To have full language support for any website the data must be stored in both languages instead of automatically translated. Decentralized storage systems are at their infancy and Solana is not optimized for READ operations. It seems like providing multilingual data for the metadata of each NFT will inflate the amount of data about said NFT. You can't store two versions of the same NFT.
Is there a better way to start with supporting two languages for an NFT collection based on the metaplex metadata standard?
The easiest and the fastest way to store multilingual data for an NFT on Solana would be to store it off-chain using the external_url field, on the on-chain metadata.
According to the (non-fungible) Token Metadata Standard by Metaplex: https://docs.metaplex.com/programs/token-metadata/token-standard#the-non-fungible-standard, this URI points to an external URL defining the asset.
This off-chain JSON could then store any number of language conversions of the metadata of your NFT and can easily (and quickly) be queried as the data remains off-chain.