diff --git a/.gitignore b/.gitignore index e75435c..69e2a14 100644 --- a/.gitignore +++ b/.gitignore @@ -1,49 +1,37 @@ # History files .Rhistory .Rapp.history - # Session Data files .RData .RDataTmp - # User-specific files .Ruserdata - # Example code in package build process *-Ex.R - # Output files from R CMD build /*.tar.gz - # Output files from R CMD check /*.Rcheck/ - # RStudio files .Rproj.user/ - # produced vignettes vignettes/*.html vignettes/*.pdf - # OAuth2 token, see https://github.com/hadley/httr/releases/tag/v0.3 .httr-oauth - # knitr and R markdown default cache directories *_cache/ /cache/ - # Temporary files created by R markdown *.utf8.md *.knit.md - # R Environment Variables .Renviron - # pkgdown site docs/ - # translation temp files po/*~ - # RStudio Connect folder rsconnect/ +record_linkage.Rproj +deduplicate.html diff --git a/missing_gender_result.csv b/data-raw/missing_gender_result.csv similarity index 100% rename from missing_gender_result.csv rename to data-raw/missing_gender_result.csv diff --git a/deduplicate.R b/deduplicate.R deleted file mode 100644 index fa3e6c0..0000000 --- a/deduplicate.R +++ /dev/null @@ -1,774 +0,0 @@ -################################################## -### A script workflow for Record linkage ---------- -################################################## - -library(tidyverse) -# install.packages("fastLink") -library(fastLink) - -## Load the data - -# which is here already the results of merging multiple list from different excel files -data <- readxl::read_excel(here::here("data-raw", "Registros2.xlsx"), - sheet = "Sheet1", - col_types = c("numeric", - "text", "text", "text", "text", "text", - "text", "text", "date", "numeric", - "numeric", "numeric", "text", "text", - "text", "text", "text", "text", "text", - "text", "text", "text", "text", "text", - "text", "text", "text", "text", "text", - "text", "text")) |> - janitor::clean_names() - - -#dput(names(data)) -## Cleaning functions ############### - - -#' clean_age -#' -#' This functions cleans the age identifiers in the context of record linkage. -#' -#' If the date of birth is present, it will parse it, extract year and month, -#' and recalculate both age and age range -#' -#' If the date of birth is not mentioned, then it will use age and date_record -#' to reconstruct an estimated date of birth - and recalculate the rest -#' -#' @param frame frame with the data -#' @param date_birth variable name for the date of birth in the frame -#' @param date_record variable name for the date of birth in the frame -#' @param age variable name for the date of birth in the frame -#' @param age_range variable name for the date of birth in the frame -#' -#' @return the same column but cleaned... -#' -#' @export -clean_age <- function(frame, - date_birth, - date_record, - age, - age_range){ - - #frame$date_birth - frame2 <- frame |> - ## Rename variable - dplyr::rename( date_birth = paste0(date_birth), - date_record = paste0(date_record), - age = paste0(age), - age_range = paste0(age_range) ) |> - dplyr::mutate( - ## In case there is no date of birth - but we have an age, we recalculate it.. - - date_birth = dplyr::case_when( - # Case we age but no DOB and date registration - is.na(date_birth) & !(is.na(age)) & !(is.na(date_record)) ~ - lubridate::as_date(date_record) - lubridate::dyears(age) , - - # Case we age but no DOB and date registration - is.na(date_birth) & !(is.na(age)) & is.na(date_record) ~ - today() - lubridate::dyears(age) , - - # Case take what we have... - TRUE ~ lubridate::as_date(date_birth) ) , - ## make sure it is in the correct format - date_birth = lubridate::as_date(date_birth), - - day_birth = as.numeric( - lubridate::day(date_birth)), - - month_birth = as.numeric( - lubridate::month(date_birth)), - - year_birth= as.numeric( - lubridate::year(date_birth)), - - age = round( as.numeric(today() - date_birth) / 365), - - age_range = dplyr::case_when( - ## if age cohort was already present and we have no DOb, retain it - # !(is.na(age_range )) & is.na( date_birth) & - age <5 ~ "0-4", - - # !(is.na(age_range )) & is.na( date_birth) & - age >=5 & age <= 11 ~ "4-11", - - #!(is.na(age_range )) & is.na( date_birth) & - age >= 12 & age <= 17 ~ "12-17", - # !(is.na(age_range )) & is.na( date_birth) & - age >= 18 & age <= 59 ~ "18-59", - - # !(is.na(age_range )) & is.na( date_birth) & - age >= 60 ~ "60+", - - TRUE ~ age_range ) ) - - return(frame2) -} - - -## Testing... -# frame <- data |> -# dplyr::select(fecha_de_nacimiento, -# date_record, edad, age_range) |> -# clean_age( date_birth = "fecha_de_nacimiento", -# date_record = "date_record", -# age = "edad", -# age_range = "age_range") - - - -#' remove_spaces_based_on_patterns -#' Utility sub-function to remove spaces based on patterns -#' This function helps cleaning name decomposition - in case it is not included in the original data -#' -#' In case, there's only 2 elements, it will fill only firstname and fathername -#' The function also identify family prefix to be bind such as for spanish -#' "DEL", 'DE", "DE LOS", "DE LAS" -#' -#' @param vector a list of string with names -remove_spaces_based_on_patterns <- function(vector, - nameprefix = data.frame( - pat1 = c( "DE LA ", "DEL ", "DE LOS ", "DE LAS ","DE ", "SAN ", "LA ", "DA "), - pat2 = c( "DE_LA_", "DEL_", "DE_LOS_", "DE_LAS_", "DE_", "SAN_", "LA_", "DA_") ) -) { - for (i in 1:nrow(nameprefix) ) { - # i <- 4 - #cat(paste0(nameprefix[i, c("pat1")],"\n", vector, "\n")) - vector <- stringr::str_replace_all(string = vector, - pattern = nameprefix[i, c("pat1")], - replacement = nameprefix[i, c("pat2")]) - # cat(paste0(vector, "\n")) - } - return(vector) -} -## Test function -remove_spaces_based_on_patterns(vector = "ADRIENNE DE LOS ANGELES MILANESE PARISIANNA") - - -#' reset_spaces_based_on_patterns -#' Utility sub-function to remove spaces based on patterns -#' This function helps cleaning name decomposition - in case it is not included in the original data -#' -#' In case, there's only 2 elements, it will fill only firstname and fathername -#' The function also identify family prefix to be bind such as for spanish -#' "DEL", 'DE", "DE LOS", "DE LAS" -#' -#' @param vector a list of string with names -reset_spaces_based_on_patterns <- function(vector, - nameprefix = data.frame( - pat1 = c( "DE LA ", "DEL ", "DE LOS ", "DE LAS ","DE ", "SAN ", "LA ", "DA "), - pat2 = c( "DE_LA_", "DEL_", "DE_LOS_", "DE_LAS_", "DE_", "SAN_", "LA_", "DA_") ) - ) { - - - for (i in 1:nrow(nameprefix) ) { - # i <- 1 - # cat(paste0( nameprefix[i, c("pat1")],"\n",vector, "\n")) - vector <- stringr::str_replace_all(string = vector, - pattern = nameprefix[i, c("pat2")], - replacement = nameprefix[i, c("pat1")]) - #cat(paste0(vector, "\n")) - } - return(vector) -} - - - - -#' separate_fullname -#' -#' This function clean name decomposition - in case it is not included in the original data -#' -#' Performing this name decomposition is important in order to enhance record linkage -#' as the name pattern can be different (either "firstname_fathername_mother_name" or -#' "fathername_mothername_firstname") which will minimise linkage probabibility -#' -#' In case, there's only 2 elements, it will fill only firstname and fathername -#' The function also identify family prefix to be bind such as for spanish -#' "DEL", 'DE", "DE LOS", "DE LAS" -#' -#' @param fullname full name including everything together -#' @param firstname first name -#' @param fathername father name -#' @param mothername mother name -#' @param namepattern either "firstname_fathername_mother_name" or -#' "fathername_mothername_firstname" -#' -#' @return a clean list with c("firstname","fathername","mothername") -#' -#' @export -separate_fullname <- function(frame, - fullname, - firstname, - fathername, - mothername, - namepattern - ){ - ## Let's go! - framesp <- frame |> - ## Rename variable - dplyr::rename( fullname = paste0(fullname), - firstname = paste0(firstname), - fathername = paste0(fathername), - mothername = paste0(mothername), - namepattern = paste0(namepattern) ) |> - - ### Lets clean all spaces and get everything to upper - dplyr::mutate( fullname_or = fullname, - fullname = trimws(stringr::str_squish(fullname), - which = "both", - whitespace = "[ \t\r\n]"), - fullname = toupper(fullname)) |> - ### Lets apply the prefix space replacement... - dplyr::mutate( fullname_pref = remove_spaces_based_on_patterns(vector = fullname), - fullname = fullname_pref ) |> - ## Counting th enumber of space to understand the structure of the full name - dplyr::mutate( numspace = stringr::str_count(fullname, ' ')) |> - tidyr::separate_wider_delim(fullname, - delim = " ", - names_sep = "", - too_few = "align_start") |> - - ### Now reconstruct the first name, father and mother name based on cases - - ## Let summarize the logic...! - ## Based on the number of componnent in the full name - ranging from 0 to 4 - - ## Case A: "firstname_fathername_mother_name" - # numspace == 0 ## only firstname = fullname1 - # numspace = 1 ## firstname = fullname1 & fathername = fullname2 - # numspace = 2 ## firstname = fullname1 & fathername = fullname2 & mothername = fullname3 - # numspace = 3 ## firstname = paste0(fullname1, " ", fullname2) & fathername = fullname3 & mothername = fullname4 - # numspace = 4 ## firstname = paste0(fullname1, " ", fullname2, " ", fullname3) & fathername = fullname4 & mothername = fullname5 - - # ## Case B: "fathername_mothername_firstname" - # numspace == 0 ## only fathername = fullname1 - # numspace = 1 ## fathername = fullname1 & firstname = fullname2 - # numspace = 2 ## fathername = fullname1 & mothername = fullname2 & firstname = fullname3 - # numspace = 3 ## fathername = fullname1 & mothername = fullname2 & firstname = paste0(fullname3, " ", fullname3) - # numspace = 4 ## fathername = fullname1 & mothername = fullname2 & firstname = paste0(fullname3, " ", fullname4, " ", fullname5) - - ### Now reconstruct fathername - dplyr::mutate( fathername = dplyr::case_when( - is.na(fathername) & namepattern == "firstname_fathername_mother_name" & - numspace == 0 ~ "", - - is.na(fathername) & namepattern == "firstname_fathername_mother_name" & - numspace == 1 ~ fullname2, - - is.na(fathername) & namepattern == "firstname_fathername_mother_name" & - numspace == 2 ~ fullname2, - - is.na(fathername) & namepattern == "firstname_fathername_mother_name" & - numspace == 3 ~ fullname3, - - is.na(fathername) & namepattern == "firstname_fathername_mother_name" & - numspace == 4 ~ fullname4, - - - is.na(fathername) & namepattern == "fathername_mothername_firstname" & - numspace == 0 ~ fullname1, - - is.na(fathername) & namepattern == "fathername_mothername_firstname" & - numspace == 1 ~ fullname1, - - is.na(fathername) & namepattern == "fathername_mothername_firstname" & - numspace == 2 ~ fullname1, - - is.na(fathername) & namepattern == "fathername_mothername_firstname" & - numspace == 3 ~ fullname1, - - is.na(fathername) & namepattern == "fathername_mothername_firstname" & - numspace == 4 ~ fullname1, - - TRUE ~ fathername )) |> - - ### Now reconstruct mothername - dplyr::mutate( mothername = dplyr::case_when( - is.na(mothername) & namepattern == "firstname_fathername_mother_name" & - numspace == 0 ~ "", - - is.na(mothername) & namepattern == "firstname_fathername_mother_name" & - numspace == 1 ~ "", - - is.na(mothername) & namepattern == "firstname_fathername_mother_name" & - numspace == 2 ~ fullname3, - - is.na(mothername) & namepattern == "firstname_fathername_mother_name" & - numspace == 3 ~ fullname4, - - is.na(mothername) & namepattern == "firstname_fathername_mother_name" & - numspace == 4 ~ fullname5, - - - is.na(mothername) & namepattern == "fathername_mothername_firstname" & - numspace == 0 ~ "", - - is.na(mothername) & namepattern == "fathername_mothername_firstname" & - numspace == 1 ~ "", - - is.na(mothername) & namepattern == "fathername_mothername_firstname" & - numspace == 2 ~ fullname2, - - is.na(mothername) & namepattern == "fathername_mothername_firstname" & - numspace == 3 ~ fullname2, - - is.na(mothername) & namepattern == "fathername_mothername_firstname" & - numspace == 4 ~ fullname2, - - TRUE ~ mothername )) |> - - - ### Now reconstruct firstname - dplyr::mutate( firstname = dplyr::case_when( - is.na(firstname) & namepattern == "firstname_fathername_mother_name" & - numspace == 0 ~ fullname1, - - is.na(firstname) & namepattern == "firstname_fathername_mother_name" & - numspace == 1 ~ fullname1, - - is.na(firstname) & namepattern == "firstname_fathername_mother_name" & - numspace == 2 ~ fullname1, - - is.na(firstname) & namepattern == "firstname_fathername_mother_name" & - numspace == 3 ~ paste0(fullname1, " ", fullname2 ), - - is.na(firstname) & namepattern == "firstname_fathername_mother_name" & - numspace == 4 ~ paste0(fullname1, " ", fullname2, " ", fullname3), - - is.na(firstname) & namepattern == "fathername_mothername_firstname" & - numspace == 0 ~ "", - - is.na(firstname) & namepattern == "fathername_mothername_firstname" & - numspace == 1 ~ fullname2, - - is.na(firstname) & namepattern == "fathername_mothername_firstname" & - numspace == 2 ~ fullname3, - - is.na(firstname) & namepattern == "fathername_mothername_firstname" & - numspace == 3 ~ paste0(fullname3, " ", fullname4 ) , - - is.na(firstname) & namepattern == "fathername_mothername_firstname" & - numspace == 4 ~ paste0(fullname3, " ", fullname4, " ", fullname5) , - - TRUE ~ firstname )) |> - - - ### Lets reset the prefix space replacement... - dplyr::mutate( - firstname = reset_spaces_based_on_patterns(vector = firstname), - fathername = reset_spaces_based_on_patterns(vector = fathername), - mothername = reset_spaces_based_on_patterns(vector = mothername)) |> - - ## then clean intermediate variables - dplyr::select( - fullname1, - fullname2, - fullname3, - fullname4, - - fullname5, fullname_or, - fullname_pref, - numspace ) - - - return(framesp) -} - -## Testing... -# frame <- data |> -# dplyr::filter( is.na(nombres) ) |> -# # dplyr::select(nombre_completo, name_pattern, -# # nombres, apellido_paterno, apellido_materno) |> -# separate_fullname(fullname= "nombre_completo", -# namepattern= "name_pattern", -# firstname = "nombres", -# fathername = "apellido_paterno", -# mothername = "apellido_materno") - - - - - -#' separate_firstname -#' -#' This function decomposition - in case there is a space in -#' -#' Performing this name decomposition is important in order to enhance record linkage -#' -#' @param firstname first name -#' @return a clean list with c("firstname1","firstname2","firstname3") -#' -#' @export -separate_firstname <- function(frame, - firstname ) { - ## Let's go! - framesp <- frame |> - ## Rename variable - dplyr::rename( firstname = paste0(firstname) ) |> - - ### Lets clean all spaces and get everything to upper - dplyr::mutate( firstname_or = firstname, - firstname = trimws(stringr::str_squish(firstname), - which = "both", - whitespace = "[ \t\r\n]"), - firstname = toupper(firstname)) |> - dplyr::mutate( firstname = remove_spaces_based_on_patterns(vector = firstname) ) |> - - ## Counting the number of space to understand the structure of the full name - tidyr::separate_wider_delim(firstname, - delim = " ", - names_sep = "", - too_few = "align_start") |> - dplyr::mutate( firstname1 = reset_spaces_based_on_patterns(vector = firstname1), - firstname2 = reset_spaces_based_on_patterns(vector = firstname2), - firstname3 = reset_spaces_based_on_patterns(vector = firstname3)) - - return(framesp) -} - -framefirstname <- data |> - separate_firstname(firstname = "nombres" ) - - - -#' separate_familyname -#' -#' This function helps in decomposing names in case there is a space -#' in the father name and the mother name is empty -#' -#' Performing this name decomposition is important in order to enhance record linkage -#' -#' @param fathernamevar father name -#' @param mothernamevar mother name -#' @return a clean list with c("firstname1","firstname2","firstname3") -#' -#' @export -separate_familyname <- function(frame, - fathernamevar, - mothernamevar) { - ## Let's go! - framesp <- frame |> - ## Rename variable - dplyr::rename( fathername = paste0(fathernamevar), - mothername = paste0(mothernamevar) ) |> - - ### Lets clean all spaces and get everything to upper - dplyr::mutate( fathername_or = fathername, - fathername= trimws(stringr::str_squish( fathername), - which = "both", - whitespace = "[ \t\r\n]"), - fathername = toupper( fathername)) |> - dplyr::mutate( fathername = remove_spaces_based_on_patterns(vector = fathername) ) |> - - ## Counting the number of space to understand the structure of the full name - tidyr::separate_wider_delim( fathername, - delim = " ", - names_sep = "", - too_few = "align_start") |> - dplyr::mutate( fathername1 = reset_spaces_based_on_patterns(vector = fathername1), - fathername2 = reset_spaces_based_on_patterns(vector = fathername2), - fathername3 = reset_spaces_based_on_patterns(vector = fathername3)) |> - - dplyr::mutate( fathername = dplyr::if_else( is.na(mothername), fathername1, fathername_or ) ) |> - dplyr::mutate( mothername = dplyr::if_else( is.na(mothername), fathername2, mothername)) |> - dplyr::select( - fathername1, - fathername2, - fathername3, - fathername_or) - - return(framesp) -} - -# frame <- data |> -# dplyr::select(apellido_paterno, apellido_materno) |> -# separate_familyname( fathernamevar = "apellido_paterno", -# mothernamevar = "apellido_materno" ) - - -#' cleanvar -#' -#' function for data cleaning with additional name removal logic -#' -#' @param names_column name of the column to treat -#' @param toRemove default vector with stuff to remove from name -#' c(" JR", " SR", " IV", " III", " II") -#' @return names_column_new name of the column treat -#' -#' @export -cleanvar <- function(names_column, - toRemove = c(" JR", " JUNIOR", " SR", " IV", " III", " II")) { - # Convert to uppercase - names_column_new <- toupper(names_column) - # Remove specified name suffixes - for (tR in toRemove) { - names_column_new <- gsub(tR, "", names_column_new) - } - # Convert special characters to ASCII equivalents - names_column_new <- iconv(names_column_new, "latin1", "ASCII//TRANSLIT", sub = "") - # Remove punctuation, digits, and all spaces - names_column_new <- gsub("[[:punct:][:digit:]][[:space:]]", "", names_column_new) - # Create a new variable with only alphabetic characters - names_column_new <- gsub("[^[:alpha:]]", "", names_column_new) - - return(names_column_new) -} - - -## Pipeline for data post processing ##################### -data.prep <- data |> - - ## Filter where the phone number is not available -- "NO REFIERE" - dplyr::filter( telefono != "NO REFIERE") |> - - ## Clean age_range - #data |> dplyr::select(age_range) |> dplyr::distinct() |> dplyr::pull() - dplyr::mutate( age_range = dplyr::case_when( - age_range == "18 