Handling overlaps in spoken term detection