Trying tokenizing for identifying similarities in criteria

misc/formative-elements/fill-missing-taxa.R

@@ -0,0 +1,66 @@
+library(dplyr, warn=FALSE)
+library(tokenizers)
+
+scu <- read.csv("~/workspace/SoilTaxonomy/misc/formative-elements/subgroup_criteria_unexplained.csv", stringsAsFactors = FALSE)
+
+scut <- scu %>% 
+  group_by(element, taxon) %>% 
+  summarize(element = unique(element), 
+            content = paste0(content, collapse = "\n"))
+
+scut$content_tokens <- tokenizers::tokenize_words(scut$content)
+
+fmt_colormatch <- lapply(split(scut, f = scut$element), function(scusub) {
+
+  nidx <- 1:nrow(scusub)
+
+  # create token comparison sets; n taxa in a element group get compared to n - 1 non-self criteria
+  res1 <- lapply(nidx, function(i)
+    lapply(nidx[nidx != i], function(j) {
+      tokens <- scusub[i, 'content_tokens'][[1]][[1]]
+      nonselftokens <- do.call('c', scusub[j, 'content_tokens'][[1]])
+      tokens[tokens %in% nonselftokens]
+    }))
+
+  # take the shortest (most restrictive) set of matches
+  res2 <- lapply(res1, function(x) x[[which.min(sapply(x, length))[1]]])
+
+  res3 <- lapply(nidx, function(i) {
+    foo <- gsub("\xbd|1 /2","1/2", gsub("&nbsp;"," ",scusub$content[i]))
+    foo2 <- NULL
+    try(foo2 <- tolower(foo[1]), silent = TRUE)
+    if (!is.null(foo2)) {
+      foo2 <- foo
+    } else {
+      cat("### ",unique(scusub$element),"\n\n")
+    }
+    if(is.null(res2[[i]])) return(NULL)
+    res <- as.list(apply(stringr::str_locate(foo2[1], res2[[i]]), 1,
+                  function(x)
+                   if (!is.na(x[1]))# && x[2] - x[1] > 0)
+                    list(
+                      old = substr(foo, x[1], x[2]),
+                      new = crayon::red(substr(foo, x[1], x[2])))
+                    ))
+    if(length(res))
+      for(j in 1:length(res)) {
+        re <- res[[j]]
+        if(!is.null(re))
+          foo <- gsub(paste0("\\b",re$old,"\\b"), re$new, foo)
+      }
+    cat(foo)
+    cat("\n\n")
+    foo
+  })
+})
+
+foo <- lapply(seq_along(fmt_colormatch), function(i) {cat("## ", names(fmt_colormatch)[i], "\n\n")
+                                                         sapply(fmt_colormatch[[i]], cat, "\n\n")})
+
+
+a_list <- as.list(rep("",93))
+names(a_list) <- names(fmt_colormatch)
+a_list[["vitrixerandic"]] <- "xeric SMR with significant amount of volcanic glass particles"
+a_list[["xereptic"]] <- "duripan strongly or less cemented, SMR borders on xeric"
+a_list[["xerertic"]] <- "xeric SMR with cracks 5mm or more wide and slickensides or wedge-shaped peds"
+a_list[["xerollic"]] <- "xeric SMR with color and base saturation of mollic epipedon"