Economic Sentiment

Data & Analysis compendium for the Economic Sentiment analysis paper.

Article

View the article (open access).

Code

The main analysis code is located in the src/data-processing. Of interest might be:

• 10_apply_dictionaries.py Sentiment Dictionaries (Python)
• 11_apply_dictionaries_quanteda.R Sentiment Dictionaries (Quanteda)
• 20_svm.py Support Vector Machines
• 22_nb.py Naive Bayes
• 40_cnn_predict.py Convolutional Neural Network

Data

The following data files might be of interest:

• metadata.csv Headline, date, and source of each article
• gold.csv Gold standard (expert) coding
• crowdcodings.csv Crowd codings
• manual_coding.csv Manual (student) coding of headlines

See the files in src/data-processing for details on how these files were constructed.

Results as presented in the article

• Performance, learning curve, correlations

Online Appendix / Supplementary Results

• Crowd coding instructions
• Confusion Matrices of all methods
• Grid Search for CNN (code: 30_cnn_gridsearch.py
• Grid Search for SVM (code: 19_svm_gridsearch.py
• Validation of ML results against student codings
• Error analysis