cleanNLP package in R

cnlp_get_dependency (obj, get_token = FALSE) and cnlp_get_dependency (obj, get_token = TRUE) gives the right result with word and target word with R 3.3. and Stanford CoreNLP. I do not remember the exact version 3.3.X.

I have recently upgraded to version 3.5.3 and had to install the external packages. Now, the same code gives a completely useless answer for get_token = TRUE.

for example.
with get_token = TRUE used before to get a data frame with 3K lines. With the latest update R, I have 3 million lines. The results look good with get_token = FALSE for both versions.

Has anyone encountered this situation?