These instructions work in a bash shell on a MacOs 10.13.2, with java 10.0.1. Should work on any Linux, with a recent java.
Ensure you have Stanford CoreNLP jars downloaded, say in dir /path/to/corenlp. This should have many jars, including stanford-corenlp-3.8.0.jar.
Then:
$ javac -cp ".:/path/to/corenlp/stanford-corenlp-3.8.0.jar" Main.java
This should run successfully, produces no output. After this your current directory should have Main.class.
To run, the system will need more jar files from the corenlp directory. So you want to tell the system where to find all these jars:
$ java -cp ".:/path/to/corenlp/*" Main
The Glue Semantics Workbench
copyright 2018 Moritz Messmer & Mark-Matthias Zymla
Starting interactive dependency mode...
Enter sentence to be analyzed or enter 'quit' to exit the program.
Now entering a simple sentence like John smiles will give you an adequate response:
John smiles.
[main] INFO edu.stanford.nlp.parser.lexparser.LexicalizedParser - Loading parser from serialized file edu/stanford/nlp/models/lexparser/englishPCFG.ser.gz ... done [0.6 sec].
[nsubj(smiles-2, John-1), root(ROOT-0, smiles-2)]
John/NNP This is a subject
smiles/VBZ has arity 1
[g : j[0], (g ⊸ f) : λx_e.smiles(x)[1]] => null
Searching for valid proofs...
Agenda: [g : j[0], (g ⊸ f) : λx_e.smiles(x)[1]]
Combining premises (g ⊸ f) : λx_e.smiles(x)[1] and g : j[0]
-->f : smiles(j)[0, 1]
Found valid deduction(s):
f : smiles(j)[0, 1]
Done!
Enter sentence to be analyzed or enter 'quit' to exit the program.