As @Steven Jeuris has said, the phenomena is best known as semantic satiation. It's not as popular a topic of study as it used to be (most references I can find for it come from the 1960s), but, to the best of my memory, the actual cause of the phenomena is down to how meaning is represented in the brain.
I'll explain this by loosely paraphrasing an account of the phenomena in Michael Spivey's book, The Continuity of Mind, but there are some (interesting) theoretical claims in that book which I won't bother with for the scope of this answer.
The general consensus on neural representations of meaning is that semantic information (i.e. the meaning of a word, although this can include much more than that) is encoded as a pattern of activation across many neurons. Exposure any given word (let's use giraffe as our example, following Spivey) causes that specific pattern to become partially activated, and so we experience the "meaning" of the word - the more accurately the pattern is recreated, the more intense our conscious experience of the meaning.
However, individual neurons are subject to adaptation, or fatigue (Wikipedia): if a particular neuron has been firing for a prolonged period of time, it begins to weaken, and it's firing rate slows, until it has time to recover (think of a drummer's arm muscles, performing the same action repeatedly until he tires out). When you repeat a specific word (giraffe), the neurons used in the corresponding pattern become fatigued in this way, and so the corresponding meaning ("long-necked African mammal") fades, or is replaced by other, possibly related meanings.
