Merge tag '0.6'

CRAN 0.6 release

DESCRIPTION

@@ -1,7 +1,6 @@
 Package: text2vec
 Type: Package
-Version: 0.5.1.5
-Date: 2018-05-28
+Version: 0.6
 Title: Modern Text Mining Framework for R
 License: GPL (>= 2) | file LICENSE
 Description: Fast and memory-friendly tools for text vectorization, topic
@@ -18,37 +17,34 @@ Authors@R: c(
       comment = "Author of the WaprLDA C++ code"))
 Maintainer: Dmitriy Selivanov <selivanov.dmitriy@gmail.com>
 Encoding: UTF-8
-SystemRequirements: GNU make, C++11
+SystemRequirements: C++11
 Depends:
-    R (>= 3.2.0),
+    R (>= 3.6.0),
     methods
 Imports:
     Matrix (>= 1.1),
-    Rcpp (>= 0.11),
-    RcppParallel (>= 4.3.14),
-    digest (>= 0.6.8),
-    foreach(>= 1.4.3),
+    Rcpp (>= 1.0.3),
+    R6 (>= 2.3.0), 
     data.table(>= 1.9.6),
-    irlba (>= 2.2.1),
-    R6 (>= 2.1.2), 
-    futile.logger (>= 1.4.3), 
+    rsparse (>= 0.3.3.4), 
     stringi (>= 1.1.5), 
-    mlapi (>= 0.1.0)
-LinkingTo: Rcpp,
-    RcppParallel,
-    digest,
-    sparsepp (>= 0.2.0)
+    mlapi (>= 0.1.0), 
+    lgr (>= 0.2), 
+    digest (>= 0.6.8)
+LinkingTo: 
+    Rcpp,
+    digest (>= 0.6.8)
 Suggests:
-    parallel,
-    doParallel,
-    glmnet,
+    magrittr,
     udpipe (>= 0.6),
+    glmnet,
     testthat,
     covr,
     knitr,
-    rmarkdown
+    rmarkdown,
+    proxy
 URL: http://text2vec.org
 BugReports: https://github.com/dselivanov/text2vec/issues
 VignetteBuilder: knitr
 LazyData: true
-RoxygenNote: 6.0.1
+RoxygenNote: 6.1.1

NAMESPACE

@@ -2,18 +2,16 @@
 
 S3method(create_dtm,itoken)
 S3method(create_dtm,itoken_parallel)
-S3method(create_dtm,list)
 S3method(create_tcm,itoken)
 S3method(create_tcm,itoken_parallel)
 S3method(create_vocabulary,character)
 S3method(create_vocabulary,itoken)
 S3method(create_vocabulary,itoken_parallel)
-S3method(create_vocabulary,list)
 S3method(itoken,character)
 S3method(itoken,iterator)
 S3method(itoken,list)
 S3method(itoken_parallel,character)
-S3method(itoken_parallel,ifiles_parallel)
+S3method(itoken_parallel,iterator)
 S3method(itoken_parallel,list)
 S3method(print,text2vec_vocabulary)
 export(BNS)
@@ -38,13 +36,13 @@ export(create_vocabulary)
 export(dist2)
 export(fit)
 export(fit_transform)
-export(glove)
 export(hash_vectorizer)
 export(idir)
 export(ifiles)
 export(ifiles_parallel)
 export(itoken)
 export(itoken_parallel)
+export(jsPCA_robust)
 export(normalize)
 export(pdist2)
 export(perplexity)
@@ -65,15 +63,8 @@ import(digest)
 import(methods)
 import(mlapi)
 importFrom(R6,R6Class)
-importFrom(RcppParallel,RcppParallelLibs)
-importFrom(foreach,"%do%")
-importFrom(foreach,"%dopar%")
-importFrom(foreach,foreach)
-importFrom(futile.logger,flog.debug)
-importFrom(futile.logger,flog.error)
-importFrom(futile.logger,flog.info)
-importFrom(futile.logger,flog.warn)
 importFrom(methods,as)
+importFrom(rsparse,GloVe)
 importFrom(utils,setTxtProgressBar)
 importFrom(utils,txtProgressBar)
 useDynLib("text2vec", .registration=TRUE)

