compute_PicaultRenault_scores: Compute scores using the Picault-Renault lexicon

In sentopics: Tools for Joint Sentiment and Topic Analysis of Textual Data

Compute scores using the Picault-Renault lexicon

Description

Computes Monetary Policy and Economic Condition scores using the Picault-Renault lexicon for central bank communication.

Usage

compute_PicaultRenault_scores(x, min_ngram = 2, return_dfm = FALSE)

Arguments

x	a quanteda::corpus object.
min_ngram	the minimum length of n-grams considered in the computation
return_dfm	if TRUE, returns the scaled word-per-document score under two quanteda::dfm, on for the Monetary Policy and one for the Economic Condition categories. If FALSE, returns the sum of all word scores per document.

Details

The computation is done on a per-document basis, such as each document is scored with a value between -1 and 1. This is relevant to the computation of the denominator of the score.

It is possible to compute the score for paragraphs and sentences for a quanteda::corpus segmented using quanteda::corpus_reshape. Segmenting a corpus using quanteda's helpers retain track to which document each paragraph/sentence belong. However, in that case, it is possible that paragraphs or sentences are scored outside the (-1,1) interval. In any case, the of the paragraph/sentences scores averaged over documents will be contained in the (-1,1) interval.

Value

A matrix with two columns, indicating respectively the MP (Monetary Policy) and EC (Economic Condition) scores of each document.

References

Picault, M. & Renault, T. (2017). Words are not all created equal: A new measure of ECB communication. Journal of International Money and Finance, 79, 136–156. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1016/j.jimonfin.2017.09.005")}

See Also

PicaultRenault

Examples

# on documents
docs <- quanteda::corpus_reshape(ECB_press_conferences, "documents")
compute_PicaultRenault_scores(docs)

# on paragraphs
compute_PicaultRenault_scores(ECB_press_conferences)

sentopics documentation built on Sept. 20, 2024, 5:06 p.m.