A 59 AÑOS" ~ "0-4", - TRUE ~ age_range )) |> - - ## Clean DOb & Age - clean_age( date_birth = "fecha_de_nacimiento", - date_record = "date_record", - age = "edad", - age_range = "age_range") |> - - ## Clean the names - separate_fullname(fullname= "nombre_completo", - namepattern= "name_pattern", - firstname = "nombres", - fathername = "apellido_paterno", - mothername = "apellido_materno") |> - - separate_firstname(firstname = "firstname") |> - - separate_familyname( fathername = "fathername", - mothername = "mothername") |> - - - ## Clean the gender variable - # data |> dplyr::select(genero) |> dplyr::distinct() |> dplyr::pull() - dplyr::mutate(gender = dplyr::case_when( - genero %in% c("F" ,"FEMENINO" ,"f", "Femenino") ~ "F", - genero %in% c("M" , "MASCULINO" , "Masculino") ~ "M", - genero %in% c("X", "Otro") ~ "Ot", - TRUE ~ NA )) |> - - ## Only retain the nationality of interest - # data |> dplyr::select(nacionalidad) |> dplyr::distinct() |> dplyr::pull() - dplyr::mutate(nationality = dplyr::case_when( - nacionalidad %in% c("Venzuela", "venezuela", "Venezolana", - "VENEZUELA", - "Venezuela", "VENEZOLANO", "VENEZOLANA") ~ "VEN", - nacionalidad %in% c("COLOMBIANO", "COLOMBIANA", "COLOMBIA", - "colombia", "Colombia", "Nac. Colombia", - "Colombiana" ) ~ "COL", - TRUE ~ "other" )) |> - dplyr::filter( nationality %in% c("VEN", "COL" )) |> - ## Apply cleanvar() - # Perform data cleaning on dfA using the clean_names function - dplyr::mutate_at( dplyr::vars(firstname1, firstname2, fathername, mothername, - asistencia, departamento, #telefono, - planilla, socio), - list(new = cleanvar)) |> - - - ### identify single data source - dplyr::mutate(datasource = paste0(socio_new, "_", planilla_new)) #|> - - # # ## Retain only fields for record linkage - # dplyr::select(datasource, nationality, - # firstname_new, fathername_new, mothername_new, - # asistencia_new, departamento_new, - # telefono, gender) - - - -dput(names(data.prep)) - - -table(data.prep$gender, useNA = "ifany") -table(data.prep$nationality, useNA = "ifany") - -## Check the datasource that we will compare -table(data.prep$datasource, useNA = "ifany") - -## See if can use departamento for blocking -table(data.prep$datasource, data.prep$departamento_new, useNA = "ifany") - -alldatasource <- data.prep |> - dplyr::select(datasource) |> - dplyr::distinct() |> - dplyr::pull() - -alldatasource - - -## Let's get 2 comparison dataset... ########## -dfA <- data.prep |> - dplyr::filter(datasource == alldatasource[9]) |> - dplyr::select( - datasource) - -dfB <- data.prep |> - dplyr::filter(datasource == alldatasource[6])|> - dplyr::select( - datasource) - -matches.out <- fastLink::fastLink( - dfA = dfA, - dfB = dfB, - # Specify the vector of variable names to be used for matching. - # These variable names should exist in both dfA and dfB - varnames = c("nationality" , - "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new", - "day_birth", "month_birth", "year_birth", - # "asistencia_new" , "departamento_new", "telefono_new" , - "gender" ), - - # Specify which variables among varnames should be compared using string distance - stringdist.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables present in stringdist.match can be partially matched - partial.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables should be matched numerically - # Must be a subset of 'varnames' and must not be present in 'stringdist.match'. - numeric.match = c( "day_birth", "month_birth", "year_birth" - # "telefono_new" - #'dob_day', 'dob_month', 'dob_year' - ), - - # Specify the number of CPU cores to utilize (parallel processing). - ## Get the number of detected cores minus 1, Reserve one core for - #non-computational tasks to help prevent system slowdowns or unresponsiveness - n.cores = parallel::detectCores() - 1, - return.all = TRUE, - return.df = TRUE) - -## Review the matching in each orginal frame -matchedA <- dfA[matches.out$matches$inds.a, ] -matchedB <- dfB[matches.out$matches$inds.b, ] - -# Confusion Matrice -fastLink::confusion(matches.out, threshold = 0.98) - -# Examine the EM object: -#matches.out$EM - -# Summarize the accuracy of the match: -# each column gives the match count, match rate, -# false discovery rate (FDR) and false negative rate (FNR) -# under different cutoffs for matches based on the posterior -# probability of a match. -summary(matches.out) - -## Gt the output... -matched_dfs <- fastLink::getMatches( - dfA = dfA, - dfB = dfB, - fl.out = matches.out, - threshold.match = 0.85 -) - -# Display the matches ################### -# convert dfA rownames to a column -dfA_clean <- dfA |> rownames_to_column() - -# convert dfB rownames to a column -dfB_clean <- dfB |> rownames_to_column() - -# convert all columns in matches dataset to character, -#so they can be joined to the rownames -matches_clean <- matched_dfs |> - dplyr::mutate(dplyr::across(dplyr::everything(), as.character)) - -# Join matches to dfA, then add dfB -# column "inds.b" is added to dfA -complete <- dplyr::left_join(dfA_clean, - matches_clean, - by = c("rowname" = "inds.a")) - -# column(s) from dfB are added -complete <- dplyr::left_join(complete, - dfB_clean, - by = c("inds.b" = "rowname")) - - - -# Preprocessing Matches via Blocking ################# -blockgender_out <- fastLink::blockData(dfA, dfB, varnames = "gender") - -## Subset dfA into blocks -dfA_block1 <- dfA[blockgender_out$block.1$dfA.inds,] -dfA_block2 <- dfA[blockgender_out$block.2$dfA.inds,] - -## Subset dfB into blocks -dfB_block1 <- dfB[blockgender_out$block.1$dfB.inds,] -dfB_block2 <- dfB[blockgender_out$block.2$dfB.inds,] - -## Run fastLink on each -link.1 <- fastLink::fastLink( - dfA = dfA_block1, - dfB = dfB_block1, - # Specify the vector of variable names to be used for matching. - # These variable names should exist in both dfA and dfB - varnames = c("nationality" , - "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new", - "day_birth", "month_birth", "year_birth", - # "asistencia_new" , "departamento_new", "telefono_new" , - "gender" ), - - # Specify which variables among varnames should be compared using string distance - stringdist.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables present in stringdist.match can be partially matched - partial.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables should be matched numerically - # Must be a subset of 'varnames' and must not be present in 'stringdist.match'. - numeric.match = c( "day_birth", "month_birth", "year_birth" - # "telefono_new" - #'dob_day', 'dob_month', 'dob_year' - ), - - # Specify the number of CPU cores to utilize (parallel processing). - ## Get the number of detected cores minus 1, Reserve one core for - #non-computational tasks to help prevent system slowdowns or unresponsiveness - n.cores = parallel::detectCores() - 1, - return.df = TRUE) - -link.2 <- fastLink::fastLink( - dfA = dfA_block2, - dfB = dfB_block2, - # Specify the vector of variable names to be used for matching. - # These variable names should exist in both dfA and dfB - varnames = c("nationality" , - "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new", - "day_birth", "month_birth", "year_birth", - # "asistencia_new" , "departamento_new", "telefono_new" , - "gender" ), - - # Specify which variables among varnames should be compared using string distance - stringdist.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables present in stringdist.match can be partially matched - partial.match = c( "firstname1_new" ,"firstname2_new" , - "fathername_new" ,"mothername_new"), - - # Specify which variables should be matched numerically - # Must be a subset of 'varnames' and must not be present in 'stringdist.match'. - numeric.match = c( "day_birth", "month_birth", "year_birth" - # "telefono_new" - #'dob_day', 'dob_month', 'dob_year' - ), - - # Specify the number of CPU cores to utilize (parallel processing). - ## Get the number of detected cores minus 1, Reserve one core for - #non-computational tasks to help prevent system slowdowns or unresponsiveness - n.cores = parallel::detectCores() - 1, - return.df = TRUE) - -## aggregate multiple matches into a single summary with aggregateEM() -agg.out <- fastLink::aggregateEM(em.list = list(link.1, link.2)) - - - -# Preprocessing Matches via Blocking ################# diff --git a/deduplicate.Rmd b/deduplicate.Rmd index 95c1ee0..9699b05 100644 --- a/deduplicate.Rmd +++ b/deduplicate.Rmd @@ -648,11 +648,11 @@ missing_gender <- data.prep |> ## let's us an API to identify gender based on firstname # devtools::install_github("coccopuffs/GenderGuesser") #missing_gender_result <- GenderGuesser::guessGender(missing_gender) -# write.csv(missing_gender_result, "missing_gender_result.csv", row.names = FALSE) +# write.csv(missing_gender_result, here::here("data-raw","missing_gender_result.csv"), row.names = FALSE) # Results are saved locally to avoid calling too many times the API -missing_gender_result <- readr::read_csv("missing_gender_result.csv") +missing_gender_result <- readr::read_csv(here::here("data-raw","missing_gender_result.csv")) missing_gender_results <- missing_gender_result |> dplyr::filter(!