#Author: Dr Will Parry 
#Web: http://willparry.net

#Load packages
pkgs <- c('data.table', 'purrr', 'stringr', 'stringdist')
sapply(pkgs, require, character.only = TRUE)

#Read database extract
db_extract <- 'csv-longtermco-set.csv' #example extract from pubmed
#NB: must have a Title field on which to match

#Read extract into data table and add a string for title matching
db_table <- fread(db_extract) %>%
  .[, match_string := toupper(str_remove_all(Title, '[:punct:]| '))]
#string is made from Title field, made upper case, with all punctuation and spaces removed

#Create table of 'distances' between pairs of match_string values using {stringdist} package
dist_table <- db_table[['match_string']] %>% #using just the vector of match_strings 
  combn(2) %>% t() %>% #get all combinations and transpose result to long format
  data.table() %>% #make into a data table
  setnames(c('V1', 'V2'), c('title1', 'title2')) %>% #rename columns
  .[, dist := stringdist(title1, title2)] %>% #add string distance metric
  .[order(dist)] #order in increasing distance (top of table will be most similar)

View(head(dist_table)) #view the most similar results

#Distance of zero is an exact match
#In the example, there is:
# - one exact match (duplicated paper)
# - one patient comment on a listed paper
# - one erratum for a listed paper