significant drop in performance [80] Sanderson [70] runs a simulation experiment where he replaces groups of words at random with a pseudo word to introduce artificial ambiguity and then measures the change in performance of the IR system. He finds that one can add a large amount of ambiguity before performance is extensively degraded. These ....

Mark Sanderson. Word sense disambiguation and information retrieval. In Proc. 17th Int'l Conference on R&D in IR (SIGIR), pages 142--150, 1994.

....make use of keywords and their stems in retrieving documents related to the query. These systems end up retrieving a lot of irrelevant information along with some useful information that the query was intended to elicit. An example Given in Word Sense DisambiGuation and Information letrieval [20] by Sanderson demonstrates this problem clearly. Sanderson Gives an example where a number of users tried to retrieve articles about the British Prime Minister usin G the query major . This query caused many articles about John Major to be retrieved. However, in addition many more articles ....

....disambiGuation of the word major prior to I1 would resolve the problem of retrievin G irrelevant documents. However, there is evidence which contradicts this claim. While Schutze and Pedersen [21] describe a sense disambiGuator that improves the precision of an II system by 4 , Sanderson [20] presents results which show that the disambiGuation process usually effects the performance of the II system negatively. He proves that in order to be of any practical use and in order to improve the performance of an II system, a disambiGuation algorithm has to work with at least 90 accuracy. ....

[Article contains additional citation context not shown here]

M. Sanderson. Word-sense Disambiguation and Information Retrieval. 1994 SIGIR. In Proceedings of the TOC, pages 142-151.

....initial results have not been promising, with both programs reporting deterioration in performance when the disambiguator is included. This agrees with the current wisdom in the IR community that unless disambiguation is highly accurate, it might not improve the retrieval system s perfor mance [13]. ....

Sanderson, Mark. 1994. "Word Sense Disambiguation and Information Retrieval," in Proceedings of $1GIR'94.

....[16] a type of query ambiguity that arises regularly in information retrieval. That type of ambiguity is often resolved implicitly when queries are long enough the additional words provide su#cient context to clear up confusion but is still a critical problem when queries are short [20]. Users of Web search engines generally provide short queries [15] and we are focusing on that situation in this study. There is a range of information retrieval interface ideas that attempt to help the user deal with ambiguous queries. Sanderson, Lawrie, and Croft have been working on concept ....

Mark Sanderson. Word sense disambiguation and information retrieval. In Proceedings of the ACM Conference on Research in Information Retrieval (SIGIR), pages 142--151, 1994.

....then the correct meaning of shot (action in sports) will be represented. This fact reduces the potential benefit of word sense disambiguation. Secondly, a mistake in word sense disambiguation can be very costly resulting in the retrieval engine retrieving many incorrect documents. Sanderson [18] conducted experiments to see if word sense disambiguation alone improves information retrieval performance. He used pseudo words, a mechanism introduced by Yarowsky [25] Pseudo words are constructs such as banana rifle whose actual meaning is one of the component words. By substituting pseudo ....

M. Sanderson. Word-sense disambiguation and information retrieval. In Proceedings of SIGIR-94, 17th ACM International Conference on Reasearch and Development in Information Retrieval, 1997. 59

....to another, as indicated in Figure i and Table 2. Average query length per query set was varied because we wanted to monitor to what extent performance would change relative to query length, since it has been shown that longer query statements reduce ambiguity associated with very short queries [28]. 3.2.2 Logical Test Set In the second phase, we wanted to strengthen our approach by using a query set that was characterized by (a) fewer positive relevance judgments per query (i.e. sparser hits) and (b) slightly longer queries (thus better simulating the issues driven information needs that ....

M. Sanderson. Word sense disambiguation and information retrieval. In Proc. of SIGIR '9J, pages 142-151. SpringerVerlag, July 1994.

....the translation disambiguation problem bears a strong resemblence to term disambiguation in a 18 monolingual setting. In fact, a number of researchers [8, 3] have used cross language relationships to help with disambiguation. Given the limited success of term disambiguation as a tool for IR [23], there is some question about whether we can hope to gain any benefits out of translation filtering. Translation disambiguation may well work better as an interactive tool. If the user has some familiarity with the target language, he or she could be given a list of items and asked to select the ....

