In skip-gram models, what does the window size determine?

In skip-gram models, what does the window size determine?

• A. The number of surrounding words considered as context.
• B. The maximum length of a sentence.
• C. The number of surrounding words used as context.
• D. The dimensionality of the word vectors.

Answer: (C)

The window size sets how many surrounding words are used as context for predicting the target word in a skip-gram model. In training, the model tries to predict each context word from the center word, and the window size defines the range of words around the center that count as context. This means, for any target word, you can have up to 2 times the window size context words (left and right), though the exact number can be smaller at the start or end of a sentence. It does not affect how long the sentence is or the dimensionality of the word vectors. A larger window brings in more distant context and tends to capture broader semantic relationships, while a smaller window focuses on nearby, often more syntactic relationships.

The window size sets how many surrounding words are used as context for predicting the target word in a skip-gram model. In training, the model tries to predict each context word from the center word, and the window size defines the range of words around the center that count as context. This means, for any target word, you can have up to 2 times the window size context words (left and right), though the exact number can be smaller at the start or end of a sentence. It does not affect how long the sentence is or the dimensionality of the word vectors. A larger window brings in more distant context and tends to capture broader semantic relationships, while a smaller window focuses on nearby, often more syntactic relationships.