(is.na(gender))) |> @@ -829,8 +829,8 @@ dup <- dplyr::rows_append(dup, matchedA ) dup1 <- dup |> dplyr::distinct() -write.csv(dup1, "dup.csv", row.names = FALSE ) -write.csv(data.prep, "data.prep.csv", row.names = FALSE ) +write.csv(dup1, here::here("data-raw","dup.csv"), row.names = FALSE ) +write.csv(data.prep, here::here("data-raw","data.prep.csv"), row.names = FALSE ) # Examine the EM object: #matches.out$EM diff --git a/fastlink.R b/fastlink.R deleted file mode 100644 index 75c5530..0000000 --- a/fastlink.R +++ /dev/null @@ -1,260 +0,0 @@ -# Clear the workspace by removing all objects -rm(list = ls()) - -# Load the pacman package if it is not already installed. -# if (!requireNamespace("pacman", quietly = TRUE)) { -# install.packages("pacman") -# } -# -# pacman::p_load( -# dplyr, # Data manipulation and transformation -# data.table, # Fast data manipulation with data tables -# stringi, # Character string processing -# lubridate, # Date and time handling -# janitor, # Data cleaning and tabulation functions -# parallel, # Parallel computing -# fastLink, # Record linkage and deduplication -# stringdist # String distance computation -# ) - -# Get the number of detected cores minus 1 -# Reserve one core for non-computational tasks to help prevent system slowdowns or unresponsiveness -numCores <- parallel::detectCores() - 1 - -# Example data frames dfA and dfB -# Replace these with your actual data frames - - -#data(samplematch) - -dfA <- data.frame( - FIRST_NAME = c("John", "Mary", "Robert", "Michael", "Jennifer", "David", "Karen", "Maria", "Carlos", "James"), - LAST_NAME = c("Smith III", "Johnson!", "Williams123", "Brown", "Jones", "Davis", "Miller", "Garcia", "Martinez", "Andrson"), - BIRTH_DATE = c("1981-05-20", "1990-05-15", "1978-12-10", "1985-08-02", "1993-11-25", "1977-03-30", "1988-06-18", "1991-02-05", "1980-09-12", "1982-07-09"), - gender = c("M", "F", "M", "M", "F", "M", "F", "F", "M", "M") -) - -dfB <- data.frame( - RecipientNameFirst = c("John", "Mary", "Robert", "Michael", "Jennifer", "David", "Karenn", "Carloas", "Mariaa", "James"), - RecipientNameLast = c("Smith iv", "Brown-", "Williams", "Jones", "John son", "No Name", "Miller", "Martinez", "Garcia", "Anderson"), - RecipientDateOfBirth = c("1981-05-21", "1992-09-25", "1978-10-12", "1985-08-02", "1993-11-25", "1977-03-30", "1988-06-18", "1980-09-12", "1991-02-05", "1982-07-09"), - gender = c("M", "F", "M", "M", "F", "M", "F", "M", "F", "M") -) - -# # Example data frames dfA and dfB -# dfA <- fread(file = 'dfA.csv', -# sep = ",", -# header = T, -# nThread = numCores) -# -# dfB <- fread(file = 'dfB.csv', -# sep = ",", -# header = T, -# nThread = numCores) - -# Assign a unique key ID to each row -dfA <- dfA |> - dplyr::mutate(row_idA = paste("dfA_", dplyr::row_number(), sep = "")) - -dfB <- dfB |> - dplyr::mutate(row_idB = paste("dfB_", dplyr::row_number(), sep = "")) - -# Convert the date column to a valid date format. -dfA$BIRTH_DATE <- as.Date(dfA$BIRTH_DATE, format = "%Y-%m-%d") # "%m/%d/%Y" -dfB$RecipientDateOfBirth <- as.Date(dfB$RecipientDateOfBirth, format = "%Y-%m-%d") # "%m/%d/%Y" -# -# # Create new date variables by splitting date of birth into three different parts. -dfA <- dfA |> - dplyr::mutate(dob_day = as.numeric(lubridate::day(BIRTH_DATE)), - dob_month = as.numeric(lubridate::month(BIRTH_DATE)), - dob_year = as.numeric(lubridate::year(BIRTH_DATE)), - DOB=BIRTH_DATE) - -dfB <- dfB |> - dplyr::mutate(dob_day = as.numeric(lubridate::day(RecipientDateOfBirth)), - dob_month = as.numeric(lubridate::month(RecipientDateOfBirth)), - dob_year = as.numeric(lubridate::year(RecipientDateOfBirth)), - DOB=RecipientDateOfBirth) - - -# Define a function for data cleaning with additional name removal logic -clean_names <- function(names_column) { - # Step 0: Convert to uppercase - names_column_new <- toupper(names_column) - # Step 1: Remove specified name suffixes - toRemove <- c(" JR", " SR", " IV", " III", " II") - for (tR in toRemove) { - names_column_new <- gsub(tR, "", names_column_new) - } - # Step 2: Convert special characters to ASCII equivalents - names_column_new <- iconv(names_column_new, "latin1", "ASCII//TRANSLIT", sub = "") - # Step 3: Remove punctuation, digits, and all sapces - names_column_new <- gsub("[[:punct:][:digit:]][[:space:]]", "", names_column_new) - # Step 4: Create a new variable with only alphabetic characters - names_column_new <- gsub("[^[:alpha:]]", "", names_column_new) - return(names_column_new) -} - -# Perform data cleaning on dfA using the clean_names function -dfA <- dfA |> - dplyr::mutate_at(dplyr::vars(FIRST_NAME, LAST_NAME), list(new = clean_names)) |> - dplyr::mutate(FN = FIRST_NAME_new, LN = LAST_NAME_new) - -# Perform data cleaning on dfB using the clean_names function -dfB <- dfB |> - dplyr::mutate_at(dplyr::vars(RecipientNameFirst, RecipientNameLast), list(new = clean_names)) |> - dplyr::mutate(FN = RecipientNameFirst_new, LN = RecipientNameLast_new) - - -rm(clean_names) - -# # Create no name list -NoNameList <- c( - "NICKNAME", - "NOFAMILYNAME", - "NOFIRSTNAME", - "NOLASTNAME", - "NOMIDDLENAME", - "NONAME", - "NO", - "UNKNOWN", - "UNK", - "UN", - "NA" -) -# -# # Blank out the names in the data if they match any of the strings in the NoNameList -dfA <- dfA |> - dplyr::mutate(FN = dplyr::case_when( - FN %in% NoNameList~ "", - TRUE ~ FN), - LN = dplyr::case_when( - LN %in% NoNameList ~ "", - TRUE ~ LN)) - -dfB <- dfB |> - dplyr::mutate(FN = dplyr::case_when( - FN %in% NoNameList~ "", - TRUE ~ FN), - LN = dplyr::case_when( - LN %in% NoNameList ~ "", - TRUE ~ LN)) -rm(NoNameList) - -# Delete rows that have missing First Name (FN), Last Name (LN), or Date of Birth (DOB). -dfA <- dfA |> - dplyr::filter(!is.na(FN) & FN != "" & !is.na( LN) & LN != "" & !is.na(DOB) ) - -dfB <- dfB |> - dplyr::filter(!is.na(FN) & FN != "" & !is.na( LN) & LN != "" & !is.na(DOB) ) - - -# Exact Matching -Exact <- merge(dfA, dfB, by=c("FN","LN","DOB", "gender")) - - -# Using the fastLink R package for record linkage -matches.out <- fastLink::fastLink( - dfA = dfA, dfB = dfB, - # Specify the vector of variable names to be used for matching. - # These variable names should exist in both dfA and dfB - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year', 'gender'), - # Specify which variables among varnames should be compared using string distance - stringdist.match = c('FN', 'LN'), - # Specify which variables present in stringdist.match can be partially matched - partial.match = c('FN', 'LN'), - # Specify which variables should be matched numerically - # Must be a subset of 'varnames' and must not be present in 'stringdist.match'. - numeric.match = c('dob_day', 'dob_month', 'dob_year'), - # Specify the number of CPU cores to utilize (parallel processing). The default value is NULL. - n.cores = numCores, - return.all = TRUE, - return.df = TRUE -) - -# The output from fastLink() will be a list that includes a matrix where each row -# is a match with the relevant indices of dfA (column 1) and dfB (column 2). - -# Confusion Matrice -fastLink::confusion(matches.out, threshold = 0.98) - -# Examine the EM object: -matches.out$EM - -# Summarize the accuracy of the match: -# each column gives the match count, match rate, -# false discovery rate (FDR) and false negative rate (FNR) -# under different cutoffs for matches based on the posterior -# probability of a match. -summary(matches.out) - - -# Get fuzzy matches using the results from fastLink -# A threshold of 0.98 is set for match classification -fuzzy_matches <- fastLink::getMatches(dfA, dfB, - fl.out = matches.out, - threshold.match = 0.98) -fuzzy_matches #|> knitr::kable() - -blockdata_out <- fastLink::blockData(dfA, dfB, - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year'), - # window blocking for numeric variables, where a given observation in dataset A - #will be compared to all observations in dataset B where the value of the blocking - #variable is within ±K of the value of the same variable in dataset A. - #The value of K is the size of the window - # window.block = "birthyear", - # window.size = 1, - # using k-means clustering, so that similar values of string and numeric - # variables are blocked together. - kmeans.block = "FN", - nclusters = 2) - - -## Aggregating Multiple Matches Together ### - -#Often, we run several different matches for a single data set - -#for instance, when blocking by gender or by some other criterion to reduce the -#number of pairwise comparisons. Here, we walk through how to aggregate those -#multiple matches into a single summary. -blockgender_out <- fastLink::blockData(dfA, dfB, varnames = "gender") -## Subset dfA into blocks -dfA_block1 <- dfA[blockgender_out$block.1$dfA.inds,] -dfA_block2 <- dfA[blockgender_out$block.2$dfA.inds,] - -## Subset dfB into blocks -dfB_block1 <- dfB[blockgender_out$block.1$dfB.inds,] -dfB_block2 <- dfB[blockgender_out$block.2$dfB.inds,] - -## Run fastLink on each -fl_out_block1 <- fastLink::fastLink( - dfA_block1, dfB_block1, - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year'), - n.cores = numCores -) -fl_out_block2 <- fastLink::fastLink( - dfA_block2, dfB_block2, - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year'), - n.cores = numCores -) - -#Here, we run fastLink() on the subsets of data defined by blocking on gender in the previous section: - -## Run fastLink on each -link.1 <- fastLink::fastLink( - dfA_block1, dfB_block1, - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year'), - n.cores = numCores - ) - -link.2 <- fastLink::fastLink( - dfA_block2, dfB_block2, - varnames = c('FN', 'LN', 'dob_day', 'dob_month', 'dob_year'), - n.cores = numCores -) - -#To aggregate the two matches into a single