This image illustrates the transformation of word relationships in the sentence “What is the capital of India?” before and after passing through a Feed Forward Network (FFN). On the left side (“Before FFN”), words have weak interconnections and are roughly annotated. On the right side (“After FFN”), the FFN strengthens semantic links and assigns more precise roles, such as “Question Word,” “City’s Admin Center,” and “Country Name.” The diagram highlights how FFNs improve contextual understanding in language models.