Existing short text similarity approaches discard syntactic information associated with words. In this paper, we propose Semantic-Syntactic Short text similarity Approach based on Keyword extraction and Syntactical Scaled Mapping Method for (SAG) System. First Phase, the Statistical Scaled Syntactical Relation (SSSR) Model based on the Link Parser (LP) uses statistical to Scale SR in short answer structure. Second phase: proposed Key Words Extraction and Syntactic Relation Mapping (WESRM) Approach to extract main words from each ESC and assign the SSR to each extracted word. Finally, Semantic-Syntactical Matrix (SSM) to integrate the scaled syntactic relations with WUP knowledge-based model to grade SA. To evaluate our system, Difference Measurement…show more content… 2011). The proposed method is combining syntactic and semantic information to decide the semantic resemblance. The syntactic information attained through bottom-up projective parser with pseudo-projective transformations. Then, a syntactic model based on the statistical analysis used to obtain a list of candidate syntactic. The method computed the semantic similarity between words under same syntactic functions in both sentences. LIN Knowledge-based model with WordNet used in the semantic phase. The combined syntactic and semantics scaled with constants assigned scales to verbs, subjects, objects, and adverbial. These scales assigned to the major syntactic arguments cannot reflect the actual importance of the syntactic structure. Similarly, analogous scales such as a semantic glossary, as an example, “I have a dog” and “I have a cat”, where both words cat and dog treated semantically the…show more content… The relations calculated based on the investigating of particular words encapsulated with their associated syntactic relations. SSSR counts the word appears in other encapsulation to determine the number of syntactical relation. This is the contrast to other algorithms which relies upon a fixed scale. The algorithm takes all syntactical relation to count to scale it actually based on its appearance in the structure. The following pseudo-code clarifies how the method works automatically to calculate syntactical scale for each relation (SR_n) in each encapsulated syntactical component (