Accurate trust evaluation is crucial for the success of Electronic-commerce systems. Position-based trust representations are mostly used and user’s assessment rankings are combined to find out vendor’s position. One major problem with this is that identifying whether everything is good with the vendor or not and it is tough for customers to find vendors who are believable, because rankings of vendors in present system is normally high for each vendor. In this paper, by observing the remarks of the customers, which were expressed freely in the form of text, Comment-based multi-dimensional trust model, is a fine-grained multi-dimensional trust evaluation model by extracting user’s reply statements in electronic-commerce, and a procedure to extract user’s remarks for dimension rankings is proposed. With this model comprehensive trust profiles are computed automatically for vendors, including dimension trust scores and weights, as well as overall trust ratings by combining dimension positional rankings. This model provides an approach that combines dependency relation analysis and lexicon-based opinion mining techniques and further proposes an algorithm based on dependency relation analysis and Latent Dirichlet Allocation (LDA) topic modeling technique to cluster aspect expressions into dimensions and compute combined dimension rankings and weights. This algorithm is named as Lexical-LDA. This model can capably address the problem of identifying the good vendor and rank vendors efficiently.