CoreNLP: Parsing of sentence failed, possibly because of out of memory

Question

While running Stanford's CoreNLP 3.7.0, I'm ocaisionally seeing this:

WARNING: Parsing of sentence failed, possibly because of out of memory.
Will ignore and continue: Just a year ago , the public outrage was over 
Sovaldi , a new hepatitis C drug being sold by Gilead Sciences for 
$ 1,000 a pill , or $ 84,000 for a course of treatment . 

I've seen this before, but that was when sentence splitting messed up, and gave a sentence that was very long. The cases I'm seeing now, like the one above, have reasonably sized, correctly split sentences.

Why might this happen, and what should I do to fix it?

Answer 1

Whilst this isn't exactly the answer to why this happens, I worked around this by using the 64-bit JRE and increasing the available heap. I'm not sure what environment and IDE you're using but to do this in Eclipse on Windows you have to do three things:

• Download the 64-bit JRE (choose Windows Offline / 64 bit)
• In Eclipse, click Window..Preferences..Java..Installed JREs then ensure that the newly installed JRE is selected (you will likely have to hit Search to find it first)
• Right-click your Java project, click Properties..Run/Debug Settings , select the launch configuration you're interested in then hit Edit . On the Arguments tab, under VM arguments enter -Xmx30G

This sets the maximum heap to 30Gb (I'm not sure it's restricted by physical memory as I don't have that much) and your project should run again. Be aware that those things it was falling over on before will likely be processed rather slowly.