Data Labelling

[viswajithiii]

As a first pass, we will do the following labelling.

Bookkeeping spreadsheet: here. Make sure to add to this every time you create a new article, creating a new id for it and putting in the link.

For each article, create a text file with name the article_id.txt with the following format:

Line 1: article_id (a001, a002 ...)
Line 2: URL
Line 3: Headline
Line 4: Byline (If multiple authors, separate by semicolon ("Poorna Kumar; Viswajith Venugopal"))
Line 5 onwards: Body

(For TSVs, article_id can be a001p, a002p, ... for Poorna's and a001v, a002v, ... for Viswa's.)

Now, the annotation is in a text file with name article_id.tsv, and is of the following format (one line per person mentioned):

Full Name, Gender, Number of times mentioned (only by part or full name, NOT by pronoun), Says something (yes/no), Number of words quoted, Source/subject (src/sub), Adjectives (a comma separated list), Expert/non-expert, Profession/Role(s)

UPDATE (Week of March 13th): As per Maneesh's instructions, I'm also adding a 'Quotes' column at the end of new articles I annotate. These contain the raw quotes that the person says. Different quotes that a person says are delimited by the special token ''.

[poorna-kumar]

List of small problems, to be looked at by Viswa if possible:

• a068: Article on Afghan cricket scene. Really not clear on who is a source versus a subject.
• a069: Article on American citizens detained in North Korea. Not clear about whether the detainees are subjects or neither source nor subject.
• a070: Article on Met's Opera House. Should we mark Verdi and Strauss as mentions?
• a096: Is Hughes a source? (I think yes). In that case, this is a good example of an article where the source is in the first paragraph. Also, in general, is Sean Spicer an expert source? In this article I have called him a source (debatable) and a non-expert (debatable).