191 Commits (3cdb68b1965bd863b514c9d54f777e277b0be618)

Author	SHA1	Message	Date
Your Name	3cdb68b196	out_parser: updated fncols function	4 years ago
Your Name	dc40fbbb19	elasticizer: update rbindlist implementation	4 years ago
Your Name	18d47762d2	actor_merger: overhaul to include cutoffs at sentence level as intended, also included options to generate sentiment for text only (don't provide actors_meta or actor_groups)	4 years ago
Your Name	74909ca3a0	sentencizer: removed text sentiment computation from script, because of incorrect implementation	4 years ago
Your Name	c99ac23bb5	actor_merger: fixed absence of publication_date in some cases	4 years ago
Your Name	cc7fa5bffa	actor_merger: added aggregations of all individual actors and all party mentions in an article	4 years ago
Your Name	d9d578c06a	actor_merger: mult fix	4 years ago
Your Name	771145faf7	actor_merger: added mult='first' to metadata join for parties_actors to deal with duplicate partyIds (see 50Plus, Conservatives and Labour)	4 years ago
Your Name	1c14646e8f	actor_merger: dont deselect sent_words and sent_sum columns	4 years ago
Your Name	9bd382f955	actor_merger: fix to generate bogus sentiment columns	4 years ago
Your Name	b7f1afddd1	actor_merger: total rewrite based on data.table for performance reasons. Added some exceptions due to non-existing partyIds that some individual actors have in the actor database	4 years ago
Your Name	2c8a88f9a0	elasticizer: switched from bind_rows to rbindlist for composing result ... actor_merger: added noactor.* sentiment columns, and switched to data.table for matching actor metadata with articles	4 years ago
Your Name	559199bb97	sentencizer: totally removed sent_lemmas field	4 years ago
Your Name	36f2b341a8	sentencizer: removed derived output from function	4 years ago
Your Name	80ec0be1f8	actorizer: updated to account for token start offset in udpipe output. Sometimes, the first token in an article doesn't start at character position 1 (or 2 if the article starts with a whitespace), but at position 16 and possibly other positions.	4 years ago
Your Name	336567732c	elastic_update: added more debug output	4 years ago
Your Name	df7631b9f1	sentencizer: Changed output, removed lemma list and added separate positive and negative sentiment sums	4 years ago
Your Name	ecdb5be3b4	actorizer: moved some code	4 years ago
Your Name	69d4b6f5b0	actorizer: updated to data.table for conditional joins ... DESCRIPTION: added data.table dependency	4 years ago
Your Name	085855908c	query_gen_actors: switched from Minister to Min	4 years ago
Your Name	b406304c80	actorizer: Removed nested parallelization function ... query_gen_actors: Integrated startDate and endDate for parties, changed party exception method from abbreviation only to both full names and abbreviations for NL and BE	4 years ago
Your Name	5de4e1488c	estimator, modelizer, preproc: Removed experimental we-vector support, and disabled inefficiently implemented preproc.R	4 years ago
Your Name	77eb51a1bf	actorizer: totally revamped way of finding actors ... elasticizer: updated dump handling to create a dump for every batch, instead of one big file at the end out_parser: streamlined code query_gen_actors: only include relevant fields ud_update: changed function parameters to work with elasticizer dump function	4 years ago
Your Name	0e593075ee	query_gen_actors: only retrieve ud field from source	4 years ago
Your Name	6eb405f8bd	merger: selecting only relevant columns	4 years ago
Your Name	38ff4dcbf0	ud_update: small fix to file naming	4 years ago
Your Name	4b4d860235	class_update: remove dfm_gen multicore option ... dfm_gen: remove multicore, update merger() code elasticizer: changed filenaming scheme for dump option merger: Fixed bug where an NA lemma would cause the entire document to become NA. Now the NA lemmas are filtered out before merging ud_update: removed parallel processing, changed script to save bulk updates in .Rds files instead of sending them straight away	4 years ago
Your Name	5d99ec9509	elasticizer: added option to dump data frames to rds files ... out_parser: changed to single core, due to performance increase sentencizer: corrected documentation for sent_dict parameter	4 years ago
Your Name	aa6587b204	dupe_detect: fix for quotation marks	4 years ago
Your Name	2a220ded5d	dupe_detect: fix to query string for multi-word doctype names	4 years ago
Your Name	5bd36dcb44	dupe_detect: Changed query from json to query_string style, and added filter for already detected duplicates ... cv_generator: Changed code to use a generic vector of true values to draw the conditional random sample, instead of dfm/docvars specifically	4 years ago
Your Name	e499d70671	actor_merger: added ungroup() calls at the start and end of function, to speed up processing ... sentencizer: added ungroup() call at the end of the function to speed up processing	4 years ago
Your Name	8634d549a3	sentencizer: updates to collect sentence word counts and number of sentences also when no sent_dict is provided	4 years ago
Your Name	61e0581595	actor_merger: removed debug line	4 years ago
Your Name	f022312485	actor_merger: added function for generating actor-document data frames ... actor_fetcher: removed from package other: major update to documentation	4 years ago
Your Name	4e867214dd	sentencizer: commented code	4 years ago
Your Name	ec8afc4990	sentencizer: fixed actorsDetail coding error	4 years ago
Your Name	9ccfd2952e	sentencizer: minor updates	4 years ago
Your Name	98325bde8f	sentencizer: added new function for sentiment coding and actor collection	4 years ago
Your Name	7f958bbc11	actor_fetcher: small fixes	4 years ago
Your Name	8eedec8bb5	actor_fetcher: added option for using dictionaries with just lemmas, besides the option of using lemma_upos dictionaries	4 years ago
Your Name	057d225a7a	actor_fetcher: Allow generation of actor df containing only specified actor ids and aggregations	4 years ago
Your Name	9eae486a80	separated data preprocessing routines ... class_update: check if there are idf values associated with model, before applying weights estimator: make use of preproc() function for data preprocessing preproc: function containing all logic with regards to text data preprocessing and weighting	5 years ago
Your Name	a3b6e19646	revised modeling pipeline: ... cv_generator: generate folds for nested cv dfm_gen: added optional lowercasing parameter estimator: estimate model and performance based on parameters feat_select: select features based on textstat_keyness metric_gen: convert output from estimator to model performance metrics modelizer: updated for new pipeline modelizer_old: old model pipeline out_parser: now correctly exported	5 years ago
Your Name	e76a914dd2	actor_fetcher: Updated to tidyr 1.0.0, no longer using preserve, slightly different approach to keeping ids_list, and not removing actorsDetail anymore because it does not exist	5 years ago
Your Name	a01a53f105	class_update: added cores parameter for multicore processing of sources when using lemmas	5 years ago
Your Name	d9f936c566	modelizer: tf-idf application updated, final model now also includes idf values from training set, explicitly setting positive category in binary classification for confusion matrices, minor code fixes ... dfm_gen: added old junk codes for recoding, and removed deprecated ngrams parameter from dfm function class_update: removed dfm_words parameter, which is replaced by the force = T parameter in predict(), training/model idf is now applied to unseen data DESCRIPTION: added quanteda.textmodels as new dependency, since these have been separated from base quanteda 2.0.0 onwards	5 years ago
Erik de Vries	06bfec71bc	lemma_writer: unlist lemmas before writing	5 years ago
Erik de Vries	a83ee5dfd0	lemma_writer: update to write lemma instead of full document text	5 years ago
Erik de Vries	e594185719	dfm_gen: set default cores to 1	5 years ago