NEWS.md

@@ -1,4 +1,8 @@
-# text2vec 0.5.1.*
+# text2vec 0.6
+1. 2019-12-17
+    *  **breaking change** - removed construction of a vocabulary in parallel on windows
+    * use `rsparse` package for SVD and GloVe factorizations
+    * uodated RWMD implementation (hopefully bug free)
 1. 2018-09-10
     *  **breaking change** - changed IDF formula - see #280 for details.
 1. 2018-05-28

R/RcppExports.R

@@ -1,30 +1,6 @@
 # Generated by using Rcpp::compileAttributes() -> do not edit by hand
 # Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
 
-cpp_glove_create <- function(params) {
-    .Call(`_text2vec_cpp_glove_create`, params)
-}
-
-cpp_glove_get_word_vectors <- function(ptr) {
-    .Call(`_text2vec_cpp_glove_get_word_vectors`, ptr)
-}
-
-cpp_glove_set_cost_zero <- function(ptr) {
-    invisible(.Call(`_text2vec_cpp_glove_set_cost_zero`, ptr))
-}
-
-cpp_glove_partial_fit <- function(ptr, x_irow, x_icol, x_val, iter_order) {
-    .Call(`_text2vec_cpp_glove_partial_fit`, ptr, x_irow, x_icol, x_val, iter_order)
-}
-
-cpp_glove_get_sparsity_level <- function(ptr) {
-    .Call(`_text2vec_cpp_glove_get_sparsity_level`, ptr)
-}
-
-cpp_glove_dump_model <- function(ptr) {
-    .Call(`_text2vec_cpp_glove_dump_model`, ptr)
-}
-
 hasher <- function(x, hash_size) {
     .Call(`_text2vec_hasher`, x, hash_size)
 }

R/analogies.R

@@ -39,7 +39,7 @@ prepare_analogy_questions = function(questions_file_path, vocab_terms) {# nocov
 
   questions_number = sum(sapply(q, nrow))
 
-  flog.info("%d full questions found out of %d total",
+  logger$info("%d full questions found out of %d total",
             questions_number,
             length(lines) - length(section_name_ind))
 
@@ -56,8 +56,8 @@ prepare_analogy_questions = function(questions_file_path, vocab_terms) {# nocov
 #' @param m_word_vectors word vectors \code{numeric matrix}. Each row should
 #'   represent a word.
 #' @description This function checks how well the GloVe word embeddings do on
-#'   the analogy task. For full examples see \link{glove}.
-#' @seealso \link{prepare_analogy_questions}, \link{glove}
+#'   the analogy task. For full examples see \link{GloVe}.
+#' @seealso \link{prepare_analogy_questions}, \link{GloVe}
 #' @export
 check_analogy_accuracy = function(questions_list, m_word_vectors) {
 
@@ -90,7 +90,7 @@ check_analogy_accuracy = function(questions_list, m_word_vectors) {
     act = q_mat[, 4]
     correct_number = sum(preds == act)
 
-    flog.info("%s: correct %d out of %d, accuracy = %.4f",
+    logger$info("%s: correct %d out of %d, accuracy = %.4f",
                    category,
                    correct_number,
                    q_number,
@@ -104,6 +104,6 @@ check_analogy_accuracy = function(questions_list, m_word_vectors) {
       )
   }
   res = rbindlist(res)
-  flog.info("OVERALL ACCURACY = %.4f", sum(res[['predicted']] == res[['actual']]) / nrow(res) )
+  logger$info("OVERALL ACCURACY = %.4f", sum(res[['predicted']] == res[['actual']]) / nrow(res) )
   res
 }# nocov end