Mark Sanderson. Word sense disambiguation and information retrieval. In Proc. of the 17th ACM/SIGIR Conference, pages 142--150, 1994.

....to the local search engine (this is the case in which L(E i ) # L(E 1j ) L(Emj ) and when it is translated into the pivot language in order to be dispatched to other search engines. The typically large size of the category profile usually ensures high quality disambiguation [18]. 1.2 Query refinement No matter whether they are issued from within categories or from semantically neutral environments, queries always return less than perfect results: some irrelevant documents are ranked high in the list, and some relevant documents are ranked low. It is thus necessary to ....

M. Sanderson. Word sense disambiguation and information retrieval. In Proceedings of SIGIR94, 17th ACM International Conference on Research and Development in Information Retrieval, pages 49--57, Dublin, IE, 1994.

....until after the evaluation had been completed. Nonetheless, the system consistently gave results that appeared reasonable to onlookers, and when informally compared with the results with the stemmer disabled, there was little or no difference. This is perhaps not surprising when the conclusions of Sanderson (1994) are considered i.e. the effectiveness of traditional IR techniques is remarkably resilient to ambiguity in the term space. Twelve volunteer users were recruited. They were all postgraduate level educated, and familiar with windowed environments, graphical interfaces, and had experience with a ....

Sanderson M (1994). Word Sense Disambiguation and Information Retrieval. In: Croft W B and van Rijsbergen C J, eds. Proceedings of ACM SIGIR conference, 1994. pp. 142-151.

....15] Most of the research has concentrated on how large the impact of semantical variation and its inaccurate resolution is on IR e ectiveness. It is estimated that if word sense disambiguation is performed with less than 90 accuracy the retrieval results are worse than not disambiguating at all [16]. Poor retrieval results were blamed on this reason in previous research [14] Conversely, in the same experiments [16] word sense ambiguity was shown to produce only minor e ects on retrieval accuracy, apparently suggesting that query document matching strategies already perform an implicit ....

....is on IR e ectiveness. It is estimated that if word sense disambiguation is performed with less than 90 accuracy the retrieval results are worse than not disambiguating at all [16] Poor retrieval results were blamed on this reason in previous research [14] Conversely, in the same experiments [16] word sense ambiguity was shown to produce only minor e ects on retrieval accuracy, apparently suggesting that query document matching strategies already perform an implicit disambiguation. In this experimental setup, ambiguity was introduced arti cially by substituting randomly selected word ....

M. Sanderson. Word Sense Disambiguation and Information Retrieval. In W. Bruce Croft and C. J. van Rijsbergen, editors, Proceedings of the 17th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 142-151, Dublin, Ireland, June 1994. ACM Press.

....reproduced on the IR Semcor test collection (derived from Semcor) where both queries and documents are hand tagged with phrases, Part Of Speech and WordNet 1.5 senses. Our results indicate that a) Word Sense Disambiguation can be more beneficial to Information Retrieval than the experiments of Sanderson (1994) with artificially ambiguous pseudo words suggested. b) PartOf Speech tagging does not seem to help improving retrieval, even if it is manually annotated. c) Using phrases as indexing terms is not a good strategy if no partial credit is given to the phrase components. 1 Introduction A major ....

....also harmed retrieval efficiency. Krovetz noted than more than half of the words in a dictionary that differ in POS are related in meaning, but he could not decide whether the decrease of performance was due to the loss of such semantic relatedness or to automatic POS tagging errors. 3. In (Sanderson, 1994), the problem of discerning the effects of differentiating word senses from the effects of inaccurate disambiguation was overcome using artificially created pseudo words (substituting, for instance, all occurrences of banana or kalashnikov for banana kalashnikov) that could be disambiguated with ....

[Article contains additional citation context not shown here]

M. Sanderson. 1994. Word Sense Disambiguation and Information Retrieval. In Proceedings of 17 t h International Conference on Research and Development in Information Retrieval.

....of the what these error rates actually means on the text of the queries is reported in the appendix. 5. 3 The Information Retrieval System The system used in the context of the work reported in this paper is an experimental IR toolkit developed at Glasgow University mostly by Mark Sanderson [21]. The system is a collection of small independent modules each conducting one part of the indexing, retrieval and evaluation tasks required for classic IR experimentation. The modules are linked in a pipeline architecture communicating through a common token based language. The system is ....