summary, we use the aggregateEM() function as follows: -agg.out <- fastLink::aggregateEM(em.list = list(link.1, link.2)) - -summary(agg.out ) - - diff --git a/img/Record-linkage-example.png b/img/Record-linkage-example.png new file mode 100644 index 0000000..db03d3e Binary files /dev/null and b/img/Record-linkage-example.png differ diff --git a/concept.png b/img/concept.png similarity index 100% rename from concept.png rename to img/concept.png diff --git a/evaluation.png b/img/evaluation.png similarity index 100% rename from evaluation.png rename to img/evaluation.png diff --git a/pipe.png b/img/pipe.png similarity index 100% rename from pipe.png rename to img/pipe.png diff --git a/img/prob.png b/img/prob.png new file mode 100644 index 0000000..69c3afe Binary files /dev/null and b/img/prob.png differ diff --git a/record-linkage.png b/img/record-linkage.png similarity index 100% rename from record-linkage.png rename to img/record-linkage.png diff --git a/img/record-linkage.webp b/img/record-linkage.webp new file mode 100644 index 0000000..b120aa3 Binary files /dev/null and b/img/record-linkage.webp differ diff --git a/index.Rmd b/index.Rmd index ba2842d..dfbc542 100644 --- a/index.Rmd +++ b/index.Rmd @@ -71,7 +71,7 @@ Can be numerous ones: ... ] .pull-right[ -![](record-linkage.png) +![](img/record-linkage.png) ] @@ -94,7 +94,7 @@ The recommended workflow includes to go through specific steps: ] .pull-right[ -![](concept.png) +![](img/concept.png) ] ??? @@ -362,7 +362,7 @@ Revise the[demo script](FastLink.html) and [applied real use case notebook](htt ] .pull-right[ -![](pipe.png) +![](img/pipe.png) ] diff --git a/index.html b/index.html index 767431e..7cde382 100644 --- a/index.html +++ b/index.html @@ -1339,6 +1339,10 @@ --unhcr-logoblue: url("data:image/svg+xml;base64,<svg id="Layer_1" data-name="Layer 1" xmlns="http://www.w3.org/2000/svg" width="225.9" height="54.3" viewBox="0 0 225.9 54.3"><defs><style>.cls-1{fill:#0072bc;}</style></defs><path class="cls-1" d="M33,36.7c.6,0,.6-.4.6-1V32.4c0-.7-.1-.9.3-.9,1.4,0,1.7-.1,1.7-1.3V17.9c0-3.3-2.4-2.8-3-4-1-1.6,1.8-1.6.8-5.2a2,2,0,0,0-2.1-1.3,2,2,0,0,0-2.1,1.3c-1,3.6,1.7,3.6.8,5.2-.7,1.2-3,.7-3,4V30.2c0,1.2.4,1.3,1.7,1.3.4,0,.3.2.3.9v3.3c0,.7-.1,1,.6,1H33"/><path class="cls-1" d="M22.3,10.1c-1.2,1.3-3.1,4.3-1.9,6.8,2.5,1,2.3-8,5.5-7.9,1.6,1.4-.5,5.7-1.1,7.5-.8,2.3-1.4,7.1-2.9,9.5-1.2,2-.3,8.1-.7,10.2-1,1-3.9.3-5.1.1a81.76,81.76,0,0,0-.7-8.2c0-.8-.9-12.7-.2-14.1C16.6,11,24,5.3,25,4.2S29,0,30.3,0c.9.6.4,2,.2,2.5-1.2,3-6.5,6-8.2,7.6"/><path class="cls-1" d="M40.4,10.1c1.2,1.3,3.1,4.3,1.9,6.8-2.6,1-2.4-8-5.5-7.9-1.5,1.4.5,5.7,1.1,7.5.8,2.3,1.4,7.1,2.9,9.5,1.2,2,.3,8.1.7,10.2,1.1,1,3.9.3,5.1.1a75.58,75.58,0,0,1,.7-8.2c.1-.8.9-12.7.2-14.1-1.4-3-8.8-8.7-9.8-9.8S33.6,0,32.4,0c-.9.6-.3,2-.2,2.5,1.3,3,6.5,6,8.2,7.6"/><path class="cls-1" d="M11.5,6.1c-.7.6-2.1,1.1-3.2,2.4-.4-.7.6-2.5,1.8-3.6a17.68,17.68,0,0,1,3.2-2.1c.4.9-1,2.3-1.8,3.3"/><path class="cls-1" d="M9.4,11c-1.1,1-2.3,1.5-3.9,3.4a12.22,12.22,0,0,1-.2-5.7c.6-1.8,1.5-3.8,3-4.4a15.69,15.69,0,0,0-.9,4.3c-.7,2.6-1,2.7-.9,3,.2,0,1-1.4,2.1-2.4C9.8,8,11.4,7.7,12.4,6.9a17.3,17.3,0,0,1-3,4.1"/><path class="cls-1" d="M3.9,21.8a14.31,14.31,0,0,1-1.8-7c.1-1.6.9-5.3,2-5.9-.3,3.4.6,3.7.2,6.4,0,.1-.4,3-.4,3.2.5-.1,0-2.6,5.4-6.4-1.2,2-.6,2.6-2.7,5.3a10.74,10.74,0,0,0-2.7,4.4"/><path class="cls-1" d="M6.1,24.3a6.63,6.63,0,0,0-1.2,3.2C2.2,26-.9,19.8.2,16.7c.7,2.3,1.7,2.9,2.4,4.8,1.2,2.9,1,3.8,1.8,4.5a9,9,0,0,1,.9-4.6,32.07,32.07,0,0,1,1.9-3.1c.5,2.1.1,3.8-1.1,6"/><path class="cls-1" d="M8.5,29.6a46,46,0,0,1-.6,5.2C3.3,33.3.8,28,.8,25.4c1.4,3.2,3.5,3.2,6.2,7.7.2-1.7-.5-.4-.6-3.7.1-3.4.8-3.5.9-5.6.9.9,1.2,4.7,1.2,5.8"/><path class="cls-1" d="M12.3,36.4c.4,1.8.5,3.8.8,5.3-2.7-1-3.3-.2-6.5-2.2-1.4-.8-4-4.1-3.9-6.4,3.1,3.5,3.7,3.1,5.2,4.4,1.4,1.1,2.5,2.8,3.6,2.6.1-.9-2.3-3-2-8.8a14.15,14.15,0,0,1,2.8,5.1"/><path class="cls-1" d="M17.6,40.8c.8,1.2.6,1.7,2.7,4.1A25.54,25.54,0,0,1,14.8,46c-1.3,0-6-1.2-7.6-4.4,2.7,1.4,2.7,1.1,6.3,1.9,2.6.6,4.4,1.4,5.1,1.3-.1-.3-2.1-1.2-3.5-2.9s-1-3.9-2.3-5.4c.9-.4,4.1,3.4,4.8,4.3"/><path class="cls-1" d="M51,6.1c.6.6,2.2,1.1,3.1,2.4.4-.7-.6-2.5-1.8-3.6a14.77,14.77,0,0,0-3.2-2.1c-.4.9,1,2.3,1.9,3.3"/><path class="cls-1" d="M53.1,11c1.1,1,2.3,1.5,3.9,3.4a12.22,12.22,0,0,0,.2-5.7c-.5-1.8-1.5-3.8-3-4.4a15.69,15.69,0,0,1,.9,4.3c.7,2.6,1,2.7.9,3-.2,0-1-1.4-2.1-2.4C52.7,8,51.2,7.7,50.1,6.9a17.3,17.3,0,0,0,3,4.1"/><path class="cls-1" d="M58.5,21.8a14,14,0,0,0,1.8-7c-.1-1.6-.9-5.3-2-5.9.3,3.4-.6,3.7-.2,6.4,0,.1.3,3,.3,3.2-.4-.1,0-2.6-5.3-6.4,1.1,2,.6,2.6,2.7,5.3a9.28,9.28,0,0,1,2.7,4.4"/><path class="cls-1" d="M56.4,24.3a9.58,9.58,0,0,1,1.2,3.2c2.7-1.6,5.8-7.8,4.6-10.8-.7,2.3-1.7,2.9-2.3,4.8-1.3,2.9-1.1,3.8-1.8,4.5a11.79,11.79,0,0,0-.9-4.6,19.82,19.82,0,0,0-1.9-3.1,7.51,7.51,0,0,0,1.1,6"/><path class="cls-1" d="M54,29.6a30.92,30.92,0,0,0,.7,5.2c4.6-1.5,7.1-6.8,7.1-9.4-1.5,3.2-3.5,3.2-6.2,7.7-.2-1.7.5-.4.5-3.7s-.8-3.5-.9-5.6c-.9.9-1.2,4.7-1.2,5.8"/><path class="cls-1" d="M50.1,36.4c-.4,1.8-.5,3.8-.7,5.3,2.6-1,3.3-.2,6.4-2.2,1.4-.8,4.1-4.1,3.9-6.4-3.1,3.5-3.7,3.1-5.2,4.4-1.4,1.1-2.6,2.8-3.6,2.6-.1-.9,2.3-3,2-8.8a15.84,15.84,0,0,0-2.8,5.1"/><path class="cls-1" d="M44.9,40.8a16.19,16.19,0,0,1-2.7,4.1A26.81,26.81,0,0,0,47.7,46c1.3,0,6-1.2,7.5-4.4-2.7,1.4-2.7,1.1-6.3,1.9-2.6.6-4.4,1.4-5.1,1.3.2-.3,2.2-1.2,3.5-2.9s1-3.9,2.3-5.4c-.9-.4-4,3.4-4.7,4.3"/><path class="cls-1" d="M24.6,51.8l-.8-.4a17.2,17.2,0,0,1,13.5-5.6c8.5.5,7.9,2.5,12.5,2.1.5,0-.4.6-.4.6a15.56,15.56,0,0,1-10-.2c-5.5-1.8-9.1-2.9-14.8,3.5"/><path class="cls-1" d="M38,51.8l.8-.4a17,17,0,0,0-13.4-5.6c-8.5.5-7.9,2.5-12.5,2.1-.5,0,.4.6.4.6a15.56,15.56,0,0,0,10-.2c5.5-1.8,9-2.9,14.7,3.5"/><path class="cls-1" d="M98.3,25.5c0,8.1-4.4,12.1-13.4,12.1S71.5,33.7,71.5,25.5V7.5h9v16c0,3,0,6.8,4.4,6.8s4.3-3.8,4.3-6.8V7.5h9.1v18"/><polygon class="cls-1" points="103.7 7.4 113 7.4 121.6 23.1 121.6 23.1 121.6 7.4 130.2 7.4 130.2 36.8 121.4 36.8 112.4 20.8 112.3 20.8 112.3 36.8 103.7 36.8 103.7 7.4"/><polygon class="cls-1" points="135.8 7.4 144.8 7.4 144.8 17.6 153.3 17.6 153.3 7.4 162.3 7.4 162.3 36.8 153.3 36.8 153.3 25.2 144.8 25.2 144.8 36.8 135.8 36.8 135.8 7.4"/><path class="cls-1" d="M185.5,18.4a4.69,4.69,0,0,0-4.7-4.3c-4.2,0-5.6,4.1-5.6,8.1s1.4,8.1,5.6,8.1c3,0,4.2-2.1,4.8-4.9h8.9c0,5.9-4.8,12.2-13.4,12.2-9.5,0-14.9-6.7-14.9-15.4,0-9.3,5.8-15.4,14.9-15.4,8.1,0,12.5,4.3,13.3,11.6h-8.9"/><path class="cls-1" d="M207.3,14.4h4.2c3.5,0,4.5,1.1,4.5,2.9,0,2.6-2.3,2.9-4,2.9h-4.7Zm-9.1,22.4h9.1V26.6h4.8c3.6,0,3.8,2.9,4.1,5.7a23.38,23.38,0,0,0,.7,4.5h9c-.8-1.4-.9-4.9-1-6.3-.3-3.7-1.9-6.4-4.3-7.3,3-1.1,4.4-4.3,4.4-7.4,0-5.6-4.4-8.4-9.6-8.4H198.2V36.8"/><polygon class="cls-1" points="74.5 43.6 71.5 43.6 71.5 41.7 79.6 41.7 79.6 43.6 76.6 43.6 76.6 51.6 74.5 51.6 74.5 43.6"/><path class="cls-1" d="M80.3,41.7h2v3.7h0a2.44,2.44,0,0,1,2.1-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4H85v-4c0-1.2-.3-1.8-1.3-1.8s-1.5.6-1.5,2v3.7h-2l.1-9.7"/><path class="cls-1" d="M89.7,48.5q.15,1.8,1.8,1.8a1.65,1.65,0,0,0,1.5-.9h1.7a3.22,3.22,0,0,1-3.3,2.4,3.46,3.46,0,0,1-3.6-3.3v-.4a3.59,3.59,0,0,1,3.4-3.8h.3c2.4,0,3.6,2,3.5,4.2Zm3.2-1.2c-.2-1-.6-1.5-1.6-1.5a1.55,1.55,0,0,0-1.6,1.4v.1h3.2"/><path class="cls-1" d="M107.6,47.9c0,2.7-1.6,4-4.2,4s-4.2-1.3-4.2-4V41.8h2.2v6.1c0,1.1.3,2.1,2,2.1,1.5,0,2-.7,2-2.1V41.8h2.2v6.1"/><polygon class="cls-1" points="109 41.7 111.2 41.7 115.3 48.3 115.3 48.3 115.3 41.7 117.4 41.7 117.4 51.6 115.2 51.6 111.1 45 111.1 45 111.1 51.6 109 51.6 109 41.7"/><path class="cls-1" d="M124.2,43.4h2.4c1,0,1.5.4,1.5,1.4s-.5,1.4-1.5,1.4h-2.4Zm-2.1,8.2h2.2V47.8h2.2c1.1,0,1.5.5,1.6,1.5a6.72,6.72,0,0,0,.4,2.4h2.2a5.87,5.87,0,0,1-.4-2.3c-.1-1-.4-2-1.4-2.3h0a2.5,2.5,0,0,0,1.6-2.5,2.69,2.69,0,0,0-2.7-2.7h-5.5l-.2,9.7"/><path class="cls-1" d="M133.2,47.3a1.54,1.54,0,0,1,1.5-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.49,1.49,0,0,1-1.5.9q-1.65,0-1.8-1.8l5.2.1"/><path class="cls-1" d="M139.6,45.8h-1.2V44.5h1.2V44a2.05,2.05,0,0,1,2-2.2h.4a3.08,3.08,0,0,1,1,.1v1.5h-.7c-.5,0-.7.2-.7.8v.4H143v1.3h-1.4v5.8h-2V45.8"/><path class="cls-1" d="M150.2,51.6h-1.9v-1h0a2.75,2.75,0,0,1-2.2,1.2c-2.1,0-2.6-1.2-2.6-2.9V44.5h2v4c0,1.2.4,1.8,1.3,1.8,1.1,0,1.5-.6,1.5-2V44.5h2v7.1"/><path class="cls-1" d="M154.7,49.9c-1.2,0-1.6-1.1-1.6-2.1s.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2a1.73,1.73,0,0,1-1.5,2c-.1,0-.1,0-.2-.1m3.6-5.4h-1.9v1h0a2,2,0,0,0-2.1-1.1,3.21,3.21,0,0,0-3.1,3.5c0,1.8.9,3.6,3,3.6a2.36,2.36,0,0,0,2.1-1.1h0v.9c0,1-.5,1.8-1.6,1.8a1.33,1.33,0,0,1-1.5-1h-1.9c.1,1.6,1.7,2.3,3.1,2.3,3.3,0,3.7-2,3.7-3.2l.2-6.7"/><path class="cls-1" d="M161.2,47.3a1.54,1.54,0,0,1,1.5-1.5h.1c1,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.49,1.49,0,0,1-1.5.9q-1.65,0-1.8-1.8l5.2.1"/><path class="cls-1" d="M168.9,47.3a1.5,1.5,0,0,1,1.6-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.15,3.15,0,0,0,3.3-2.4h-1.7a1.54,1.54,0,0,1-1.6.9q-1.65,0-1.8-1.8l5.3.1"/><path class="cls-1" d="M183,44.2h0l1.2,3.6h-2.6Zm-4.8,7.4h2.2l.8-2.2h3.7l.7,2.2h2.3l-3.7-9.8H182Z"/><path class="cls-1" d="M191.5,49.9c-1.2,0-1.6-1.1-1.6-2.1s.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2a1.8,1.8,0,0,1-1.5,2c-.1,0-.2,0-.2-.1m3.5-5.4h-1.9v1h0a2.09,2.09,0,0,0-2.1-1.1,3.21,3.21,0,0,0-3.1,3.5c0,1.8.9,3.6,3.1,3.6a2.36,2.36,0,0,0,2.1-1.1h0v.9c0,1-.5,1.8-1.6,1.8a1.33,1.33,0,0,1-1.5-1h-1.9c.1,1.6,1.7,2.3,3.1,2.3,3.3,0,3.7-2,3.7-3.2l.1-6.7"/><path class="cls-1" d="M197.9,47.3a1.5,1.5,0,0,1,1.6-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.2,1.2c.1-2.2-1-4.2-3.5-4.2a3.56,3.56,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.54,1.54,0,0,1-1.6.9q-1.65,0-1.8-1.8l5.3.1"/><path class="cls-1" d="M203.9,44.5h1.9v1h.1a2.75,2.75,0,0,1,2.2-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-.3-1.8-1.3-1.8s-1.5.6-1.5,2v3.7h-2v-7"/><path class="cls-1" d="M216.6,47a1.32,1.32,0,0,0-1.4-1.2h0c-1.3,0-1.7,1.3-1.7,2.3s.4,2.2,1.6,2.2a1.62,1.62,0,0,0,1.6-1.4h1.9a3.08,3.08,0,0,1-3.4,2.9h0a3.46,3.46,0,0,1-3.6-3.3v-.3a3.43,3.43,0,0,1,3.2-3.8h.5a3.12,3.12,0,0,1,3.4,2.7l-2.1-.1"/><path class="cls-1" d="M223,52.5a2.21,2.21,0,0,1-2.5,1.6,4.87,4.87,0,0,1-1.2-.1V52.4c.4,0,.8.1,1.1.1a1.08,1.08,0,0,0,.8-1.1.37.37,0,0,0-.1-.3l-2.5-6.7h2.1l1.6,4.9h0l1.6-4.9h2L223,52.5"/></svg>"); --unhcr-logowhite: url("data:image/svg+xml;base64,<?xml version="1.0" encoding="utf-8"?>
