The Stanford Parser is a probabilistic parser, structured around pre-programmed data sets. These “hand-parsed” sentences become the lexical basis for sentence analysis. The parser assigns tags to each element in the sentence. Then, drawing from the knowledge base, it assigns values to each element according to its estimated function within the sentence (e.g. nouns, verbs, direct objects, etc.).

The most recent version of the Stanford Parser employs enhanced universal dependencies. Enhanced universal dependencies (UD) code implicit relationships between words using augmented names, making their relationship explicit. The inclusion of the enhanced UD allows for greater output accuracy by emphasizing the relationship between context words, what is called natural language understanding (NLU), over their respective places in the tree structure. You can experiment with the Stanford Parser by downloading the Java program from their website.