diff --git a/NAMESPACE b/NAMESPACE
index ca7d071..5150447 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -1,6 +1,7 @@
 # Generated by roxygen2: do not edit by hand
 
 export(actor_aggregation)
+export(actor_fetcher)
 export(actorizer)
 export(aggregator)
 export(aggregator_elastic)
diff --git a/R/actor_fetcher.R b/R/actor_fetcher.R
new file mode 100644
index 0000000..b3d21a6
--- /dev/null
+++ b/R/actor_fetcher.R
@@ -0,0 +1,95 @@
+#' Generate actor data frames (with sentiment) from database
+#'
+#' Generate actor data frames (with sentiment) from database
+#' @param out Data frame produced by elasticizer
+#' @param sent_dict Optional dataframe containing the sentiment dictionary (see sentiment paper scripts for details on format)
+#' @param cores Number of threads to use for parallel processing
+#' @param validation Boolean indicating whether human validation should be performed on sentiment scoring
+#' @return No return value, data per batch is saved in an RDS file
+#' @export
+#' @examples
+#' actor_fetcher(out, sent_dict = NULL, cores = 1)
+#################################################################################################
+#################################### Aggregate actor results ################################
+#################################################################################################
+actor_fetcher <- function(out, sent_dict = NULL, cores = 1, localhost = NULL, validation = F) {
+  plan(multiprocess, workers = cores)
+  ### Functions
+  ### Calculate sentiment scores for each actor-document
+  sent_scorer <- function(row, out_row, ud_sent) {
+    ### Contains sentiment per sentence for actor
+    actor_tone <- filter(ud_sent, sentence_id %in% unlist(out_row[row,]$sentence_id))
+
+    ### Aggregated sentiment per actor (over all sentences containing actor)
+    actor <- summarise(actor_tone,
+                       sent = sum(sent_sum)/sum(words),
+                       sent_sum = sum(sent_sum),
+                       sent_words = sum(sent_words),
+                       words = sum(words),
+                       arousal = sum(sent_words)/sum(words)
+    )
+    return(cbind(out_row[row,],data.frame(actor = actor)))
+  }
+
+  par_sent <- function(row, out, sent_dict = NULL) {
+    out_row <- out[row,]
+    ### Generating actor dataframe, unnest by actorsDetail, then by actor ids. Filter out non-relevant actor ids.
+    if (is.null(sent_dict) == F) {
+      ud_sent <- out_row$`_source.ud`[[1]] %>%
+        select(-one_of('exists')) %>%
+        unnest() %>%
+        filter(upos != 'PUNCT') %>% # For getting proper word counts
+        mutate(V1 = str_c(lemma,'_',upos)) %>%
+        left_join(sent_dict, by = 'V1') %>%
+        ### Setting binary sentiment as unit of analysis
+        mutate(V2 = V3) %>%
+        group_by(sentence_id) %>%
+        mutate(
+          V2 = case_when(
+            is.na(V2) == T ~ 0,
+            TRUE ~ V2
+          )
+        ) %>%
+        summarise(sent_sum = sum(V2),
+                  words = length(lemma),
+                  sent_words = length(na.omit(V3))) %>%
+        mutate(
+          sent = sent_sum/words,
+          arousal = sent_words/words
+        )
+      out_row <- select(out_row, -`_source.ud`) %>%
+        unnest(`_source.computerCodes.actorsDetail`, .preserve = colnames(.))
+      ### Aggregated sentiment per article (over all sentences in article)
+      text_sent <- summarise(ud_sent,
+                             sent = sum(sent_sum)/sum(words),
+                             sent_sum = sum(sent_sum),
+                             sent_words = sum(sent_words),
+                             words = sum(words),
+                             arousal = sum(sent_words)/sum(words)
+      )
+      out_row <- bind_rows(lapply(seq(1,nrow(out_row),1),sent_scorer, out_row = out_row, ud_sent = ud_sent)) %>%