R/coherence.R

@@ -133,7 +133,8 @@
 #' tcm = crossprod(sign(dtm))
 #'
 #' # check coherence
-#' futile.logger::flog.threshold(futile.logger::DEBUG)
+#' logger = lgr::get_logger('text2vec')
+#' logger$set_threshold('debug')
 #' res = coherence(tw, tcm, n_doc_tcm = N)
 #' res
 #'
@@ -234,7 +235,7 @@ res = matrix(NA_real_, nrow = n_topics, ncol = n_metrics,
     topic_i_term_indices = topic_i_term_indices[!is.na(topic_i_term_indices)]
     for(j in seq_len(n_metrics)) {
       m = metrics[j]
-      futile.logger::flog.debug("calculating coherence metric '%s' for topic %d", m, i)
+      logger$debug("calculating coherence metric '%s' for topic %d", m, i)
       res[i, j] = calc_coherence(m, topic_i_term_indices, tcm, smooth, n_doc_tcm = n_doc_tcm)
     }
   }

R/distance.R

@@ -81,15 +81,15 @@ dist2 = function(x, y = NULL, method = c("cosine", "euclidean", "jaccard"),
     method = match.arg(method)
     if(method %in% c("cosine", "jaccard")) {
       if( inherits(x, "sparseMatrix") || inherits(y, "sparseMatrix"))
-        flog.warn("Sparsity will be lost - worth to calculate similarity instead of distance.")
+        logger$warn("Sparsity will be lost - worth to calculate similarity instead of distance.")
       RESULT = 1 - sim2(x = x, y = y, method = method, norm = norm)
     }
     if (method == "euclidean") {
       if (!FLAG_TWO_MATRICES_INPUT)
         y = x
       if (!inherits(x, "matrix") || !inherits(y, "matrix")) {
         msg = "At the moment eucludian distance could be calculated only for dense matrices of class 'matrix'"
-        flog.error(msg)
+        logger$error(msg)
         stop(msg)
 
       }
@@ -109,7 +109,7 @@ dist2 = function(x, y = NULL, method = c("cosine", "euclidean", "jaccard"),
     if (inherits(method, "RWMD")) {
       if (norm != "none") {
         msg = paste(norm, "norm provided. RWMD can be computed only on bag-of-words matrices - raw word-counts")
-        flog.warn(msg)
+        logger$warn(msg)
       }
       RESULT = method$dist2(x, y)
     }
@@ -154,7 +154,7 @@ pdist2 = function(x, y, method = c("cosine", "euclidean", "jaccard"),
     if (inherits(method, "RWMD")) {
       if (norm != "none") {
         msg = paste(norm, "norm provided. RWMD can be computed only on bag-of-words matrices - raw word-counts")
-        flog.warn(msg)
+        logger$warn(msg)
       }
       RESULT = method$pdist2(x, y)
     }
@@ -217,7 +217,7 @@ sim2 = function(x, y = NULL, method = c("cosine", "jaccard"),
     if (norm != "none") {
       msg = paste(norm, "norm provided. Howewer matrix will be converted to binary (0,1) automatically.")
       msg = paste(msg, "'jaccard' can be computed only on sets which should be encoded as sparse matrices of 0, 1.")
-      flog.warn(msg)
+      logger$warn(msg)
     }
     x@x = sign(x@x)
     if (FLAG_TWO_MATRICES_INPUT) {
@@ -259,7 +259,7 @@ psim2 = function(x, y, method = c("cosine", "jaccard"), norm = c("l2", "none"))
     if (norm != "none") {
       msg = paste(norm, "norm provided. Howewer matrix will be converted to binary (0,1) automatically.")
       msg = paste(msg, "'jaccard' can be computed only on sets which should be encoded as sparse matrices of 0, 1.")
-      flog.warn(msg)
+      logger$warn(msg)
     }
 
     x@x = sign(x@x)