M. Sanderson. Word Sense Disambiguation and Information Retrieval. PhD Thesis, Department of Computing Science, University of Glasgow, Glasgow, Scotland, UK, 1996.

....Most of the research has concentrated into how large is the impact of semantical variation and its inaccurate resolution on IR e ectiveness. It is estimated that if word sense disambiguation is performed with less than 90 accuracy, the retrieval results are worse than not disambiguating at all [16]. Poor retrieval results were blamed on this reason in previous research [14] Conversely, in the same experiments [16] word sense ambiguity was shown to produce only minor e ects on retrieval accuracy, apparently suggesting that query document matching strategies already perform an implicit ....

....on IR e ectiveness. It is estimated that if word sense disambiguation is performed with less than 90 accuracy, the retrieval results are worse than not disambiguating at all [16] Poor retrieval results were blamed on this reason in previous research [14] Conversely, in the same experiments [16], word sense ambiguity was shown to produce only minor e ects on retrieval accuracy, apparently suggesting that query document matching strategies already perform an implicit disambiguation. In his experimental setup, ambiguity was introduced arti cially by substituting randomly selected word ....

M. Sanderson. Word Sense Disambiguation and Information Retrieval. In W. Bruce Croft and C. J. van Rijsbergen, editors, Proceedings of the 17th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 142-151, Dublin, Ireland, June 1994. ACM Press.

....typical in cases where automatic segmentation would be needed, would be expected to provide sufficient context for co occurrence relationships within the documents being searched to favor the correct terms over the incorrect ones. Sanderson saw a similar effect with conflated terms, for example [11]. Based on this insight, we chose to also explore a task tuned segmentation strategy that we call exhaustive segmentation. In exhaustive segmentation, every substring for which a translation is known is extracted from the query. This represents the opposite extreme from the one best segmentation ....

Mark Sanderson, "Word Sense Disambiguation and Information Retrieval", in W Bruce Croft and C. J. van Rijsbergen, editors, Proceedings of the 17th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, Springer-Verlag, July 1994, pp. 142-151.

....of subsequent information retrieval. With performance figures less than that the amount of noise introduced by incorrect disambiguation choices brings down the relative performance of retrieval when compared to indexing and retrieval based on simple word stems. A series of subsequent experiments [22] confirmed earlier results as well as exploring a number of other WSD approaches for an information retrieval application. The current thinking among those who work on using WSD for IR seems to be to use multiple sources of evidence in indexing by word senses by using partsof speech, word ....

Sanderson, M. : Word Sense Disambiguation and Information Retrieval. PhD thesis, Department of Computing Science, University of Glasgow, Scotland, 1997.

....70 accurate recognition of the unique word sense among all words in input texts seems to be about the best. In a series of information retrieval simulations in which he used the concept of a pseudo word to incorporate a synthetic value for the accuracy of a word sense disambiguator, Sanderson [21] showed that a WSD accuracy of at least 90 is required before the payback of WSD begins to have a positive effect on the effectiveness of subsequent information retrieval. With performance figures less than that the amount of noise introduced by incorrect disambiguation choices brings down the ....

Sanderson, M. : Word Sense Disambiguation and Information Retrieval. in Proc. 17th Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, Dublin, Ireland, 142-151, Springer-Verlag, 1994.

....machine readable dictionaries (MRDs) and knowledge bases (KBs) as candidates. The incorporation of MRDs into information retrieval was ongoing at the University of Massachusetts through work by Bob Krovetz and others (Krovetz and Croft, 1992)and at Glasgow University through work by Mark Sanderson (Sanderson, 1994). These sites were having mixed results in their experiments. NLP OR NLP RESOURCES FOR INFORMATION RETRIEVAL 7 An even richer resource than an MRD is a knowledge base and for largescale, domain independent knowledge bases there are only two, CYC and WordNet. Cyc is a long term project initially ....

Sanderson, M. (1994). Word Sense Disambiguation and Information Retrieval. in Proceedings of the 17th ACM-SIGIR Conference on Research and Development in Information Retrieval, Dublin, Ireland, pp. 49-57.

