NEWS.md

In GermaParl: Download and Augment the Corpus of Plenary Protocols of the German Bundestag

GermaParl 1.5.2

• The germaparl_download_lda() will check md5 sums now when downloading data.
• If downloading an LDA model using germaparl_download_lda() fails, you will now see an informative message and the return value will be FALSE.

GermaParl 1.5.1

• Functions included in older versions of the package that used functions from the RcppCWB package had been dropped. An unnecessary declaration of RcppCWB in the 'Imports:' section of the DESCRIPTION file has been removed.
• The data objects germaparl_by_lp and germaparl_by_year were included as data.table objects, making the presence of the data.table package necessary. To reduce the number of packages imported from and to avoid an error that emerged on Windows, these tables are included as data.frame objects.
• The documentation of the data objects germaparl_by_lp and germaparl_by_year now includes an explanation of what is reported in rows and columns.
• The germaparl_by_year table now includes a column unknown_total and unknown_share with the total number of tokens that cannot be lemmatized, and their share, respectively. On this basis, an error in the calculation of the aggregate unknown share for all years can be corrected.

GermaParl 1.5.0

• The package will not depend on the polmineR package any more. Higher-level functions of the polmineR package have been replaced by lower-level functions.
• The functions germaparl_encode_speeches() and germaparl_encode_lda_topics() have been moved to the (GitHub-only) polmineR.misc package. These are higher-level functions that rely on polmineR classes and methods. Keeping them in the GermaParl package would require to make polmineR a dependency of GermaParl. But as GermaParl is designed to become a dependency of polmineR, we prevent a circular dependecy by removing the functions. What is more, both functions have been designed to augmment GermaParl, but their essence is morge generic. In the long run, a cwbtools.misc package (to be created) might be the most logical place for generic functionality to augment corpora.

GermaParl 1.4.2

• Most functions now include an argument sample that defaults to FALSE. If set as TRUE, functionality to retrieve information from the corpus or to modify the corpus will be applied to the smaller GERMAPARLSAMPLE corpus rather than the GERMAPARL corpus.
• The sample workflow of the overall package documentation object will now rely on the GERMAPARLSAMPLE corpus rather than the full GERMAPARL corpus.
• A Rmarkdown document in the data-raw folder explains how the topic model for the sample corpus has been prepared.

GermaParl 1.4.1

• The 'topicmodels' package has been turned into a suggested package and has been moved from the 'Depends' section to the 'Suggests' section in the DESCRIPTION file.
• Rework of the documentation, including examples.

GermaParl 1.4.0

• To meet CRAN requirements, the corpus is not stored within the package as in previous version, but in a system corpus directory. The same is applies to supplementary data such as LDA topic models fitted on GERMAPARL.
• The core of the functionality of germaparl_download_corpus() to download the corpus has been moved to cwbtools (v0.2.0). The germaparl_download_corpus() function is now a convenience wrapper for cwbtools::corpus_install() that ensures that the correct DOI (argument doi) is passed to corpus_install().

GermaParl 1.3.0

• The GermaParl corpus is downloaded now from a storage location at zenodo. The germapar_download_corpus() function has been reworked accordingly. It now takes the argument doi.
• The GermaParl R6 class has been dropped from the package. The main method of the function ($summary()) is superfluous as the size()-method of the polmineR package produces the same output (data.table with report of sizes of subcorpora on according to an s_attribute).
• The function germaparl_add_p_attribute_stem() has been removed from the package. The functionality (adding a new p-attribute with word stems) makes sense, but the implementation should be generic and included in the cwbtools package.
• The file 'zzz.R' has been removed from the package. It moved the registry files from the GermaParl package to a central registry. This is not necessary any more as polmineR works with a temporary registry directory.
• The file 'configure' in the main directory of the package has been removed, as the polmineR approach to work with a temporary registry does not require the paths in the registry files to be set. The file 'tools/setpaths.R' has been removed for the same reason.
• The germaparl_regdir() function has been removed. Not necessary any more.
• The germaparl_search_speeches() function has been removed from the package. The functionality is nice, but there should either be a generic implementation in the polmineR package, or it might be offered as a recipe.
• Documentation for function to work with topicmodels have been integrated into one documentation object.
• Bug removed from germaparl_download_lda examples.
• germaparl_load_lda will now return NULL object (instead of throwing an error) if lda model is not present.
• germaparl_encode_lda_topics will now issue a warning (instead of crashing) if the s-attribute 'speeches' has not yet been generated.
• The tarball is downloaded from Zenodo.

GermaParl 1.2.1

• configure.win script is removed so that installation works on Windows without Rtools installation

GermaParl 1.2.0

• Package includes sample data, but not full corpus upon installation, use germaparl_download_corpus() for downloading full corpus.
• installation script setpaths.R in ./tools used for adapting paths

GermaParl 1.1.0

• Stanford CoreNLP (updated ctk package as interface) used for tokenization and annotation
• vignette "The Making of GermaParl" included that documents annotation procedure
• simplification of names of s-attributes (e.g. date instead of text_date)
• data in the GermaParl package synced with data available at github.com/PolMine/GermaParlTEI
• data includes 18th legislative period (until December 2016)

GermaParl 1.0.5

• extended README added
• English version of the package vignette prepared
• documentation generated with pkgdown

GermaParl 1.0.4

• template.json added to make use of new template mechanism (as offered by polmineR 0.7.4)

GermaParl 1.0.3

• package/corpus renamed to GermaParl
• no substantial changes

GermaParl 1.0.2

• adding a (German) vignette to the package

GermaParl documentation built on Oct. 23, 2020, 8:27 p.m.