Term suggestion is a very useful information retrieval technique that tries to suggest relevant terms for users' queries, to help advertisers find more appropriate terms relevant to their target market. This paper aims to focus on the problem of using several semantic analysis methods to implement a term suggestion system.
Three semantic analysis techniques are adopted – latent semantic indexing (LSI), probabilistic latent semantic indexing (PLSI), and a keyword relationship graph (KRG) – to implement a term suggestion system.
This paper shows that using multiple semantic analysis techniques can give significant performance improvements.
The suggested terms returned from the system may be out of date, since the system uses a batch processing mode to update the training parameter.
The paper shows that the benefit of the techniques is to overcome the problems of synonymy and polysemy over the information retrieval field, by using a vector space model. Moreover, an intelligent stopping strategy is proposed to save the required number of iterations for probabilistic latent semantic indexing.