+        cbind(., text = text_sent)
+      if (validation == T) {
+        codes_sent <- filter(ud_sent, sentence_id == out_row$`_source.codes.sentence.id`[1]) %>%
+          select(-sentence_id)
+        out_row <- cbind(out_row, codes = codes_sent)
+      }
+    } else {
+      out_row <- unnest(out_row, `_source.computerCodes.actorsDetail`, .preserve = colnames(.))
+    }
+    out_row <- out_row %>%
+      mutate(
+        year = strftime(`_source.publication_date`, format = '%Y'),
+        yearmonth = strftime(`_source.publication_date`, format = '%Y%m'),
+        yearmonthday = strftime(`_source.publication_date`, format = '%Y%m%d'),
+        yearweek = strftime(`_source.publication_date`, format = "%Y%V")
+      ) %>%
+      select(-`_source.computerCodes.actorsDetail`,
+             -`_score`,
+             -`_index`,
+             -`_type`)
+    return(out_row)
+  }
+  saveRDS(bind_rows(future_lapply(1:nrow(out), par_sent, out = out, sent_dict = sent_dict)), file = paste0('df_out',as.numeric(as.POSIXct(Sys.time())),'.Rds'))
+  return()
+}
diff --git a/man/actor_aggregation.Rd b/man/actor_aggregation.Rd
index 33e40db..8d60aff 100644
--- a/man/actor_aggregation.Rd
+++ b/man/actor_aggregation.Rd
@@ -5,7 +5,7 @@
 \title{Generate aggregated actor measures from raw data}
 \usage{
 actor_aggregation(row, actors, es_pwd, localhost,
-  default_operator = "OR")
+  default_operator = "OR", sent_dict = NULL, cores = detectCores())
 }
 \arguments{
 \item{row}{The row of the actors data frame used for aggregation}
diff --git a/man/actor_fetcher.Rd b/man/actor_fetcher.Rd
new file mode 100644
index 0000000..5c7417b
--- /dev/null
+++ b/man/actor_fetcher.Rd
@@ -0,0 +1,27 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/actor_fetcher.R
+\name{actor_fetcher}
+\alias{actor_fetcher}
+\title{Generate actor data frames (with sentiment) from database}
+\usage{
+actor_fetcher(out, sent_dict = NULL, cores = 1, localhost = NULL,
+  validation = F)
+}
+\arguments{
+\item{out}{Data frame produced by elasticizer}
+
+\item{sent_dict}{Optional dataframe containing the sentiment dictionary (see sentiment paper scripts for details on format)}
+
+\item{cores}{Number of threads to use for parallel processing}
+
+\item{validation}{Boolean indicating whether human validation should be performed on sentiment scoring}
+}
+\value{
+No return value, data per batch is saved in an RDS file
+}
+\description{
+Generate actor data frames (with sentiment) from database
+}
+\examples{
+actor_fetcher(out, sent_dict = NULL, cores = 1)
+}
diff --git a/man/aggregator_elastic.Rd b/man/aggregator_elastic.Rd
index 89a2f44..547b1dc 100644
--- a/man/aggregator_elastic.Rd
+++ b/man/aggregator_elastic.Rd
@@ -1,10 +1,10 @@
 % Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/actor_aggregation_db.R
+% Please edit documentation in R/aggregator_elastic.R
 \name{aggregator_elastic}
 \alias{aggregator_elastic}
 \title{Generate and store aggregate actor measures to elasticsearch}
 \usage{
-aggregator_elastic(out, localhost = F, actorids, ver, cores, es_super)
+aggregator_elastic(out, localhost = F, actorids, ver, es_super)
 }
 \arguments{
 \item{out}{The output provided by elasticizer()}
@@ -15,8 +15,6 @@ aggregator_elastic(out, localhost = F, actorids, ver, cores, es_super)
 
 \item{ver}{String indicating the version of the update}
 
-\item{cores}{Numeric value indicating the number of cores to use for processing}
-
 \item{es_super}{Write password for ES}
 }
 \value{
@@ -26,5 +24,5 @@ Return value is based on output of elastic_update()
 Generate and store aggregate actor measures to elasticsearch
 }
 \examples{
-aggregator_elastic(out, localhost = F, actorids, ver, cores, es_super)
+aggregator_elastic(out, localhost = F, actorids, ver, es_super)
 }