(cluttered label) How to avoid wrong key pairs value (2 field in 1 line)

I experienced having wrong key pairs value result (2 field in 1 line) and I found article in github explaining: Avoid cluttered labels Sometimes when you apply different labels within the same line of text, the service may merge those labels into one field. For example, in an address, you might label the city, state, and zip code as different fields, but during prediction those fields are not recognized separately. We understand this scenario is essential for our customers, and we are working on improving this in the future. Currently, we recommend our users to label multiple cluttered fields as one field, and then separate the terms in a post-processing of the extraction results. Is that true? how can I improve the result of 2 field in 1 line? Your response is highly appreciated. Thank you