Academia.eduAcademia.edu

Figure 5 – uploaded by Siva Reddy

See full PDFdownloadDownload figure
As discussed in Section 4, feature ( sw, ) and ( sw;, pos(sw;), pos(swo) ) are used as all the models. As k value is increased, FSCL method performs better than the baseline. We believe

Figure 5 As discussed in Section 4, feature ( sw, ) and ( sw;, pos(sw;), pos(swo) ) are used as all the models. As k value is increased, FSCL method performs better than the baseline. We believe

Related Figures (10)

Figure 1.2: Hypernym hierarchy of first synset of bill A Figure 1.3: Hindi WordNet entry of the word bilJA. The word has two synsets meaning male cat and badge. Ontological category mappings of the two senses are shown on the left side of the figure. On the right, the semantic category tree(SCT) of the word is shown. Semantic category tree (SCT) of a word is defined as the tree formed by merging the
Figure 1.2: Hypernym hierarchy of first synset of bill A Figure 1.3: Hindi WordNet entry of the word bilJA. The word has two synsets meaning male cat and badge. Ontological category mappings of the two senses are shown on the left side of the figure. On the right, the semantic category tree(SCT) of the word is shown. Semantic category tree (SCT) of a word is defined as the tree formed by merging the
preference to the words in SOC’, which are less ambiguous. The example shown in the figure 2.2 clarifies the aggregation and normalization steps In this paragraph, we discuss an alternative scoring mechanism. The same example i figure 2.2 is used to explain this mechanism. The probabilities in this are calculated as
preference to the words in SOC’, which are less ambiguous. The example shown in the figure 2.2 clarifies the aggregation and normalization steps In this paragraph, we discuss an alternative scoring mechanism. The same example i figure 2.2 is used to explain this mechanism. The probabilities in this are calculated as
assuming that this semantic Category 1S more Ukely to occur with the given feature. Step 3: To disambiguate an occurrence a word w; with it’s collocational features /; a top down walk is performed on the semantic category tree of w;. The set of categories at level h (denoted by SCT" (w,)) are disambiguated first before moving to disambiguate at level h + 1. Once a category is decided at level h, then the algorithm considers only the children of this category in level h + 1. This results in reducing the semantic category search space of disambiguation algorithm. For more details, refer to the algorithm below. Advantages of HSCL: search space of disambiguation algorithm. For more details, refer to the algorithm below. the children of this category in level h + 1. This results in reducing the semantic category e HSCL disambiguate level by level. Number of categories to be disambiguated in the
assuming that this semantic Category 1S more Ukely to occur with the given feature. Step 3: To disambiguate an occurrence a word w; with it’s collocational features /; a top down walk is performed on the semantic category tree of w;. The set of categories at level h (denoted by SCT" (w,)) are disambiguated first before moving to disambiguate at level h + 1. Once a category is decided at level h, then the algorithm considers only the children of this category in level h + 1. This results in reducing the semantic category search space of disambiguation algorithm. For more details, refer to the algorithm below. Advantages of HSCL: search space of disambiguation algorithm. For more details, refer to the algorithm below. the children of this category in level h + 1. This results in reducing the semantic category e HSCL disambiguate level by level. Number of categories to be disambiguated in the
Figure 4.1: Personalized Graph: Yellow and Blue nodes represent synsets and context words respectively. Green edges connecting the word and its synsets are assigned weights equal to Sense Ranking Scores (srs). Blue nodes are initialized with Keyword Ranking Scores (krs). Our approach uses the Personalized PageRank algorithm (Agirre and Soroa, 2009) with WordNet as the lexical knowledge base (LKB) to perform WSD. WordNet is converted to a sraph by representing each synset as a node (synset node) and the relationships in WordNet (hypernymy, hyponymy etc.) as edges between synset nodes. The graph is initialized by adding a node (word node) for each context word of the target word (including itself) thus creating a context dependent graph (personalized graph). The popular PageRank (Page st al., 1999) algorithm is employed to analyze this personalized graph (thus the algorithm is referred as personalized PageRank algorithm) and the sense for each disambiguous word is chosen by choosing the synset node which gets the highest weight after a certain number te . ~ —_™ —_ We capture domain information in the personalized graph by using sense ranking scores
Figure 4.1: Personalized Graph: Yellow and Blue nodes represent synsets and context words respectively. Green edges connecting the word and its synsets are assigned weights equal to Sense Ranking Scores (srs). Blue nodes are initialized with Keyword Ranking Scores (krs). Our approach uses the Personalized PageRank algorithm (Agirre and Soroa, 2009) with WordNet as the lexical knowledge base (LKB) to perform WSD. WordNet is converted to a sraph by representing each synset as a node (synset node) and the relationships in WordNet (hypernymy, hyponymy etc.) as edges between synset nodes. The graph is initialized by adding a node (word node) for each context word of the target word (including itself) thus creating a context dependent graph (personalized graph). The popular PageRank (Page st al., 1999) algorithm is employed to analyze this personalized graph (thus the algorithm is referred as personalized PageRank algorithm) and the sense for each disambiguous word is chosen by choosing the synset node which gets the highest weight after a certain number te . ~ —_™ —_ We capture domain information in the personalized graph by using sense ranking scores
Table 5.1: Evaluation results on Senseval-2 and Senseval-3 data-set of all words task. and Moldovan, 2001) and sense disambiguated gloss present in WordNet3.0. 5.5.3. Performance analysis
Table 5.1: Evaluation results on Senseval-2 and Senseval-3 data-set of all words task. and Moldovan, 2001) and sense disambiguated gloss present in WordNet3.0. 5.5.3. Performance analysis
Related topics:
Word Sense Disambiguation
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved