SP500207 Query Improvement in INformation Retrieval Using Genetic Algorithms - A Report on the Experiments of the TREC Project chapter J. Yang R. Korfhage E. Rasmussen National Institute of Standards and Technology Donna K. Harman (5) Precision and recall The average performance of the current genetic algorithm can be shown using precision and recall values. Here we use data provided by NIST to show the results from the ad hoc queries and the routing queries. Note that in our experiment the ad hoc queries were used only on the second dataset (disk two). Table 19 and Table 20 display the average precision at the 11 recall levels from both the ad hoc queries and the routing queries, respectively. 8. Discussion The experimental results on the TREC document collection have shown that the genetic approach in query modification can be applied in large and varied databases. Evidence has indicated that the genetic query modification leads to the convergence of the query term weights. The results have also shown that additional relevant documents can be retrieved by modifying the query term weights using the genetic approach. We have also shown, through specific examples, the effects of crossover and mutation, and how variations in query term weights affect the retrieval of relevant documents. We believe that the genetic algorithm technique introduces new information into the retrieval process, allowing the system to do a better job of query modification. Evidence has shown that even without the use of genetic techniques our system employs parallel search, although the genetic modification brings additional relevant documents. For the system designer interested in Boolean techniques this permits exploration of several Boolean variants simultaneously, and also facilitates use of Boolean combinations of the individual query variants. Table 15. Term weights of query individuals of topic 24, the second generation (generation 1) ti t2 t3 t4 t5 t6 t7 t8 t9 tio til t12 t13 t14 tistl6 t17 tig t19 t20 t21 0.18 .31 .53.95 .17.70.23 .49.12.08 .39.28 .37.96.38 .48.69.77.78.82.15 1 .60.24.45.79.08.48.15 .25.94.61 .99.48.80.76.54.77.49 .10.59.35.14 2.49.61 .42.13.26 .04.98 .11 .38 .65 .35 .55 .36.57.48.16.74.57.55 .29.87 3 .18.31 .53.95 .28 .70.23.49.12.08 .39.28 .37.98 .54.77.53.37.78.82.15 4.49.61 .42.13.26 .04.67.25.94.61 .99.48.80.75.48.16.62 .17.55.29.87 5.60.24.45 .79.08.48.47.11 .38 .65 .35 .55 .36.55 .38.48 .53.10.59.35.14 6.60.24.45.79.08.48.15 .25.94.61 .79.28.37.98.54.77.65 .77.78.69.14 7.18.31 .53.95.17.70.23.49.12 .08.59.48.80.74.38.48 .53.10.59.48.98 8.18.31 .53 .95 .17.70.23 .49.12.08 .39.28 .37.98 .54.77.65 .77.78.82.15 9.18 .31 .53.95 .17.70.23 .49.12.08 .39.28 .37 .98 .54.77 .65 .77.78.82.15 50