从itu告中提取死名实体-NLP

Question

I have a continuous strings of ads , which are extracted from some newspaper. The ads may appear in a format as shown below:My task here is to extract the deceased person's names.

John, the small son of Mr. and Mrs.<br>
Elmer Cleppfer, died at their home in<br>
Lewistown on Wednesday. The funeral<br>
will He held on Saturday afternoon<br>
from the home of the grandparents<br>
on the child, Mr. and Mrs. John<br>
Kiopper, 224 Locust street, tortiorrow<br>
afternoon at 2 o'clock. Interment witt<br>
take place at Oberlin.<br>

Mrs. Lydia Mintch, aged 6S years <br>
died yesterday afternoon at the home<br>
of Fred Flowerfleld at Enhaut. Mrs.<br>
Mlnlch contracted a severe attack of<br>
pneumonia aggravated by other illness<br>
Several days ago which resulted in her<br>
death. Funeral arrangements have not<br>
yet been completed.<br>

The whole of the para is made up of 2 ads.. Can any one tell me how to classify such kind of text into paragraphs if there are more than 1 such ads?

Answer 1

Well Stanford Parser is your option here.

1. First extract only the sentences that contain died, decreased or something on these terms.
2. Generate collapsed typed dependencies for these sentences using Stanford Parser.
3. You will find a pattern which will help you get the name of the decreased person.

I am intentionally not giving away the pattern here as you should put in your efforts as well.

Answer 2

Here is how I would approach the problem.

1. Get sentences POS tagged.
2. For each sentence, deep parse and create subject-verb-object model. (Left to right parse).
3. Where ever the verb points to death, the subject is the dead person.