A Semi-supervised Learning Algorithm for Web Information Extraction with Tolerance Rough Sets

Learning relational facts from the web: A tolerance rough set approach

“…The categorical extractor module is the same as in our original TPL model [8]; it uses noun phrases, contextual patterns and the co-occurrence statistics to extract credible instances. The relational module is an extension.…”

“…Like CBS and CPL, TPL is an iterative algorithm which is appointed to run indefinitely. In every iteration, it learns new trusted instances of categories and it uses its growing knowledge to make 4 Calculate the approximations U A (n i ) and L A (n i ); 5 for each candidate noun phrase n j do 6 Calculate micro(n i , n j ); 7 for each candidate noun phrase n j do 8 macro cat (n j ) = ∀n i ∈cat micro(n i , n j ); 9 Rank instances by macro cat /|cat|; 10 Promote top instances as trusted; more informed judgments. As the first iteration is based on userlabeled seeds, it forms the supervised step of the algorithm.…”

“…This work is an extension of the TPL algorithm presented in [8] and is a first attempt at using a granular-based approach to extracting relational facts. We observed a natural affinity between term-based document clustering [9] and context-based noun phrase clustering as they both rely on the co-occurrence statistics of their respective counterparts: terms and contexts.…”

Harvesting Patterns from Textual Web Sources with Tolerance Rough Sets

“…Tolerance Rough Sets Adapted to NLP We can now define a tolerance approximation space categorizing noun phrases as K = ðCP; NP; I; l; hÞ. 28 In this work, the universes of entities are:…”