Adapting StanfordCoreNLP to process noisy web text?

Question

I've been trying out the StanfordCoreNLP NER and everything manually on the website, and it seems they depend on very specific/proper English cues to detect entities, for example. When dealing with web text, though, where you could have some text like

---

John Doe

Assistant Professor of Computer Science

Stanford University

---

StanfordNLP seems to have some trouble (labeling the whole thing as one organization due to lack of prepositions/punctuation). Is there anything I can do to allow NER to better handle this kind of text (e.g. program some pre-processing of text)?

Answer 1

Adding dot(.) at the end of each line gives better results. (Since sentence splitter uses dot as delimeter)