In this paper, we present how table memorized semiring structure contributes in the vector space model in information retrieval system. We implement this new structure by generating table for each document and the first row filled with key word and the second row filled with weight for each word. This new structure implements using 242 Arabic documents which were presented in the Saudi Arabian National Computer Conference. The new method (technique) shows a new result which is more efficient than traditional structure and can save space. The results also show that when we used traditional structure system it occupies 204248 × units to implement vectors, but in new data structure system it occupies 5388×unit which means that we saved more than 198860 space units.