The Document Similarity Index based on the Jaccard Distance for Mail Filtering
Temma Seiya, Manabu Sugii, Hiroshi Matsuno · 2019 34th International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) · 2019
We propose a new index of similarity for classification of emails into ham and spam ones with the Jaccard index. It takes advantage of co-occurrence value of all pairs of two words in emails. The co-occurrence of words represents a sort of context in documents because a word is often in use with another word in the same context. Our proposed method classified emails into hams or spams with high accuracy rate than the present filtering system using appearance frequency of word. Our method could extract patterns of word usage reflecting the context of emails.