Please see my comment to the question.
A solution with a good performance is possible via a hash table: http://en.wikipedia.org/wiki/Hash_table
This structure keeps data in key-value pairs, supports uniqueness of keys, and allows to quickly find a value by a key. The computational time complexity
of search is O(1), and is highly beneficial starting some considerable volume of data. (Please see http://en.wikipedia.org/wiki/Big_O_notation
You would need to have keys of the string type, to represent a word, and a value should represent a number of words found so far (use pointer to a number). You add words one-by one (first, you would need to split your string into words, another problem to solve), and try to add each word. If a value is found by some key, increment the word count by pointer. If not found, add a new key-value pair with value of 1 (a pointer to a value object). At the end, you will have all words each represented only once, and number of occurrences for each word. Don't forget to deallocate all the memory.
Perhaps, much simple structure with acceptable time complexity (O(log(n)) would be a binary tree: http://en.wikipedia.org/wiki/Binary_tree
The idea of using binary tree is pretty much the same.
With C, it's going to be a good volume of work, but it promises you unforgettable and useful experience, especially in patience.
You can help yourself if you find existing C implementation of a tree or a hash table, which should be easy.