To take the use of natural language part of speech tagging in text information hiding research, the order relation and its transforming property were studied. A proper partial relation was defined in the collection of part of speech tagging firstly，and then a tagging string could be reflected as 0 and 1 by calculating the odd and even characteristic of the inverse order of it. Research shows that the corresponding binary sequence of the casually selected tagging strings had a better self-relation and co-relation property whose normalized relation coefficients were round 0.55, it sounds reasonable to distinguish various templates, so a new method by using the odd and even characteristic of the inverse order to hide information was proposed. And the inverse order has also been proved to have an odd and even characteristic under the condition of being exchanged, added and deleted. This method is robust against the attacking of synonyms replacement.