<!-- Generator: Adobe Illustrator 25.2.0, SVG Export Plug-In . SVG Version: 6.00 Build 0)  -->
<svg version="1.1" id="Layer_1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" x="0px" y="0px"
	 viewBox="0 0 225.9 54.3" style="enable-background:new 0 0 225.9 54.3;" xml:space="preserve">
<style type="text/css">
	.st0{fill:#FFFFFF;}
</style>
<path class="st0" d="M33,36.7c0.6,0,0.6-0.4,0.6-1v-3.3c0-0.7-0.1-0.9,0.3-0.9c1.4,0,1.7-0.1,1.7-1.3V17.9c0-3.3-2.4-2.8-3-4
	c-1-1.6,1.8-1.6,0.8-5.2c-0.3-0.9-1.2-1.4-2.1-1.3c-0.9-0.1-1.8,0.4-2.1,1.3c-1,3.6,1.7,3.6,0.8,5.2c-0.7,1.2-3,0.7-3,4v12.3
	c0,1.2,0.4,1.3,1.7,1.3c0.4,0,0.3,0.2,0.3,0.9v3.3c0,0.7-0.1,1,0.6,1H33"/>
<path class="st0" d="M22.3,10.1c-1.2,1.3-3.1,4.3-1.9,6.8c2.5,1,2.3-8,5.5-7.9c1.6,1.4-0.5,5.7-1.1,7.5c-0.8,2.3-1.4,7.1-2.9,9.5
	c-1.2,2-0.3,8.1-0.7,10.2c-1,1-3.9,0.3-5.1,0.1c-0.1-2.7-0.3-5.5-0.7-8.2c0-0.8-0.9-12.7-0.2-14.1c1.4-3,8.8-8.7,9.8-9.8
	S29,0,30.3,0c0.9,0.6,0.4,2,0.2,2.5C29.3,5.5,24,8.5,22.3,10.1"/>
<path class="st0" d="M40.4,10.1c1.2,1.3,3.1,4.3,1.9,6.8c-2.6,1-2.4-8-5.5-7.9c-1.5,1.4,0.5,5.7,1.1,7.5c0.8,2.3,1.4,7.1,2.9,9.5
	c1.2,2,0.3,8.1,0.7,10.2c1.1,1,3.9,0.3,5.1,0.1c0.1-2.7,0.3-5.5,0.7-8.2c0.1-0.8,0.9-12.7,0.2-14.1c-1.4-3-8.8-8.7-9.8-9.8
	S33.6,0,32.4,0c-0.9,0.6-0.3,2-0.2,2.5C33.5,5.5,38.7,8.5,40.4,10.1"/>
<path class="st0" d="M11.5,6.1c-0.7,0.6-2.1,1.1-3.2,2.4C7.9,7.8,8.9,6,10.1,4.9c1-0.8,2.1-1.5,3.2-2.1C13.7,3.7,12.3,5.1,11.5,6.1"
	/>
<path class="st0" d="M9.4,11c-1.1,1-2.3,1.5-3.9,3.4C5,12.5,4.9,10.6,5.3,8.7c0.6-1.8,1.5-3.8,3-4.4C7.8,5.7,7.5,7.1,7.4,8.6
	c-0.7,2.6-1,2.7-0.9,3c0.2,0,1-1.4,2.1-2.4c1.2-1.2,2.8-1.5,3.8-2.3C11.6,8.4,10.6,9.8,9.4,11"/>
<path class="st0" d="M3.9,21.8c-1.2-2.1-1.8-4.6-1.8-7c0.1-1.6,0.9-5.3,2-5.9c-0.3,3.4,0.6,3.7,0.2,6.4c0,0.1-0.4,3-0.4,3.2
	c0.5-0.1,0-2.6,5.4-6.4c-1.2,2-0.6,2.6-2.7,5.3C5.4,18.6,4.4,20.1,3.9,21.8"/>
<path class="st0" d="M6.1,24.3c-0.7,0.9-1.1,2-1.2,3.2C2.2,26-0.9,19.8,0.2,16.7c0.7,2.3,1.7,2.9,2.4,4.8c1.2,2.9,1,3.8,1.8,4.5
	c-0.1-1.6,0.2-3.2,0.9-4.6c0.6-1.1,1.2-2.1,1.9-3.1C7.7,20.4,7.3,22.1,6.1,24.3"/>
<path class="st0" d="M8.5,29.6c-0.1,1.7-0.3,3.5-0.6,5.2c-4.6-1.5-7.1-6.8-7.1-9.4c1.4,3.2,3.5,3.2,6.2,7.7c0.2-1.7-0.5-0.4-0.6-3.7
	c0.1-3.4,0.8-3.5,0.9-5.6C8.2,24.7,8.5,28.5,8.5,29.6"/>
<path class="st0" d="M12.3,36.4c0.4,1.8,0.5,3.8,0.8,5.3c-2.7-1-3.3-0.2-6.5-2.2c-1.4-0.8-4-4.1-3.9-6.4c3.1,3.5,3.7,3.1,5.2,4.4
	c1.4,1.1,2.5,2.8,3.6,2.6c0.1-0.9-2.3-3-2-8.8C10.8,32.8,11.7,34.5,12.3,36.4"/>
<path class="st0" d="M17.6,40.8c0.8,1.2,0.6,1.7,2.7,4.1c-1.8,0.6-3.6,0.9-5.5,1.1c-1.3,0-6-1.2-7.6-4.4c2.7,1.4,2.7,1.1,6.3,1.9
	c2.6,0.6,4.4,1.4,5.1,1.3c-0.1-0.3-2.1-1.2-3.5-2.9s-1-3.9-2.3-5.4C13.7,36.1,16.9,39.9,17.6,40.8"/>
<path class="st0" d="M51,6.1c0.6,0.6,2.2,1.1,3.1,2.4c0.4-0.7-0.6-2.5-1.8-3.6c-1-0.8-2-1.5-3.2-2.1C48.7,3.7,50.1,5.1,51,6.1"/>
<path class="st0" d="M53.1,11c1.1,1,2.3,1.5,3.9,3.4c0.5-1.9,0.6-3.8,0.2-5.7c-0.5-1.8-1.5-3.8-3-4.4c0.5,1.4,0.8,2.8,0.9,4.3
	c0.7,2.6,1,2.7,0.9,3c-0.2,0-1-1.4-2.1-2.4c-1.2-1.2-2.7-1.5-3.8-2.3C50.9,8.4,51.9,9.8,53.1,11"/>
<path class="st0" d="M58.5,21.8c1.2-2.1,1.8-4.5,1.8-7c-0.1-1.6-0.9-5.3-2-5.9c0.3,3.4-0.6,3.7-0.2,6.4c0,0.1,0.3,3,0.3,3.2
	c-0.4-0.1,0-2.6-5.3-6.4c1.1,2,0.6,2.6,2.7,5.3C57.1,18.6,58,20.1,58.5,21.8"/>
<path class="st0" d="M56.4,24.3c0.6,1,1,2.1,1.2,3.2c2.7-1.6,5.8-7.8,4.6-10.8c-0.7,2.3-1.7,2.9-2.3,4.8c-1.3,2.9-1.1,3.8-1.8,4.5
	c0-1.6-0.3-3.1-0.9-4.6c-0.5-1.1-1.2-2.1-1.9-3.1C54.8,20.4,55.2,22.5,56.4,24.3"/>
<path class="st0" d="M54,29.6c0.1,1.8,0.3,3.5,0.7,5.2c4.6-1.5,7.1-6.8,7.1-9.4c-1.5,3.2-3.5,3.2-6.2,7.7c-0.2-1.7,0.5-0.4,0.5-3.7
	s-0.8-3.5-0.9-5.6C54.3,24.7,54,28.5,54,29.6"/>
<path class="st0" d="M50.1,36.4c-0.4,1.8-0.5,3.8-0.7,5.3c2.6-1,3.3-0.2,6.4-2.2c1.4-0.8,4.1-4.1,3.9-6.4c-3.1,3.5-3.7,3.1-5.2,4.4
	c-1.4,1.1-2.6,2.8-3.6,2.6c-0.1-0.9,2.3-3,2-8.8C51.7,32.8,50.7,34.5,50.1,36.4"/>
<path class="st0" d="M44.9,40.8c-0.7,1.5-1.6,2.9-2.7,4.1c1.8,0.6,3.6,0.9,5.5,1.1c1.3,0,6-1.2,7.5-4.4c-2.7,1.4-2.7,1.1-6.3,1.9
	c-2.6,0.6-4.4,1.4-5.1,1.3c0.2-0.3,2.2-1.2,3.5-2.9s1-3.9,2.3-5.4C48.7,36.1,45.6,39.9,44.9,40.8"/>
<path class="st0" d="M24.6,51.8l-0.8-0.4c3.4-3.8,8.4-5.8,13.5-5.6c8.5,0.5,7.9,2.5,12.5,2.1c0.5,0-0.4,0.6-0.4,0.6
	c-3.3,1-6.8,1-10-0.2C33.9,46.5,30.3,45.4,24.6,51.8"/>
<path class="st0" d="M38,51.8l0.8-0.4c-3.4-3.8-8.3-5.8-13.4-5.6c-8.5,0.5-7.9,2.5-12.5,2.1c-0.5,0,0.4,0.6,0.4,0.6
	c3.3,1,6.8,1,10-0.2C28.8,46.5,32.3,45.4,38,51.8"/>
<path class="st0" d="M98.3,25.5c0,8.1-4.4,12.1-13.4,12.1s-13.4-3.9-13.4-12.1v-18h9v16c0,3,0,6.8,4.4,6.8s4.3-3.8,4.3-6.8v-16h9.1
	V25.5"/>
<polygon class="st0" points="103.7,7.4 113,7.4 121.6,23.1 121.6,23.1 121.6,7.4 130.2,7.4 130.2,36.8 121.4,36.8 112.4,20.8 
	112.3,20.8 112.3,36.8 103.7,36.8 "/>
<polygon class="st0" points="135.8,7.4 144.8,7.4 144.8,17.6 153.3,17.6 153.3,7.4 162.3,7.4 162.3,36.8 153.3,36.8 153.3,25.2 
	144.8,25.2 144.8,36.8 135.8,36.8 "/>
<path class="st0" d="M185.5,18.4c-0.2-2.4-2.3-4.3-4.7-4.3c-4.2,0-5.6,4.1-5.6,8.1s1.4,8.1,5.6,8.1c3,0,4.2-2.1,4.8-4.9h8.9
	c0,5.9-4.8,12.2-13.4,12.2c-9.5,0-14.9-6.7-14.9-15.4c0-9.3,5.8-15.4,14.9-15.4c8.1,0,12.5,4.3,13.3,11.6L185.5,18.4"/>
<path class="st0" d="M207.3,14.4h4.2c3.5,0,4.5,1.1,4.5,2.9c0,2.6-2.3,2.9-4,2.9h-4.7V14.4z M198.2,36.8h9.1V26.6h4.8
	c3.6,0,3.8,2.9,4.1,5.7c0.1,1.5,0.3,3,0.7,4.5h9c-0.8-1.4-0.9-4.9-1-6.3c-0.3-3.7-1.9-6.4-4.3-7.3c3-1.1,4.4-4.3,4.4-7.4
	c0-5.6-4.4-8.4-9.6-8.4h-17.2V36.8"/>
<polygon class="st0" points="74.5,43.6 71.5,43.6 71.5,41.7 79.6,41.7 79.6,43.6 76.6,43.6 76.6,51.6 74.5,51.6 "/>
<path class="st0" d="M80.3,41.7h2v3.7l0,0c0.4-0.7,1.2-1.2,2.1-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-0.3-1.8-1.3-1.8
	s-1.5,0.6-1.5,2v3.7h-2L80.3,41.7"/>
<path class="st0" d="M89.7,48.5c0.1,1.2,0.7,1.8,1.8,1.8c0.6,0,1.2-0.3,1.5-0.9h1.7c-0.4,1.5-1.8,2.5-3.3,2.4
	c-1.9,0.1-3.5-1.4-3.6-3.3v-0.4c-0.1-2,1.4-3.7,3.4-3.8c0,0,0,0,0,0h0.3c2.4,0,3.6,2,3.5,4.2H89.7z M92.9,47.3
	c-0.2-1-0.6-1.5-1.6-1.5c-0.8,0-1.5,0.6-1.6,1.4v0.1H92.9"/>
<path class="st0" d="M107.6,47.9c0,2.7-1.6,4-4.2,4s-4.2-1.3-4.2-4v-6.1h2.2v6.1c0,1.1,0.3,2.1,2,2.1c1.5,0,2-0.7,2-2.1v-6.1h2.2
	V47.9"/>
<polygon class="st0" points="109,41.7 111.2,41.7 115.3,48.3 115.3,48.3 115.3,41.7 117.4,41.7 117.4,51.6 115.2,51.6 111.1,45 
	111.1,45 111.1,51.6 109,51.6 "/>
<path class="st0" d="M124.2,43.4h2.4c1,0,1.5,0.4,1.5,1.4s-0.5,1.4-1.5,1.4h-2.4L124.2,43.4z M122.1,51.6h2.2v-3.8h2.2
	c1.1,0,1.5,0.5,1.6,1.5c0,0.8,0.1,1.6,0.4,2.4h2.2c-0.3-0.7-0.4-1.5-0.4-2.3c-0.1-1-0.4-2-1.4-2.3l0,0c1-0.4,1.7-1.4,1.6-2.5
	c0-1.5-1.2-2.7-2.7-2.7c0,0,0,0,0,0h-5.5L122.1,51.6"/>
<path class="st0" d="M133.2,47.3c0-0.8,0.7-1.5,1.5-1.5h0.1c0.9,0,1.4,0.5,1.6,1.5H133.2z M138.3,48.5c0.1-2.2-1-4.2-3.5-4.2
	c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.5,0.9c-1.1,0-1.7-0.6-1.8-1.8L138.3,48.5"/>
<path class="st0" d="M139.6,45.8h-1.2v-1.3h1.2V44c-0.1-1.1,0.8-2.1,1.9-2.2c0,0,0.1,0,0.1,0h0.4c0.3,0,0.7,0,1,0.1v1.5h-0.7
	c-0.5,0-0.7,0.2-0.7,0.8v0.4h1.4v1.3h-1.4v5.8h-2V45.8"/>
<path class="st0" d="M150.2,51.6h-1.9v-1l0,0c-0.5,0.7-1.3,1.2-2.2,1.2c-2.1,0-2.6-1.2-2.6-2.9v-4.4h2v4c0,1.2,0.4,1.8,1.3,1.8
	c1.1,0,1.5-0.6,1.5-2v-3.8h2v7.1"/>
<path class="st0" d="M154.7,49.9c-1.2,0-1.6-1.1-1.6-2.1s0.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2c0.2,0.9-0.5,1.8-1.4,2c0,0-0.1,0-0.1,0
	C154.8,50,154.8,50,154.7,49.9 M158.3,44.5h-1.9v1l0,0c-0.4-0.8-1.2-1.2-2.1-1.1c-1.8,0.1-3.2,1.5-3.1,3.3c0,0.1,0,0.1,0,0.2
	c0,1.8,0.9,3.6,3,3.6c0.8,0,1.6-0.4,2.1-1.1l0,0v0.9c0,1-0.5,1.8-1.6,1.8c-0.7,0.1-1.3-0.3-1.5-1h-1.9c0.1,1.6,1.7,2.3,3.1,2.3
	c3.3,0,3.7-2,3.7-3.2L158.3,44.5"/>
<path class="st0" d="M161.2,47.3c0-0.8,0.7-1.5,1.5-1.5h0.1c1,0,1.4,0.5,1.6,1.5H161.2z M166.3,48.5c0.1-2.2-1-4.2-3.5-4.2
	c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.5,0.9c-1.1,0-1.7-0.6-1.8-1.8L166.3,48.5"/>
<path class="st0" d="M168.9,47.3c0-0.8,0.7-1.5,1.5-1.5c0,0,0.1,0,0.1,0h0.1c0.9,0,1.4,0.5,1.6,1.5H168.9z M174,48.5
	c0.1-2.2-1-4.2-3.5-4.2c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4
	c1.5,0.1,2.9-0.9,3.3-2.4h-1.7c-0.3,0.6-0.9,1-1.6,0.9c-1.1,0-1.7-0.6-1.8-1.8L174,48.5"/>
<path class="st0" d="M183,44.2L183,44.2l1.2,3.6h-2.6L183,44.2z M178.2,51.6h2.2l0.8-2.2h3.7l0.7,2.2h2.3l-3.7-9.8H182L178.2,51.6z"
	/>
<path class="st0" d="M191.5,49.9c-1.2,0-1.6-1.1-1.6-2.1s0.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2c0.1,1-0.5,1.8-1.5,2
	C191.6,50,191.5,50,191.5,49.9 M195,44.5h-1.9v1l0,0c-0.4-0.8-1.2-1.2-2.1-1.1c-1.8,0.1-3.2,1.5-3.1,3.3c0,0.1,0,0.1,0,0.2
	c0,1.8,0.9,3.6,3.1,3.6c0.8,0,1.6-0.4,2.1-1.1l0,0v0.9c0,1-0.5,1.8-1.6,1.8c-0.7,0.1-1.3-0.3-1.5-1h-1.9c0.1,1.6,1.7,2.3,3.1,2.3
	c3.3,0,3.7-2,3.7-3.2L195,44.5"/>
<path class="st0" d="M197.9,47.3c0-0.8,0.7-1.5,1.5-1.5c0,0,0.1,0,0.1,0h0.1c0.9,0,1.4,0.5,1.6,1.5H197.9z M203.1,48.5
