A proposed parallel model for extracting and translating query keywords for bilingual information retrieval

In this work, a parallel model is proposed to analyze an input English query and translate the English keywords; word-by-word; into Arabic. The proposed model involves three main parts: the architectural model, the application model, and the cost model. In the architectural model, a multiprocessor system based on shared memory is adopted. The communication among processors is done via the shared memory. Concerning the application model, four important phases are done to translate the English extracted keywords into Arabic, and form two patterns one in English and the other in Arabic. Once the English and Arabic patterns are formed, they will be searched against the terms describing the collection of English and Arabic documents respectively, i.e. the patterns are prepared to be matched against the Web documents which is not our objective here. The proposed model does searching and also computation of the similarity value in parallel. The parallel model is going to decrease the time consumed in searching and computation. Moreover, the cost model calculates the total execution time of the query analysis and keywords´ translation. That execution time is calculated by using one processor and more than one processor. The speedup is also calculated as well as the processors´ utilization. This work was simulated on the IBM/PC with speed of 800 MHZ. The simulation was done for different query lengths from one to six words. It is concluded that, parallelism is significant for reducing the time consumed in extracting and translating the query keywords for bilingual information retrieval.