A research team is pre-training a language model with the primary objective of making it highly robust to common typographical errors found in real-world text. They are debating between two methods for damaging the input sentences during training:

• Method 1: A portion of the words in each input sentence are completely removed and substituted with a generic placeholder symbol. The model must then predict the original words that belong in those placeholder positions.
• Method 2: A portion of the words in each input sentence are replaced with different, randomly selected words from the vocabulary. The model must then learn to identify these incorrect words and restore the original sentence.

Evaluate which of these two methods is more suitable for achieving the team's specific goal. Justify your choice by explaining the strengths and weaknesses of each method in the context of learning to handle typographical errors.