convert categorical data into bounded integers using hash E.g., author names "Polar" for O, "Fadi420" for 1, etc. E.g., "worldnews" for 0, etc. embedding news topics: you find all words in all topics, and convert each topic into a bit vector (containing 0 or 1, 0 mean the topic doesn't have the word, and 1 means the topic has the word