Hypothesis Discovery via Knowledge Extraction and Generalization

点击翻译按钮获取中文摘要

Owing to the recent advances in computer technologies, a large amount of texts have become available. This research attempted to automate highly intellectual tasks analyzing such texts, which usually require labor-intensive work and domain knowledge. Specifically, we investigated gene function annotation and hypothesis discovery. For the former, we proposed an approach based on semantic kernels and achieved both efficient and effective gene annotation as compared with existing works. For the latter, we defined event similarity to spot valid hypotheses. Evaluative experiments revealed that our proposed approach was more stable and effective than previous frequency-based approaches.