#include <bits/stdc++.h>

#include <variant>

#define rep2(i, k, n) for (i64 i = (i64)(k); i < (i64)(n); i++)
#define rep(i, n) rep2(i, 0, n)
#define all(x) begin(x), end(x)
#ifdef ENV_LOCAL
#define dump \
  if (1) cerr
#else
#define dump \
  if (0) cerr
#endif

using namespace std;
using namespace std::string_literals;

using i32 = int32_t;
using i64 = int64_t;
using f64 = double;
using f80 = long double;
using vi32 = vector<i32>;
using vi64 = vector<i64>;

/*
 * harudake::modint
 *
 * Copyright (c) 2021 prime number
 *
 * This software is released under the MIT license.
 * see https://opensource.org/licenses/MIT
 *
 */
namespace harudake {

template <typename base_t, base_t MOD>
class static_modint;

using modint_1000000007 = static_modint<uint64_t, 1'000'000'007>;
using modint_998244353 = static_modint<uint64_t, 998'244'353>;

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> add(const static_modint<base_t, MOD> lhs,
                               const static_modint<base_t, MOD> rhs) {
  base_t tmp = lhs.val + rhs.val;
  if (tmp >= MOD) tmp -= MOD;
  return static_modint<base_t, MOD>::make_raw(tmp);
}

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> sub(const static_modint<base_t, MOD> lhs,
                               const static_modint<base_t, MOD> rhs) {
  base_t tmp = lhs.val + MOD - rhs.val;
  if (tmp >= MOD) tmp -= MOD;
  return static_modint<base_t, MOD>::make_raw(tmp);
}

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> mul(const static_modint<base_t, MOD> lhs,
                               const static_modint<base_t, MOD> rhs) {
  base_t tmp = lhs.val * rhs.val;
  tmp %= MOD;
  return static_modint<base_t, MOD>::make_raw(tmp);
}

// data types must be signed integer
inline int64_t inv(const int64_t a, const int64_t p) {
  return (a == 1 ? 1 : (1 - p * inv(p % a, a)) / a + p);
};

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> div(const static_modint<base_t, MOD> lhs,
                               const static_modint<base_t, MOD> rhs) {
  base_t tmp = lhs.val * inv(rhs.val, MOD);
  tmp %= MOD;
  return static_modint<base_t, MOD>::make_raw(tmp);
}

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> pow(const static_modint<base_t, MOD> base,
                               const uint64_t index) {
  if (index == 0) return static_modint<base_t, MOD>{1};
  if (index % 2 == 1) {
    return pow(base, index - 1) * base;
  } else {
    auto half = pow(base, index / 2);
    return half * half;
  }
}

template <typename base_t, base_t MOD>
class static_modint {
  using modint = static_modint;

 public:
  static constexpr base_t mod = MOD;
  static_modint() : val(0) {}
  static_modint(const base_t x) : val(x % MOD) {}
  static_modint(const modint&) = default;
  modint& operator=(const modint&) = default;
  modint& operator=(const base_t x) {
    val = x % MOD;
    return *this;
  }
  modint operator-() const { return val == 0 ? 0 : MOD - val; }
  explicit operator base_t() const { return val; }
  base_t get() const { return val; }
  constexpr base_t get_mod() const { return mod; }
  static modint make_raw(const base_t raw) {
    modint res;
    res.val = raw;
    return res;
  }

 private:
  friend modint add<base_t, MOD>(const modint, const modint);
  friend modint sub<base_t, MOD>(const modint, const modint);
  friend modint mul<base_t, MOD>(const modint, const modint);
  friend modint div<base_t, MOD>(const modint, const modint);
  base_t val;
};

template <typename base_t, base_t MOD>
static_modint<base_t, MOD> operator+(const static_modint<base_t, MOD> lhs,
                                     const static_modint<base_t, MOD> rhs) {
  return add(lhs, rhs);
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD> operator-(const static_modint<base_t, MOD> lhs,
                                     const static_modint<base_t, MOD> rhs) {
  return sub(lhs, rhs);
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD> operator*(const static_modint<base_t, MOD> lhs,
                                     const static_modint<base_t, MOD> rhs) {
  return mul(lhs, rhs);
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD> operator/(const static_modint<base_t, MOD> lhs,
                                     const static_modint<base_t, MOD> rhs) {
  return div(lhs, rhs);
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD> operator^(const static_modint<base_t, MOD> lhs,
                                     const uint64_t rhs) {
  return pow(lhs, rhs);
}

template <typename base_t, base_t MOD>
static_modint<base_t, MOD>& operator+=(static_modint<base_t, MOD>& lhs,
                                       const static_modint<base_t, MOD> rhs) {
  return lhs = lhs + rhs;
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD>& operator-=(static_modint<base_t, MOD>& lhs,
                                       const static_modint<base_t, MOD> rhs) {
  return lhs = lhs - rhs;
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD>& operator*=(static_modint<base_t, MOD>& lhs,
                                       const static_modint<base_t, MOD> rhs) {
  return lhs = lhs * rhs;
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD>& operator/=(static_modint<base_t, MOD>& lhs,
                                       const static_modint<base_t, MOD> rhs) {
  return lhs = lhs / rhs;
}
template <typename base_t, base_t MOD>
static_modint<base_t, MOD>& operator^=(static_modint<base_t, MOD>& lhs,
                                       const uint64_t rhs) {
  return lhs = lhs ^ rhs;
}

}  // namespace harudake
using namespace harudake;
using modint = modint_998244353;

int main() {
  ios::sync_with_stdio(false);
  cin.tie(nullptr);
  i64 n, m, k;
  cin >> n >> m >> k;
  string s;
  cin >> s;
  modint ans(0);
  set<char> cs;
  vector<modint> dp(27);
  rep(i, m) {
    vector<modint> nxt(27);
    if (i < n) {
      rep2(ch, 'a', s[i]) {
        i64 idx = ch - 'a';
        if (cs.count(ch)) {
          nxt[size(cs)] += modint(1);
        } else {
          nxt[size(cs) + 1] += modint(1);
        }
      }
      cs.insert(s[i]);
      if (i < n - 1 && size(cs) >= k) {
        ans += modint(1);
      }
    }
    rep(j, 27) {
      nxt[j] += dp[j] * modint(j);
      if (j < 26) {
        nxt[j + 1] += dp[j] * modint(26 - j);
      }
    }
    rep(j, 27) {
      if (j >= k) {
        ans += nxt[j];
      }
    }
    swap(dp, nxt);
  }
  cout << ans.get() << endl;
  return 0;
}