IMPLEMENTASI METODE PROBABILISTIC LATENT SEMANTIC ANALYSIS UNTUK OPINION RETRIEVAL

Abstract

Opinion retrieval is a search system by the user, where in the information needed is more of opinion than a fact. The method used for opinion retrieval system is probabilistic latent semantic analysis (PLSA). The search proses opinion sentences with opinion retrieval system have some steps, the documents processed in text processing, should be form matrix value term. The PLSA method gives matrix values to calculate e-step, m-step, decomposition matrix, and likelihood value. The similarities calculating process in opinion sentence and query used cosine similarity formula. So it has similarity value as identified opinion sentence. The result of testing in document with 5 query words, it has highest score for kappa statistic testing 0.152432875 with kappa slight interpretation.