	c0.1-2.2-1-4.2-3.5-4.2c-2,0-3.6,1.5-3.6,3.5V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.6,0.9c-1.1,0-1.7-0.6-1.8-1.8L203.1,48.5"/>
<path class="st0" d="M203.9,44.5h1.9v1h0.1c0.5-0.7,1.3-1.2,2.2-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-0.3-1.8-1.3-1.8
	s-1.5,0.6-1.5,2v3.7h-2V44.5"/>
<path class="st0" d="M216.6,47c-0.1-0.7-0.7-1.2-1.4-1.2l0,0c-1.3,0-1.7,1.3-1.7,2.3s0.4,2.2,1.6,2.2c0.8,0,1.5-0.6,1.6-1.4h1.9
	c-0.1,1.7-1.5,3-3.2,2.9c-0.1,0-0.1,0-0.2,0l0,0c-1.9,0.1-3.5-1.4-3.6-3.3v-0.3c-0.2-1.9,1.1-3.6,3-3.8c0.1,0,0.1,0,0.2,0h0.5
	c1.7-0.2,3.2,1,3.4,2.7L216.6,47"/>
<path class="st0" d="M223,52.5c-0.3,1.1-1.4,1.8-2.5,1.6c-0.4,0-0.8,0-1.2-0.1v-1.6c0.4,0,0.8,0.1,1.1,0.1c0.5-0.1,0.8-0.6,0.8-1.1
	c0-0.1,0-0.2-0.1-0.3l-2.5-6.7h2.1l1.6,4.9l0,0l1.6-4.9h2L223,52.5"/>
</svg>
"); } +:root { +--unhcr-logoblue: url("data:image/svg+xml;base64,<svg id="Layer_1" data-name="Layer 1" xmlns="http://www.w3.org/2000/svg" width="225.9" height="54.3" viewBox="0 0 225.9 54.3"><defs><style>.cls-1{fill:#0072bc;}</style></defs><path class="cls-1" d="M33,36.7c.6,0,.6-.4.6-1V32.4c0-.7-.1-.9.3-.9,1.4,0,1.7-.1,1.7-1.3V17.9c0-3.3-2.4-2.8-3-4-1-1.6,1.8-1.6.8-5.2a2,2,0,0,0-2.1-1.3,2,2,0,0,0-2.1,1.3c-1,3.6,1.7,3.6.8,5.2-.7,1.2-3,.7-3,4V30.2c0,1.2.4,1.3,1.7,1.3.4,0,.3.2.3.9v3.3c0,.7-.1,1,.6,1H33"/><path class="cls-1" d="M22.3,10.1c-1.2,1.3-3.1,4.3-1.9,6.8,2.5,1,2.3-8,5.5-7.9,1.6,1.4-.5,5.7-1.1,7.5-.8,2.3-1.4,7.1-2.9,9.5-1.2,2-.3,8.1-.7,10.2-1,1-3.9.3-5.1.1a81.76,81.76,0,0,0-.7-8.2c0-.8-.9-12.7-.2-14.1C16.6,11,24,5.3,25,4.2S29,0,30.3,0c.9.6.4,2,.2,2.5-1.2,3-6.5,6-8.2,7.6"/><path class="cls-1" d="M40.4,10.1c1.2,1.3,3.1,4.3,1.9,6.8-2.6,1-2.4-8-5.5-7.9-1.5,1.4.5,5.7,1.1,7.5.8,2.3,1.4,7.1,2.9,9.5,1.2,2,.3,8.1.7,10.2,1.1,1,3.9.3,5.1.1a75.58,75.58,0,0,1,.7-8.2c.1-.8.9-12.7.2-14.1-1.4-3-8.8-8.7-9.8-9.8S33.6,0,32.4,0c-.9.6-.3,2-.2,2.5,1.3,3,6.5,6,8.2,7.6"/><path class="cls-1" d="M11.5,6.1c-.7.6-2.1,1.1-3.2,2.4-.4-.7.6-2.5,1.8-3.6a17.68,17.68,0,0,1,3.2-2.1c.4.9-1,2.3-1.8,3.3"/><path class="cls-1" d="M9.4,11c-1.1,1-2.3,1.5-3.9,3.4a12.22,12.22,0,0,1-.2-5.7c.6-1.8,1.5-3.8,3-4.4a15.69,15.69,0,0,0-.9,4.3c-.7,2.6-1,2.7-.9,3,.2,0,1-1.4,2.1-2.4C9.8,8,11.4,7.7,12.4,6.9a17.3,17.3,0,0,1-3,4.1"/><path class="cls-1" d="M3.9,21.8a14.31,14.31,0,0,1-1.8-7c.1-1.6.9-5.3,2-5.9-.3,3.4.6,3.7.2,6.4,0,.1-.4,3-.4,3.2.5-.1,0-2.6,5.4-6.4-1.2,2-.6,2.6-2.7,5.3a10.74,10.74,0,0,0-2.7,4.4"/><path class="cls-1" d="M6.1,24.3a6.63,6.63,0,0,0-1.2,3.2C2.2,26-.9,19.8.2,16.7c.7,2.3,1.7,2.9,2.4,4.8,1.2,2.9,1,3.8,1.8,4.5a9,9,0,0,1,.9-4.6,32.07,32.07,0,0,1,1.9-3.1c.5,2.1.1,3.8-1.1,6"/><path class="cls-1" d="M8.5,29.6a46,46,0,0,1-.6,5.2C3.3,33.3.8,28,.8,25.4c1.4,3.2,3.5,3.2,6.2,7.7.2-1.7-.5-.4-.6-3.7.1-3.4.8-3.5.9-5.6.9.9,1.2,4.7,1.2,5.8"/><path class="cls-1" d="M12.3,36.4c.4,1.8.5,3.8.8,5.3-2.7-1-3.3-.2-6.5-2.2-1.4-.8-4-4.1-3.9-6.4,3.1,3.5,3.7,3.1,5.2,4.4,1.4,1.1,2.5,2.8,3.6,2.6.1-.9-2.3-3-2-8.8a14.15,14.15,0,0,1,2.8,5.1"/><path class="cls-1" d="M17.6,40.8c.8,1.2.6,1.7,2.7,4.1A25.54,25.54,0,0,1,14.8,46c-1.3,0-6-1.2-7.6-4.4,2.7,1.4,2.7,1.1,6.3,1.9,2.6.6,4.4,1.4,5.1,1.3-.1-.3-2.1-1.2-3.5-2.9s-1-3.9-2.3-5.4c.9-.4,4.1,3.4,4.8,4.3"/><path class="cls-1" d="M51,6.1c.6.6,2.2,1.1,3.1,2.4.4-.7-.6-2.5-1.8-3.6a14.77,14.77,0,0,0-3.2-2.1c-.4.9,1,2.3,1.9,3.3"/><path class="cls-1" d="M53.1,11c1.1,1,2.3,1.5,3.9,3.4a12.22,12.22,0,0,0,.2-5.7c-.5-1.8-1.5-3.8-3-4.4a15.69,15.69,0,0,1,.9,4.3c.7,2.6,1,2.7.9,3-.2,0-1-1.4-2.1-2.4C52.7,8,51.2,7.7,50.1,6.9a17.3,17.3,0,0,0,3,4.1"/><path class="cls-1" d="M58.5,21.8a14,14,0,0,0,1.8-7c-.1-1.6-.9-5.3-2-5.9.3,3.4-.6,3.7-.2,6.4,0,.1.3,3,.3,3.2-.4-.1,0-2.6-5.3-6.4,1.1,2,.6,2.6,2.7,5.3a9.28,9.28,0,0,1,2.7,4.4"/><path class="cls-1" d="M56.4,24.3a9.58,9.58,0,0,1,1.2,3.2c2.7-1.6,5.8-7.8,4.6-10.8-.7,2.3-1.7,2.9-2.3,4.8-1.3,2.9-1.1,3.8-1.8,4.5a11.79,11.79,0,0,0-.9-4.6,19.82,19.82,0,0,0-1.9-3.1,7.51,7.51,0,0,0,1.1,6"/><path class="cls-1" d="M54,29.6a30.92,30.92,0,0,0,.7,5.2c4.6-1.5,7.1-6.8,7.1-9.4-1.5,3.2-3.5,3.2-6.2,7.7-.2-1.7.5-.4.5-3.7s-.8-3.5-.9-5.6c-.9.9-1.2,4.7-1.2,5.8"/><path class="cls-1" d="M50.1,36.4c-.4,1.8-.5,3.8-.7,5.3,2.6-1,3.3-.2,6.4-2.2,1.4-.8,4.1-4.1,3.9-6.4-3.1,3.5-3.7,3.1-5.2,4.4-1.4,1.1-2.6,2.8-3.6,2.6-.1-.9,2.3-3,2-8.8a15.84,15.84,0,0,0-2.8,5.1"/><path class="cls-1" d="M44.9,40.8a16.19,16.19,0,0,1-2.7,4.1A26.81,26.81,0,0,0,47.7,46c1.3,0,6-1.2,7.5-4.4-2.7,1.4-2.7,1.1-6.3,1.9-2.6.6-4.4,1.4-5.1,1.3.2-.3,2.2-1.2,3.5-2.9s1-3.9,2.3-5.4c-.9-.4-4,3.4-4.7,4.3"/><path class="cls-1" d="M24.6,51.8l-.8-.4a17.2,17.2,0,0,1,13.5-5.6c8.5.5,7.9,2.5,12.5,2.1.5,0-.4.6-.4.6a15.56,15.56,0,0,1-10-.2c-5.5-1.8-9.1-2.9-14.8,3.5"/><path class="cls-1" d="M38,51.8l.8-.4a17,17,0,0,0-13.4-5.6c-8.5.5-7.9,2.5-12.5,2.1-.5,0,.4.6.4.6a15.56,15.56,0,0,0,10-.2c5.5-1.8,9-2.9,14.7,3.5"/><path class="cls-1" d="M98.3,25.5c0,8.1-4.4,12.1-13.4,12.1S71.5,33.7,71.5,25.5V7.5h9v16c0,3,0,6.8,4.4,6.8s4.3-3.8,4.3-6.8V7.5h9.1v18"/><polygon class="cls-1" points="103.7 7.4 113 7.4 121.6 23.1 121.6 23.1 121.6 7.4 130.2 7.4 130.2 36.8 121.4 36.8 112.4 20.8 112.3 20.8 112.3 36.8 103.7 36.8 103.7 7.4"/><polygon class="cls-1" points="135.8 7.4 144.8 7.4 144.8 17.6 153.3 17.6 153.3 7.4 162.3 7.4 162.3 36.8 153.3 36.8 153.3 25.2 144.8 25.2 144.8 36.8 135.8 36.8 135.8 7.4"/><path class="cls-1" d="M185.5,18.4a4.69,4.69,0,0,0-4.7-4.3c-4.2,0-5.6,4.1-5.6,8.1s1.4,8.1,5.6,8.1c3,0,4.2-2.1,4.8-4.9h8.9c0,5.9-4.8,12.2-13.4,12.2-9.5,0-14.9-6.7-14.9-15.4,0-9.3,5.8-15.4,14.9-15.4,8.1,0,12.5,4.3,13.3,11.6h-8.9"/><path class="cls-1" d="M207.3,14.4h4.2c3.5,0,4.5,1.1,4.5,2.9,0,2.6-2.3,2.9-4,2.9h-4.7Zm-9.1,22.4h9.1V26.6h4.8c3.6,0,3.8,2.9,4.1,5.7a23.38,23.38,0,0,0,.7,4.5h9c-.8-1.4-.9-4.9-1-6.3-.3-3.7-1.9-6.4-4.3-7.3,3-1.1,4.4-4.3,4.4-7.4,0-5.6-4.4-8.4-9.6-8.4H198.2V36.8"/><polygon class="cls-1" points="74.5 43.6 71.5 43.6 71.5 41.7 79.6 41.7 79.6 43.6 76.6 43.6 76.6 51.6 74.5 51.6 74.5 43.6"/><path class="cls-1" d="M80.3,41.7h2v3.7h0a2.44,2.44,0,0,1,2.1-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4H85v-4c0-1.2-.3-1.8-1.3-1.8s-1.5.6-1.5,2v3.7h-2l.1-9.7"/><path class="cls-1" d="M89.7,48.5q.15,1.8,1.8,1.8a1.65,1.65,0,0,0,1.5-.9h1.7a3.22,3.22,0,0,1-3.3,2.4,3.46,3.46,0,0,1-3.6-3.3v-.4a3.59,3.59,0,0,1,3.4-3.8h.3c2.4,0,3.6,2,3.5,4.2Zm3.2-1.2c-.2-1-.6-1.5-1.6-1.5a1.55,1.55,0,0,0-1.6,1.4v.1h3.2"/><path class="cls-1" d="M107.6,47.9c0,2.7-1.6,4-4.2,4s-4.2-1.3-4.2-4V41.8h2.2v6.1c0,1.1.3,2.1,2,2.1,1.5,0,2-.7,2-2.1V41.8h2.2v6.1"/><polygon class="cls-1" points="109 41.7 111.2 41.7 115.3 48.3 115.3 48.3 115.3 41.7 117.4 41.7 117.4 51.6 115.2 51.6 111.1 45 111.1 45 111.1 51.6 109 51.6 109 41.7"/><path class="cls-1" d="M124.2,43.4h2.4c1,0,1.5.4,1.5,1.4s-.5,1.4-1.5,1.4h-2.4Zm-2.1,8.2h2.2V47.8h2.2c1.1,0,1.5.5,1.6,1.5a6.72,6.72,0,0,0,.4,2.4h2.2a5.87,5.87,0,0,1-.4-2.3c-.1-1-.4-2-1.4-2.3h0a2.5,2.5,0,0,0,1.6-2.5,2.69,2.69,0,0,0-2.7-2.7h-5.5l-.2,9.7"/><path class="cls-1" d="M133.2,47.3a1.54,1.54,0,0,1,1.5-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.49,1.49,0,0,1-1.5.9q-1.65,0-1.8-1.8l5.2.1"/><path class="cls-1" d="M139.6,45.8h-1.2V44.5h1.2V44a2.05,2.05,0,0,1,2-2.2h.4a3.08,3.08,0,0,1,1,.1v1.5h-.7c-.5,0-.7.2-.7.8v.4H143v1.3h-1.4v5.8h-2V45.8"/><path class="cls-1" d="M150.2,51.6h-1.9v-1h0a2.75,2.75,0,0,1-2.2,1.2c-2.1,0-2.6-1.2-2.6-2.9V44.5h2v4c0,1.2.4,1.8,1.3,1.8,1.1,0,1.5-.6,1.5-2V44.5h2v7.1"/><path class="cls-1" d="M154.7,49.9c-1.2,0-1.6-1.1-1.6-2.1s.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2a1.73,1.73,0,0,1-1.5,2c-.1,0-.1,0-.2-.1m3.6-5.4h-1.9v1h0a2,2,0,0,0-2.1-1.1,3.21,3.21,0,0,0-3.1,3.5c0,1.8.9,3.6,3,3.6a2.36,2.36,0,0,0,2.1-1.1h0v.9c0,1-.5,1.8-1.6,1.8a1.33,1.33,0,0,1-1.5-1h-1.9c.1,1.6,1.7,2.3,3.1,2.3,3.3,0,3.7-2,3.7-3.2l.2-6.7"/><path class="cls-1" d="M161.2,47.3a1.54,1.54,0,0,1,1.5-1.5h.1c1,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.49,1.49,0,0,1-1.5.9q-1.65,0-1.8-1.8l5.2.1"/><path class="cls-1" d="M168.9,47.3a1.5,1.5,0,0,1,1.6-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.1,1.2c.1-2.2-1-4.2-3.5-4.2a3.5,3.5,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.15,3.15,0,0,0,3.3-2.4h-1.7a1.54,1.54,0,0,1-1.6.9q-1.65,0-1.8-1.8l5.3.1"/><path class="cls-1" d="M183,44.2h0l1.2,3.6h-2.6Zm-4.8,7.4h2.2l.8-2.2h3.7l.7,2.2h2.3l-3.7-9.8H182Z"/><path class="cls-1" d="M191.5,49.9c-1.2,0-1.6-1.1-1.6-2.1s.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2a1.8,1.8,0,0,1-1.5,2c-.1,0-.2,0-.2-.1m3.5-5.4h-1.9v1h0a2.09,2.09,0,0,0-2.1-1.1,3.21,3.21,0,0,0-3.1,3.5c0,1.8.9,3.6,3.1,3.6a2.36,2.36,0,0,0,2.1-1.1h0v.9c0,1-.5,1.8-1.6,1.8a1.33,1.33,0,0,1-1.5-1h-1.9c.1,1.6,1.7,2.3,3.1,2.3,3.3,0,3.7-2,3.7-3.2l.1-6.7"/><path class="cls-1" d="M197.9,47.3a1.5,1.5,0,0,1,1.6-1.5h.1c.9,0,1.4.5,1.6,1.5Zm5.2,1.2c.1-2.2-1-4.2-3.5-4.2a3.56,3.56,0,0,0-3.6,3.5V48a3.4,3.4,0,0,0,3.2,3.7h.4a3.22,3.22,0,0,0,3.3-2.4h-1.7a1.54,1.54,0,0,1-1.6.9q-1.65,0-1.8-1.8l5.3.1"/><path class="cls-1" d="M203.9,44.5h1.9v1h.1a2.75,2.75,0,0,1,2.2-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-.3-1.8-1.3-1.8s-1.5.6-1.5,2v3.7h-2v-7"/><path class="cls-1" d="M216.6,47a1.32,1.32,0,0,0-1.4-1.2h0c-1.3,0-1.7,1.3-1.7,2.3s.4,2.2,1.6,2.2a1.62,1.62,0,0,0,1.6-1.4h1.9a3.08,3.08,0,0,1-3.4,2.9h0a3.46,3.46,0,0,1-3.6-3.3v-.3a3.43,3.43,0,0,1,3.2-3.8h.5a3.12,3.12,0,0,1,3.4,2.7l-2.1-.1"/><path class="cls-1" d="M223,52.5a2.21,2.21,0,0,1-2.5,1.6,4.87,4.87,0,0,1-1.2-.1V52.4c.4,0,.8.1,1.1.1a1.08,1.08,0,0,0,.8-1.1.37.37,0,0,0-.1-.3l-2.5-6.7h2.1l1.6,4.9h0l1.6-4.9h2L223,52.5"/></svg>"); +--unhcr-logowhite: url("data:image/svg+xml;base64,<?xml version="1.0" encoding="utf-8"?>