....is no guarantee that one s top ranked documents will have been assessed for relevance though it is impossible to see how anything could be done here. Many aspects of our overall implementation have not been fine tuned. For example, the word sense disambiguator has not been evaluated and we know [Krov92, Sand94] how crucial sense disambiguation is to the quality of retrieval. This compares with the performance enhanced, for the WSJ) variation of a standard tf IDF system we used as a benchmark. There are many occurrences of proper nouns in TREC queries which do not occur in WordNet. In fact there is a ....

: M. Sanderson, "Word Sense Disambiguation and Information Retrieval", Proceedings of the Seventeenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, Dublin, 49-57, 1994

....fails. Similar problems arise in IR. Take the example of a synonym based constraint. If the term referred to in this constraint is polysemic, disambiguation is necessary to ensure that the constraint can be used. In IR, disambiguation unfortunately is not always successful or even possible [23, 24]. This implies that the use of such a constraint will often be uncertain, thus leading to an uncertain flow. The more uncertain the flow, the less relevant the document. In IR, it is usually insufficient to say that a document is relevant to a query; knowing the extent of relevance is mandatory ....

M. Sanderson. Word sense disambiguation and information retrieval. In W. Bruce Croft and C.J. van Rijsbergen, editors, Proceedings of the Seventeenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 142--151. ACM, Springer-Verlag, July 1994.

....test collections to Harman s work. However, results were not consistent across all collections. Interestingly, evidence from other fields is also mixed. For example, Riloff reported that stemming had a deleterious effect on term classification [14] whereas Sanderson s work on sense disambiguation [16] recorded that random ambiguation of terms had minimal impact on retrieval and that only very accurate re disambiguation regained lost performance, suggesting that stemming should at worst have no effect and could potentially result in significant gains. Intuitively these results are quite ....

Mark Sanderson. Word sense disambiguation and information retrieval. In Croft and van Rijsbergen [3], pages 142--151.

....for evaluating the effectiveness of WSD. The general conclusion is that WSD in IR is worthwhile even though their results are not compelling. Another view is that word sense disambiguation do not affect retrieval results significantly unless the IR system is retrieving from very short queries. [San94] This makes sense intuitively because the more words there are in the context, the easier it is to find clues about an ambiguous word. For example, joint venture is a meaningful term but joint and venture alone refer to many irrelevant senses. Two striking observations Yarowsky [Yar95] ....

Mark Sanderson. Word sense disambiguation and information retrieval. Proceedings of SIGIR`94, 17 th ACM International Conference on Research and Development in Information Retrieval, pp.142-151, 1994.

....20 30 Query words (w) Baseline Unmodified figure 4 2.3 Simple experiment To illustrate the use of the collection, a set of experiments were performed which compared the effect on retrieval performance of two different stemming algorithms. A more comprehensive set of experiments are outlined in [Sanderson 94] The stemming algorithms chosen were a modified version of the Porter stemmer [Porter 80] and the stemmer provided with the WordNet thesaurus [Miller 90] In total, three experiments were run, one experiment for each stemmer and an experiment where no stemming was used. Figure 5 shows the ....

M. Sanderson (1994) "Word sense disambiguation and information retrieval" Proceedings of the ACM SIGIR Conference, Vol. 17

....tasks required for classic retrieval experimentation. The modules are linked in a pipeline architecture communicating through a common token based language. SIRE was initially used in research examining the relationship between word sense ambiguity, disambiguation, and retrieval effectiveness [17]. It proved to be a flexible tool as it not only provided retrieval functionality but a number of its core modules were used to build a word sense disambiguator as well. It was also used in the experiments for the Glasgow IR group submissions to TREC 4 and TREC 5 and is currently being used in a ....

M. Sanderson. Word Sense Disambiguation and Information Retrieval. PhD Thesis, Department of Computing Science, University of Glasgow, Glasgow, Scotland, UK, 1996.

....tasks required for classic retrieval experimentation. The modules are linked in a pipeline architecture communicating through a common token based language. SIRE was initially used in research examining the relationship between word sense ambiguity, disambiguation, and retrieval effectiveness [8]. It proved to be a flexible tool as it not only provided retrieval functionality but a number of its core modules were used to build a word sense disambiguator as well. It was also used in the experiments for the Glasgow IR group submissions to TREC 4 and TREC 5 and is currently being used in a ....

