Verb Based Conceptual Common Sense Extraction

The knowledge in the knowledge bases such as Freebase, Knowledge Vault and so on are all facts which record the relationships between two entities. It may lead to following two problems. First, this form of knowledge limits the scale of the existing knowledge bases. When extracting new facts, no good patterns with a good ability of summarization can be used. Second, when applied in some real tasks, the knowledge may always suffer the problem of data sparsity. To solve these two problems, in this paper, we define the problem of extracting common senses in a concept level. We evaluate our solutions on Google N-Grams data set, and the results shows a great improvement.