<!-- Generator: Adobe Illustrator 25.2.0, SVG Export Plug-In . SVG Version: 6.00 Build 0)  -->
<svg version="1.1" id="Layer_1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" x="0px" y="0px"
	 viewBox="0 0 225.9 54.3" style="enable-background:new 0 0 225.9 54.3;" xml:space="preserve">
<style type="text/css">
	.st0{fill:#FFFFFF;}
</style>
<path class="st0" d="M33,36.7c0.6,0,0.6-0.4,0.6-1v-3.3c0-0.7-0.1-0.9,0.3-0.9c1.4,0,1.7-0.1,1.7-1.3V17.9c0-3.3-2.4-2.8-3-4
	c-1-1.6,1.8-1.6,0.8-5.2c-0.3-0.9-1.2-1.4-2.1-1.3c-0.9-0.1-1.8,0.4-2.1,1.3c-1,3.6,1.7,3.6,0.8,5.2c-0.7,1.2-3,0.7-3,4v12.3
	c0,1.2,0.4,1.3,1.7,1.3c0.4,0,0.3,0.2,0.3,0.9v3.3c0,0.7-0.1,1,0.6,1H33"/>
<path class="st0" d="M22.3,10.1c-1.2,1.3-3.1,4.3-1.9,6.8c2.5,1,2.3-8,5.5-7.9c1.6,1.4-0.5,5.7-1.1,7.5c-0.8,2.3-1.4,7.1-2.9,9.5
	c-1.2,2-0.3,8.1-0.7,10.2c-1,1-3.9,0.3-5.1,0.1c-0.1-2.7-0.3-5.5-0.7-8.2c0-0.8-0.9-12.7-0.2-14.1c1.4-3,8.8-8.7,9.8-9.8
	S29,0,30.3,0c0.9,0.6,0.4,2,0.2,2.5C29.3,5.5,24,8.5,22.3,10.1"/>
<path class="st0" d="M40.4,10.1c1.2,1.3,3.1,4.3,1.9,6.8c-2.6,1-2.4-8-5.5-7.9c-1.5,1.4,0.5,5.7,1.1,7.5c0.8,2.3,1.4,7.1,2.9,9.5
	c1.2,2,0.3,8.1,0.7,10.2c1.1,1,3.9,0.3,5.1,0.1c0.1-2.7,0.3-5.5,0.7-8.2c0.1-0.8,0.9-12.7,0.2-14.1c-1.4-3-8.8-8.7-9.8-9.8
	S33.6,0,32.4,0c-0.9,0.6-0.3,2-0.2,2.5C33.5,5.5,38.7,8.5,40.4,10.1"/>
<path class="st0" d="M11.5,6.1c-0.7,0.6-2.1,1.1-3.2,2.4C7.9,7.8,8.9,6,10.1,4.9c1-0.8,2.1-1.5,3.2-2.1C13.7,3.7,12.3,5.1,11.5,6.1"
	/>
<path class="st0" d="M9.4,11c-1.1,1-2.3,1.5-3.9,3.4C5,12.5,4.9,10.6,5.3,8.7c0.6-1.8,1.5-3.8,3-4.4C7.8,5.7,7.5,7.1,7.4,8.6
	c-0.7,2.6-1,2.7-0.9,3c0.2,0,1-1.4,2.1-2.4c1.2-1.2,2.8-1.5,3.8-2.3C11.6,8.4,10.6,9.8,9.4,11"/>
<path class="st0" d="M3.9,21.8c-1.2-2.1-1.8-4.6-1.8-7c0.1-1.6,0.9-5.3,2-5.9c-0.3,3.4,0.6,3.7,0.2,6.4c0,0.1-0.4,3-0.4,3.2
	c0.5-0.1,0-2.6,5.4-6.4c-1.2,2-0.6,2.6-2.7,5.3C5.4,18.6,4.4,20.1,3.9,21.8"/>
<path class="st0" d="M6.1,24.3c-0.7,0.9-1.1,2-1.2,3.2C2.2,26-0.9,19.8,0.2,16.7c0.7,2.3,1.7,2.9,2.4,4.8c1.2,2.9,1,3.8,1.8,4.5
	c-0.1-1.6,0.2-3.2,0.9-4.6c0.6-1.1,1.2-2.1,1.9-3.1C7.7,20.4,7.3,22.1,6.1,24.3"/>
<path class="st0" d="M8.5,29.6c-0.1,1.7-0.3,3.5-0.6,5.2c-4.6-1.5-7.1-6.8-7.1-9.4c1.4,3.2,3.5,3.2,6.2,7.7c0.2-1.7-0.5-0.4-0.6-3.7
	c0.1-3.4,0.8-3.5,0.9-5.6C8.2,24.7,8.5,28.5,8.5,29.6"/>
<path class="st0" d="M12.3,36.4c0.4,1.8,0.5,3.8,0.8,5.3c-2.7-1-3.3-0.2-6.5-2.2c-1.4-0.8-4-4.1-3.9-6.4c3.1,3.5,3.7,3.1,5.2,4.4
	c1.4,1.1,2.5,2.8,3.6,2.6c0.1-0.9-2.3-3-2-8.8C10.8,32.8,11.7,34.5,12.3,36.4"/>
<path class="st0" d="M17.6,40.8c0.8,1.2,0.6,1.7,2.7,4.1c-1.8,0.6-3.6,0.9-5.5,1.1c-1.3,0-6-1.2-7.6-4.4c2.7,1.4,2.7,1.1,6.3,1.9
	c2.6,0.6,4.4,1.4,5.1,1.3c-0.1-0.3-2.1-1.2-3.5-2.9s-1-3.9-2.3-5.4C13.7,36.1,16.9,39.9,17.6,40.8"/>
<path class="st0" d="M51,6.1c0.6,0.6,2.2,1.1,3.1,2.4c0.4-0.7-0.6-2.5-1.8-3.6c-1-0.8-2-1.5-3.2-2.1C48.7,3.7,50.1,5.1,51,6.1"/>
<path class="st0" d="M53.1,11c1.1,1,2.3,1.5,3.9,3.4c0.5-1.9,0.6-3.8,0.2-5.7c-0.5-1.8-1.5-3.8-3-4.4c0.5,1.4,0.8,2.8,0.9,4.3
	c0.7,2.6,1,2.7,0.9,3c-0.2,0-1-1.4-2.1-2.4c-1.2-1.2-2.7-1.5-3.8-2.3C50.9,8.4,51.9,9.8,53.1,11"/>
<path class="st0" d="M58.5,21.8c1.2-2.1,1.8-4.5,1.8-7c-0.1-1.6-0.9-5.3-2-5.9c0.3,3.4-0.6,3.7-0.2,6.4c0,0.1,0.3,3,0.3,3.2
	c-0.4-0.1,0-2.6-5.3-6.4c1.1,2,0.6,2.6,2.7,5.3C57.1,18.6,58,20.1,58.5,21.8"/>
<path class="st0" d="M56.4,24.3c0.6,1,1,2.1,1.2,3.2c2.7-1.6,5.8-7.8,4.6-10.8c-0.7,2.3-1.7,2.9-2.3,4.8c-1.3,2.9-1.1,3.8-1.8,4.5
	c0-1.6-0.3-3.1-0.9-4.6c-0.5-1.1-1.2-2.1-1.9-3.1C54.8,20.4,55.2,22.5,56.4,24.3"/>
<path class="st0" d="M54,29.6c0.1,1.8,0.3,3.5,0.7,5.2c4.6-1.5,7.1-6.8,7.1-9.4c-1.5,3.2-3.5,3.2-6.2,7.7c-0.2-1.7,0.5-0.4,0.5-3.7
	s-0.8-3.5-0.9-5.6C54.3,24.7,54,28.5,54,29.6"/>
<path class="st0" d="M50.1,36.4c-0.4,1.8-0.5,3.8-0.7,5.3c2.6-1,3.3-0.2,6.4-2.2c1.4-0.8,4.1-4.1,3.9-6.4c-3.1,3.5-3.7,3.1-5.2,4.4
	c-1.4,1.1-2.6,2.8-3.6,2.6c-0.1-0.9,2.3-3,2-8.8C51.7,32.8,50.7,34.5,50.1,36.4"/>
<path class="st0" d="M44.9,40.8c-0.7,1.5-1.6,2.9-2.7,4.1c1.8,0.6,3.6,0.9,5.5,1.1c1.3,0,6-1.2,7.5-4.4c-2.7,1.4-2.7,1.1-6.3,1.9
	c-2.6,0.6-4.4,1.4-5.1,1.3c0.2-0.3,2.2-1.2,3.5-2.9s1-3.9,2.3-5.4C48.7,36.1,45.6,39.9,44.9,40.8"/>
<path class="st0" d="M24.6,51.8l-0.8-0.4c3.4-3.8,8.4-5.8,13.5-5.6c8.5,0.5,7.9,2.5,12.5,2.1c0.5,0-0.4,0.6-0.4,0.6
	c-3.3,1-6.8,1-10-0.2C33.9,46.5,30.3,45.4,24.6,51.8"/>
<path class="st0" d="M38,51.8l0.8-0.4c-3.4-3.8-8.3-5.8-13.4-5.6c-8.5,0.5-7.9,2.5-12.5,2.1c-0.5,0,0.4,0.6,0.4,0.6
	c3.3,1,6.8,1,10-0.2C28.8,46.5,32.3,45.4,38,51.8"/>
<path class="st0" d="M98.3,25.5c0,8.1-4.4,12.1-13.4,12.1s-13.4-3.9-13.4-12.1v-18h9v16c0,3,0,6.8,4.4,6.8s4.3-3.8,4.3-6.8v-16h9.1
	V25.5"/>
<polygon class="st0" points="103.7,7.4 113,7.4 121.6,23.1 121.6,23.1 121.6,7.4 130.2,7.4 130.2,36.8 121.4,36.8 112.4,20.8 
	112.3,20.8 112.3,36.8 103.7,36.8 "/>
<polygon class="st0" points="135.8,7.4 144.8,7.4 144.8,17.6 153.3,17.6 153.3,7.4 162.3,7.4 162.3,36.8 153.3,36.8 153.3,25.2 
	144.8,25.2 144.8,36.8 135.8,36.8 "/>
<path class="st0" d="M185.5,18.4c-0.2-2.4-2.3-4.3-4.7-4.3c-4.2,0-5.6,4.1-5.6,8.1s1.4,8.1,5.6,8.1c3,0,4.2-2.1,4.8-4.9h8.9
	c0,5.9-4.8,12.2-13.4,12.2c-9.5,0-14.9-6.7-14.9-15.4c0-9.3,5.8-15.4,14.9-15.4c8.1,0,12.5,4.3,13.3,11.6L185.5,18.4"/>
<path class="st0" d="M207.3,14.4h4.2c3.5,0,4.5,1.1,4.5,2.9c0,2.6-2.3,2.9-4,2.9h-4.7V14.4z M198.2,36.8h9.1V26.6h4.8
	c3.6,0,3.8,2.9,4.1,5.7c0.1,1.5,0.3,3,0.7,4.5h9c-0.8-1.4-0.9-4.9-1-6.3c-0.3-3.7-1.9-6.4-4.3-7.3c3-1.1,4.4-4.3,4.4-7.4
	c0-5.6-4.4-8.4-9.6-8.4h-17.2V36.8"/>
<polygon class="st0" points="74.5,43.6 71.5,43.6 71.5,41.7 79.6,41.7 79.6,43.6 76.6,43.6 76.6,51.6 74.5,51.6 "/>
<path class="st0" d="M80.3,41.7h2v3.7l0,0c0.4-0.7,1.2-1.2,2.1-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-0.3-1.8-1.3-1.8
	s-1.5,0.6-1.5,2v3.7h-2L80.3,41.7"/>
<path class="st0" d="M89.7,48.5c0.1,1.2,0.7,1.8,1.8,1.8c0.6,0,1.2-0.3,1.5-0.9h1.7c-0.4,1.5-1.8,2.5-3.3,2.4
	c-1.9,0.1-3.5-1.4-3.6-3.3v-0.4c-0.1-2,1.4-3.7,3.4-3.8c0,0,0,0,0,0h0.3c2.4,0,3.6,2,3.5,4.2H89.7z M92.9,47.3
	c-0.2-1-0.6-1.5-1.6-1.5c-0.8,0-1.5,0.6-1.6,1.4v0.1H92.9"/>
<path class="st0" d="M107.6,47.9c0,2.7-1.6,4-4.2,4s-4.2-1.3-4.2-4v-6.1h2.2v6.1c0,1.1,0.3,2.1,2,2.1c1.5,0,2-0.7,2-2.1v-6.1h2.2
	V47.9"/>
<polygon class="st0" points="109,41.7 111.2,41.7 115.3,48.3 115.3,48.3 115.3,41.7 117.4,41.7 117.4,51.6 115.2,51.6 111.1,45 
	111.1,45 111.1,51.6 109,51.6 "/>
<path class="st0" d="M124.2,43.4h2.4c1,0,1.5,0.4,1.5,1.4s-0.5,1.4-1.5,1.4h-2.4L124.2,43.4z M122.1,51.6h2.2v-3.8h2.2
	c1.1,0,1.5,0.5,1.6,1.5c0,0.8,0.1,1.6,0.4,2.4h2.2c-0.3-0.7-0.4-1.5-0.4-2.3c-0.1-1-0.4-2-1.4-2.3l0,0c1-0.4,1.7-1.4,1.6-2.5
	c0-1.5-1.2-2.7-2.7-2.7c0,0,0,0,0,0h-5.5L122.1,51.6"/>
<path class="st0" d="M133.2,47.3c0-0.8,0.7-1.5,1.5-1.5h0.1c0.9,0,1.4,0.5,1.6,1.5H133.2z M138.3,48.5c0.1-2.2-1-4.2-3.5-4.2
	c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.5,0.9c-1.1,0-1.7-0.6-1.8-1.8L138.3,48.5"/>
<path class="st0" d="M139.6,45.8h-1.2v-1.3h1.2V44c-0.1-1.1,0.8-2.1,1.9-2.2c0,0,0.1,0,0.1,0h0.4c0.3,0,0.7,0,1,0.1v1.5h-0.7
	c-0.5,0-0.7,0.2-0.7,0.8v0.4h1.4v1.3h-1.4v5.8h-2V45.8"/>
<path class="st0" d="M150.2,51.6h-1.9v-1l0,0c-0.5,0.7-1.3,1.2-2.2,1.2c-2.1,0-2.6-1.2-2.6-2.9v-4.4h2v4c0,1.2,0.4,1.8,1.3,1.8
	c1.1,0,1.5-0.6,1.5-2v-3.8h2v7.1"/>
<path class="st0" d="M154.7,49.9c-1.2,0-1.6-1.1-1.6-2.1s0.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2c0.2,0.9-0.5,1.8-1.4,2c0,0-0.1,0-0.1,0
	C154.8,50,154.8,50,154.7,49.9 M158.3,44.5h-1.9v1l0,0c-0.4-0.8-1.2-1.2-2.1-1.1c-1.8,0.1-3.2,1.5-3.1,3.3c0,0.1,0,0.1,0,0.2
	c0,1.8,0.9,3.6,3,3.6c0.8,0,1.6-0.4,2.1-1.1l0,0v0.9c0,1-0.5,1.8-1.6,1.8c-0.7,0.1-1.3-0.3-1.5-1h-1.9c0.1,1.6,1.7,2.3,3.1,2.3
	c3.3,0,3.7-2,3.7-3.2L158.3,44.5"/>
<path class="st0" d="M161.2,47.3c0-0.8,0.7-1.5,1.5-1.5h0.1c1,0,1.4,0.5,1.6,1.5H161.2z M166.3,48.5c0.1-2.2-1-4.2-3.5-4.2
	c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.5,0.9c-1.1,0-1.7-0.6-1.8-1.8L166.3,48.5"/>
<path class="st0" d="M168.9,47.3c0-0.8,0.7-1.5,1.5-1.5c0,0,0.1,0,0.1,0h0.1c0.9,0,1.4,0.5,1.6,1.5H168.9z M174,48.5
	c0.1-2.2-1-4.2-3.5-4.2c-1.9-0.1-3.5,1.5-3.6,3.4c0,0,0,0.1,0,0.1V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4
	c1.5,0.1,2.9-0.9,3.3-2.4h-1.7c-0.3,0.6-0.9,1-1.6,0.9c-1.1,0-1.7-0.6-1.8-1.8L174,48.5"/>
<path class="st0" d="M183,44.2L183,44.2l1.2,3.6h-2.6L183,44.2z M178.2,51.6h2.2l0.8-2.2h3.7l0.7,2.2h2.3l-3.7-9.8H182L178.2,51.6z"
	/>
<path class="st0" d="M191.5,49.9c-1.2,0-1.6-1.1-1.6-2.1s0.5-2,1.6-2c1.3,0,1.7,1,1.7,2.2c0.1,1-0.5,1.8-1.5,2
	C191.6,50,191.5,50,191.5,49.9 M195,44.5h-1.9v1l0,0c-0.4-0.8-1.2-1.2-2.1-1.1c-1.8,0.1-3.2,1.5-3.1,3.3c0,0.1,0,0.1,0,0.2
	c0,1.8,0.9,3.6,3.1,3.6c0.8,0,1.6-0.4,2.1-1.1l0,0v0.9c0,1-0.5,1.8-1.6,1.8c-0.7,0.1-1.3-0.3-1.5-1h-1.9c0.1,1.6,1.7,2.3,3.1,2.3
	c3.3,0,3.7-2,3.7-3.2L195,44.5"/>
<path class="st0" d="M197.9,47.3c0-0.8,0.7-1.5,1.5-1.5c0,0,0.1,0,0.1,0h0.1c0.9,0,1.4,0.5,1.6,1.5H197.9z M203.1,48.5
	c0.1-2.2-1-4.2-3.5-4.2c-2,0-3.6,1.5-3.6,3.5V48c-0.2,1.9,1.2,3.5,3.1,3.7c0,0,0.1,0,0.1,0h0.4c1.5,0.1,2.9-0.9,3.3-2.4h-1.7
	c-0.3,0.6-0.9,1-1.6,0.9c-1.1,0-1.7-0.6-1.8-1.8L203.1,48.5"/>
<path class="st0" d="M203.9,44.5h1.9v1h0.1c0.5-0.7,1.3-1.2,2.2-1.2c2.1,0,2.6,1.2,2.6,2.9v4.4h-2v-4c0-1.2-0.3-1.8-1.3-1.8
	s-1.5,0.6-1.5,2v3.7h-2V44.5"/>
<path class="st0" d="M216.6,47c-0.1-0.7-0.7-1.2-1.4-1.2l0,0c-1.3,0-1.7,1.3-1.7,2.3s0.4,2.2,1.6,2.2c0.8,0,1.5-0.6,1.6-1.4h1.9
	c-0.1,1.7-1.5,3-3.2,2.9c-0.1,0-0.1,0-0.2,0l0,0c-1.9,0.1-3.5-1.4-3.6-3.3v-0.3c-0.2-1.9,1.1-3.6,3-3.8c0.1,0,0.1,0,0.2,0h0.5
	c1.7-0.2,3.2,1,3.4,2.7L216.6,47"/>
<path class="st0" d="M223,52.5c-0.3,1.1-1.4,1.8-2.5,1.6c-0.4,0-0.8,0-1.2-0.1v-1.6c0.4,0,0.8,0.1,1.1,0.1c0.5-0.1,0.8-0.6,0.8-1.1
	c0-0.1,0-0.2-0.1-0.3l-2.5-6.7h2.1l1.6,4.9l0,0l1.6-4.9h2L223,52.5"/>
</svg>
"); +}