....this experiment, after submitting to TREC, some further expansion strategies were attempted on the 251 300 queries and a strategy was found that improved upon previous strategies, though still caused a drop in effectiveness, albeit a small one. The strategy was motivated from work reported by [8] which showed how the frequency of occurrence of the senses of words was skewed so that the most common sense of a word typically accounted for the majority of occurrences of that word. With this information in mind, it was surmised that query words used in their commonest sense did not need ....

[Article contains additional citation context not shown here]

M. Sanderson. Word Sense Disambiguation and Information Retrieval. PhD Thesis, Department of Computing Science, University of Glasgow, Glasgow, Scotland, UK, 1996.

....for classic retrieval experimentation [7, pp. 144 183] The modules are linked in a pipeline architecture communicating through a common token based language. SIRE was initially used in research examining the relationship between word sense ambiguity, disambiguation, and retrieval effectiveness [8]. It proved to be a flexible tool as it not only provided retrieval functionality but a number of its core modules were also used to build a word sense disambiguator. It was further used in the Glasgow IR group submissions to TRECs 4 5 [9, 10] and is currently being used in a number of research ....

Sanderson, M. Word Sense Disambiguation and Information Retrieval, PhD Thesis, Technical Report (TR-1997-7) of the Department of Computing Science at the University of Glasgow, Glasgow G12 8QQ, UK.

No context found.

Sanderson, M. (1994) Word Sense Disambiguation and Information Retrieval. Proceedings of the 17th ACM SIGIR Conference, 142-151.

No context found.

M. Sanderson. Word-sense disambiguation and information retrieval. In SIGIR 1994.

No context found.

Mark Sanderson. Word sense disambiguation and information retrieval. In Proceedings of ACM-SIGIR , pages 49--57, Dublin, IE, 1994.

No context found.

Sanderson, M. Word sense disambiguation and information retrieval. In Proceedings of ACM SIGIR Conference on Research and Development in Information Retrieval (Dublin, Ireland, 1994), W. B. Croft and C. J. van Rijsbergen, Eds., pp. 142--151.

No context found.

M. Sanderson. Word Sense Disambiguation and Information Retreival. In Proceedings of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieaval, pages 142--151. Springer-Verlag, 1994.

No context found.

M. Sanderson. Word sense disambiguation and information retrieval. In Proceedings of SIGIR, pages 142--151, 1994.

No context found.

, pp. 142-151, 1994.

No context found.

Sanderson, M.: Word sense disambiguation and information retrieval. In: Proceedings of the 17 ACM SIGIR Conference, Dublin, Ireland (1994) 142--151.

No context found.

Mark Sanderson. Word Sense Disambiguation and Information Retrieval. PhD thesis, University of Glasgow, 1997. Technical Report (TR-1997-7).

No context found.

Mark Sanderson. Word sense disambiguation and information retrieval. In Proceedings of the 17th Conference on Research and Development in Information Retrieval (SIGIR{94), pages 49-57, 1994.

No context found.

M. Sanderson. Word Sense Disambiguation and Information Retreival. In Proceedings of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieaval, pages 142--151. Springer-Verlag, 1994.

No context found.

M. Sanderson. Word Sense Disambiguation and Information Retreival. In Proceedings of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieaval, pages 142-151. Springer-Verlag, 1994.

No context found.

M. Sanderson. Word Sense Disambiguation and Information Retrieval. PhD thesis, Department of Computing Science at the University of Glasgow, 1997.

No context found.

SANDERSON, M., Word Sense Disambiguation and Information Retrieval. Ph.D. Thesis, Technical Report (TR-1997-7) of the Department of Computing Science, University of Glasgow.

No context found.

Sanderson, M. Word sense disambiguation and information retrieval. Proceedings of the 17th Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, pages 142-151, Springer-Verlag, 1994.

No context found.

Sanderson, M.: Word sense disambiguation and information retrieval. Ph.D. Thesis, Department of Computing Science, University of University of Glasgow. 1996.

No context found.

Sanderson M, "Word Sense Disambiguation and Information Retrieval", in Proceedings of the Seventeenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 142--151, 1994.

No context found.

Mark Sanderson. 1994. Word sense disambiguation and information retrieval. In Proceedings, ACM Special Interest Group on Information retrieval, pages 142--151.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC