cran-comments.md

In hope-data-science/akc: Automatic Knowledge Classification

0.9.9

Update the citation in The R Journal.

0.9.8.1

The data.table has an update that affects the package, fix to make it work again. Details: akc ===== Error: package or namespace load failed for ‘akc’: object ‘update.dev.pkg’ is not exported by 'namespace:data.table'

Yes we have changed the name to update_dev_pkg(). This function upgrades the user's installation to the very latest commit of data.table in development. As such I'm surprised your package depends on it. It is supposed to be called by users, yes. But it is not supposed to be called by packages.

0.9.8

1.Using the latest version of roxygen2 to process the package submission, so as to deal with the CRAN notification. 2.Remove outdated URL in the document.

0.9.7

1. Update the ORCID.
2. Better styling for the display (e.g. white space between brackets and texts).
3. In make_dict function, ensure the words in lower case.
4. Add a vignette to show benchmark for different algorithms and network sizes.

0.9.6

1. We have added the parameter lemmatize_dict in the keyword_clean function, users can now use a customized dictionary in this function.
2. Added lemmatize_dict and stem_lang parameter for keyword_merge function, so as to utilize other language and dictionary options when using 'textstem' package.
3. Update codes in the keyword_vis and keyword_network function, the former code guides(fill = FALSE) has changed to guides(fill = "none")
4. Import tidyfst::pairwise_count_dt instead of widyr::pairwise_count, so as to suppress the warning and run faster.
5. Add unit test for the package.

0.9.5

1. Add "group_no" to keyword_cloud to get single group of word cloud.
2. Add doc_group function to classify documents automatically.
3. Fix problem mentioned in https://github.com/hope-data-science/akc/pull/4.
4. Suppress warnings caused by widyr::pairwise_count,see https://stackoverflow.com/questions/63977452/warning-message-pairwise-count-function (keyword_group).
5. Suppress warnings caused by group_by in the new dplyr version(keyword_vis & keyword_network).
6. Urls in keyword_group are removed because these links might be unstable. However, through the reference users could still find the source paper via searching.

0.9.4 20200130

1. Add "stopword" parameter to keyword_extract.
2. Add keyword_cloud function for visualization of keywords.
3. Add keyword_network function to provide more flexible network visualization.
4. Add a new vignette as tutorial titled as "Tutorial for knowledge classification from raw text".

v0.9.3 20200126

1. Rearrange the column order of keyword_extract,put "id" first.
2. When use the "partof" method in keyword_merge, iterate this process until no more "partof" method to make new changes to the data.frame.
3. Add parameter "alpha" to keyword_vis to control transparency.

v0.9.2

Add a new option for keyword_merge,namely "partof". This option could help merge tokens with relatively same meaning in the same document. For instance, "time series" and "time series analysis" would be merged into "time series". This method could only be applied to more-than-one-word phrases, which means "time series" and "time" would never be merged into "time" (which is considered as information loss). This alternative is an advanced method and should be used with caution.

v0.9.1 20200109

1. For keyword_extract,give the alternative option to let parameter dict = NULL, then the keywords would not be filtered by the dictionary. This could be useful in some circumstances.
2. Because data.table has some deprecated functions, we remove some of the functions imported from data.table. This might be urgent. Details are listed in https://github.com/hope-data-science/akc/issues/1#issuecomment-572388600.
3. For keyword_vis,change mutate(Group = str_c("Group ",group)) to mutate(Group = str_c("Group ",group) %>% reorder(group)). This means that when there are more than 10 plots, it would be ordered as 1~10, but not 1,10,2...

v0.9.0 20200107

Fix the Description, adding quotes to API name. In the future the frame work will be extended to use more methods for keyword clustering in the tidy framework.

v0.8.9 20191217

1.There might be two max values in keyword_merge, previously keep both, now randomly save only one. 2.Correct some typos in the vignette. 3.Document standardization,e.g. use TRUE and FALSE instead of T and F. 4.Update the description of package for more details. 5.Use \donttest{} and \dontrun{} for codes demanding longer time.This package could deal with relatively big data, but it demands computer with larger RAM.

v0.8.8

Change back to unnest_token for high performance

v0.8.7 20191211

1.Stop word should not be used in keyword_extract, hense I remove it. Because pre-defined stop words might hurt the following task of keyword extraction. 2.The previous keyword_extract did not implement the tokenizatin correctly (did not segment the sentence first, then extract n-grams), so I have adjusted the codes to make it right (there might be something wrong in the unnest_tokens in "ngram" mode, I use tokenizers::tokennize_ngrams directly to debug). 3.Add examples to make_dict function.

v0.8.6 20191210

Export the tidygraph community detection functions. Change the detection function parameter name to "com_detect_fun". Add github url information and maintainer information.

Test environments

• local OS X install, R 3.6.1
• ubuntu 14.04 (on travis-ci), R 3.6.1
• win-builder (devel and release)

R CMD check results

0 errors | 0 warnings | 1 note

• This is a new release.

hope-data-science/akc documentation built on Feb. 8, 2023, 8:49 a.m.