Apappu writeup on Wang IJCNLP '09
From Cohen Courses
Jump to navigationJump to searchThis is a review of Wang_2009_automatic_set_instance_extraction_using_the_web by user:Apappu.
- Automatic set instance extractor builds up on the previous work noise-resistant version of SEAL.
- Authors propose a near-language independent and state-of-art approach that did well on benchmark tests.
- This approach consists of three major components that function in the following fashion:
- Set-Expander takes seeds of target concept as input and gives out an expanded set representing the target concept.
- Reranker ranks the noisy candidates using noise-resistant SEAL.
- Bootstrapper, after each iteration adds highly ranked output candidates to its input data.
Comment: I liked the idea of using back-off strategy to compensate for the lack of web documents that contain hyponym phrases.
