Ferreira, R.S.; Graça Pimentel, M. de; Cristo, M.: ¬A wikification prediction model based on the combination of latent, dyadic, and monadic features (2018)
0.01
0.007663213 = product of:
0.015326426 = sum of:
0.015326426 = product of:
0.030652853 = sum of:
0.030652853 = weight(_text_:web in 4119) [ClassicSimilarity], result of:
0.030652853 = score(doc=4119,freq=2.0), product of:
0.17002425 = queryWeight, product of:
3.2635105 = idf(docFreq=4597, maxDocs=44218)
0.052098576 = queryNorm
0.18028519 = fieldWeight in 4119, product of:
1.4142135 = tf(freq=2.0), with freq of:
2.0 = termFreq=2.0
3.2635105 = idf(docFreq=4597, maxDocs=44218)
0.0390625 = fieldNorm(doc=4119)
0.5 = coord(1/2)
0.5 = coord(1/2)
- Abstract
- Considering repositories of web documents that are semantically linked and created in a collaborative fashion, as in the case of Wikipedia, a key problem faced by content providers is the placement of links in the articles. These links must support user navigation and provide a deeper semantic interpretation of the content. Current wikification methods exploit machine learning techniques to capture characteristics of the concepts and its associations. In previous work, we proposed a preliminary prediction model combining traditional predictors with a latent component which captures the concept graph topology by means of matrix factorization. In this work, we provide a detailed description of our method and a deeper comparison with a state-of-the-art wikification method using a sample of Wikipedia and report a gain up to 13% in F1 score. We also provide a comprehensive analysis of the model performance showing the importance of the latent predictor component and the attributes derived from the associations between the concepts. Moreover, we include an analysis that allows us to conclude that the model is resilient to ambiguity without including a disambiguation phase. We finally report the positive impact of selecting training samples from specific content quality classes.
