for token_i in tokens:
    for token_j in tokens:
        compute_similarity(token_i, token_j)