In regards to computing lexical similarity, the two fundamental problems are respectively concerned with how to explore concept relationships predefined and enumerated in lexical knowledge bases and how to statistically induce and learn context relationships from word co-occurrences. To address these problems, this book focuses on approaching both taxonomic similarity through the semantic networks in WordNet and distributional similarity through syntactically constrained context. The taxonomic similarity model we proposed outperforms most popular similarity methods with respect to simulating human similarity judgments. In relation to distributional similarity, we thoroughly investigated the semantic properties of grammatical relationships in regulating word meanings, whereby over 80% precision can be reached in extracting synonyms or near-synonyms. This book provides a systematic guidance on computing taxonomic similarity and distributional similarity. It is appropriate for system developers and researchers working in language technology.