Umbrella ticket: Output Format and Exploration

[dcsw2]

Add requests here for research requirements

[kmcdono2]

Two core approaches to working with the toponyms:

1. analyzing toponyms independently --> see discussion of this as viz in #166 This could provide sample-level information such as:
   • average #/min/max of toponyms per article and per title
   • spatial "center" of resolved toponyms (e.g. where do they cluster) per article and per title
   • how many toponyms of each type (location, building, street, etc.) in total sample

And, for every resolved and unresolved toponym in the corpus/sample:

• NER result, type, and confidence score
• candidates for each toponym from ranker + conf score
• confidence score for linking/resolution
• candidates from linker + conf score --> #168
• be sure to include toponyms that could not be resolved in output

2. analyzing toponyms dependent on their linguistic context Basic format would allow us to analyze toponyms within the context of other words in the article. Literally what format this should be (e.g. XML) is what I'm not sure about. For the paper we are working on, this would ideally include (not a complete list, but a start):
   • access to POS tags --> answer is to run sentences with nltk POS tagger --> documented in #167
   • the ability to analyze text within a window to either side of any toponym

[kallewesterling]

I'd be happy to talk through the format question, if you want, @kmcdono2!

[dcsw2]

I sent an invite

[kallewesterling]

Thank you @dcsw2 - looking forward to chatting!

[kmcdono2]

Decision to divide this ticket into separate tasks. #166 for the items under (1). Next week we will discuss how to prepare for (2).

[fedenanni]

Hi all - I am picking up the "access to POS tags" and will keep track of progress in here