Weighing scheme for Term-Term matrices

  • Weight contexts according to the distance from the word → further away, lower the weight
  • For a window size ±k, multiply the context word’s frequency at each position by (k−distance)/k
    • distance of word wj from wt = t - j -1