class FullyIndexableDictionary():

    def __init__(self, size, bit = None):
        self.size = size
        self.block = (size + 31) >> 5
        self.b = [0] * self.block
        self.s = [0] * self.block
        if bit is not None:
            for i in range(self.size-1, -1, -1):
                if bit & 1:
                    self.set(i)
                bit >>= 1
                self.build()

    def set(self, p):
        self.b[p >> 5] |= 1 << (p & 31)

    def build(self):
        for i in range(1, self.block):
            self.s[i] = self.s[i - 1] + self.popcount(self.b[i - 1])

    def popcount(self, x):
        x = x - ((x >> 1) & 0x55555555)
        x = (x & 0x33333333) + ((x >> 2) & 0x33333333)
        x = (x + (x >> 4)) & 0x0f0f0f0f
        x = x + (x >> 8)
        x = x + (x >> 16)
        return x & 0x0000007f

    def access(self, p):
        return (self.b[p >> 5] >> (p & 31)) & 1

    def _rank(self, p):  # [0, p)
        mask = (1 << (p & 31)) - 1
        return self.s[p >> 5] + self.popcount(self.b[p >> 5] & mask)

    def rank(self, p, b):
        if b:
            return self._rank(p)
        else:
            return p - self._rank(p)

    def select(self, x): # not verified
        lb, ub = 0, self.block
        while ub - lb > 1:
            mid = (ub + lb) // 2
            if self.s[mid] <= x:
                lb = mid
            else:
                ub = mid
        b_id = lb
        lb = b_id * 32
        ub = (b_id + 1) * 32
        while ub - lb > 1:
            mid = (ub + lb) // 2
            if self.rank(mid) <= x:
                lb = mid
            else:
                ub = mid
        return lb

class WaveletMatrix:
    def __init__(self, lis, MAXLOG=32):
        self.N = len(lis)
        self.mat = list()
        self.zs = list()
        self.MAXLOG = MAXLOG
        self.build(lis)

    def build(self, _lis):
        lis = _lis[:]
        for dep in range(self.MAXLOG-1, -1, -1):
            ls = list()
            rs = list()
            fid = FullyIndexableDictionary(self.N + 1)
            for i in range(self.N):
                if lis[i] & (1 << dep):
                    rs.append(lis[i])
                    fid.set(i)
                else:
                    ls.append(lis[i])
            fid.build()
            self.mat.append(fid)
            self.zs.append(len(ls))
            lis = ls + rs
        return

    def rank(self, r, x):
        """ count number of occurrences of x in [0,r) """
        return self._rank(0, r, x)

    def rangemax_k(self, l, r, k):
        """ k-th largest value in [l, r) (k in 1-indexed)"""
        ret = 0
        for dep in range(self.MAXLOG):
            ret <<= 1
            cntl = self.mat[dep].rank(l, 1)
            cntr = self.mat[dep].rank(r, 1)
            if cntr - cntl >= k:
                l = cntl + self.zs[dep]
                r = cntr + self.zs[dep]
                ret |= 1
            else:
                l = l - cntl
                r = r - cntr
                k = k - (cntr - cntl)
        return ret

    def rangemin_k(self, l, r, k):
        """ k-th smallest value in [l, r) (k in 1-indexed)"""
        return self.rangemax_k(l,r,r-l-k+1)

    def rangefreq(self, l, r, x, y):
        """ count number of occurrences of [x,y) in [l,r) """
        return self._freq_dfs(0, l, r, 0, x, y)

    def rangelist(self, l, r, x, y):
        """ count number of occurrences of each x in [x,y) in [l,r) """
        return self._list_dfs(0, l, r, 0, x, y)

    def rangemax(self, l, r):
        ret = 0
        for dep in range(self.MAXLOG):
            ret <<= 1
            cntl = self.mat[dep].rank(l, 1)
            cntr = self.mat[dep].rank(r, 1)
            if cntl == cntr:
                l = l - cntl
                r = r - cntr
            else:
                l = cntl + self.zs[dep]
                r = cntr + self.zs[dep]
                ret |= 1
        return ret

    def rangemin(self, l, r):
        ret = 0
        for dep in range(self.MAXLOG):
            ret <<= 1
            cntl = self.mat[dep].rank(l, 0)
            cntr = self.mat[dep].rank(r, 0)
            if cntl == cntr:
                l = l - cntl + self.zs[dep]
                r = r - cntr + self.zs[dep]
                ret |= 1
            else:
                l = cntl
                r = cntr
        return ret

    def _rank(self, l, r, x):
        for dep in range(self.MAXLOG):
            bit = (x >> (self.MAXLOG - (dep + 1))) & 1
            l = self.mat[dep].rank(l, bit) + self.zs[dep] * bit
            r = self.mat[dep].rank(r, bit) + self.zs[dep] * bit
        return r - l

    def _freq_dfs(self, dep, l, r, val, a, b):
        if l == r:
            return 0
        if dep == self.MAXLOG:
            return r - l if a <= val < b else 0
        mid = 1 << (self.MAXLOG - dep - 1) | val
        right = (1 << (self.MAXLOG - dep - 1)) - 1 | mid
        if right < a or b <= val:
            return 0
        if a <= val and mid < b:
            return r - l
        cntl = self.mat[dep].rank(l, 1)
        cntr = self.mat[dep].rank(r, 1)
        return self._freq_dfs(dep + 1, l - cntl, r - cntr, val, a, b) + \
        self._freq_dfs(dep + 1, cntl + self.zs[dep], cntr + self.zs[dep], mid, a, b)

    def _list_dfs(self, dep, l, r, val, a, b):
        vs = list()
        if l == r or b <= val:
            return vs
        if dep == self.MAXLOG:
            if a <= val < b:
                vs.append((val, r - l))
            return vs
        mid = 1 << (self.MAXLOG - dep - 1) | val
        right = (1 << (self.MAXLOG - dep - 1)) - 1 | mid
        if right < a:
            return vs
        cntl = self.mat[dep].rank(l, 1)
        cntr = self.mat[dep].rank(r, 1)
        vs.extend(self._list_dfs(dep + 1, l - cntl, r - cntr, val, a, b))
        vs.extend(self._list_dfs(dep + 1, cntl + self.zs[dep], cntr + self.zs[dep], mid, a, b))
        return vs


###############################################################################

N,Q=map(int, input().split())
P=[]
for s in input().rstrip():
    P.append(ord(s)-ord("a"))

WM=WaveletMatrix(P)

for _ in range(Q):
    l,r,x=map(int, input().split())
    p=WM.rangemin_k(l-1,r,x)
    print(chr(p+ord("a")))