-
Notifications
You must be signed in to change notification settings - Fork 0
/
testpred 7.44.50 PM.R
243 lines (225 loc) · 9.15 KB
/
testpred 7.44.50 PM.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
# library(caret)
# library(data.table)
# library(quanteda)
# library(stringr)
# library(qdap)
# #############################################
# # #Splitting the corpus into training and validation sets
# set.seed(11152020)
# inTest <- createDataPartition(seq_len(NROW(valid.data)), p=.10, list=FALSE)
# test.data.10 <- valid.data[inTest]
# unused.test.data <- valid.data[-inTest]
#
# save(test.data.10, file = "test.data.10.Rda")
# remove(inTest)
# remove(unused.test.data)
# ###########################################
# #Text Processing function
#
# text.proc.fun <- function(text.object, stop.bad=0){
#
# text.object <- str_replace_all(text.object, "-", " ")
#
# text.object <- replace_contraction(text.object, contraction = qdapDictionaries::contractions)
#
# text.object <- tolower(text.object)
# text.object <- str_replace_all(text.object, "\\S*@+\\S*", " ")
# text.object <- str_replace_all(text.object, "\\S*#+\\S*", " ")
# text.object <- str_replace_all(text.object, "[^[a-zA-Z][:space:]]", "")
#
# text.object <- tokens(text.object)
# digit.left <- c("pm", "nd", "rd", "th")
# text.object <- tokens_select(text.object, digit.left, selection = "remove")
# abbrev <- c("aka", "appt", "approx", "apt", "asap", "ave", "afk", "bnb", "bff", "byob", "btw", "cm", "diy", "dnd", "dr", "eta", "eg", "faq", "idk", "kg", "misc", "ms", "mrs", "mr", "na", "lb", "sr", "ttyl", "tgif", "vs", "vip")
# fullword <- c("also known as","apartment","appointment","approximately","as soon as possible","avenue","away from keyboard","bed and breakfast","best friend forever","bring your own bottle","by the way","centimeter","do it yourself","do not disturb","doctor","estimated time of arrival","for example","frequently asked questions","i do not know", "kilogram","miscellaneous","miss", "misses","mister", "not applicable", "pound","senior","talk to you later","thank god its friday","versus","very important person")
# text.object <- tokens_replace(text.object, abbrev, fullword)
#
# remove.word <- read.table("badwords.txt", nrows=-1, sep="\n")
# remove.word <- as.vector(remove.word[1:447,])
# remove.word <- tolower(remove.word)
#
# if(stop.bad == 1){
# text.object <- tokens_select(text.object, pattern = stopwords("en"), selection = "remove")
# }
# if(stop.bad == 2){
# text.object <- tokens_select(text.object, remove.word, selection = "remove")
# }
# if(stop.bad == 3){
# text.object <- tokens_select(text.object, pattern = stopwords("en"), selection = "remove")
# text.object <- tokens_select(text.object, remove.word, selection = "remove")
# }
# text.object <- tokens_split(text.object)
# return(text.object)
#
# }
#
# valid.Toke <- text.proc.fun(test.data.10)
# #Gets the last 4 words of each text
#
# #Creating an empty list for the loop to gather the last four words of each text
# valid.last.4 <- vector(mode = "list")
# #
# # #gathering the last 4 words of each text. We will be predicting the last word of each list element(text).
# # #need to chaned i below to the number of rows after completion
# for (i in 1:length(valid.Toke)){
# # Only takes in strings that are greater than 3 words long & printing out the last 4 words. This will be wha t is
# #predicted on.
# if (length(valid.Toke[[i]])>3){
#
# valid.last.4[i][1] <- rev(valid.Toke[[i]])[4]
# valid.last.4[[i]][2] <- rev(valid.Toke[[i]])[3]
# valid.last.4[[i]][3] <- rev(valid.Toke[[i]])[2]
# valid.last.4[[i]][4] <- rev(valid.Toke[[i]])[1]
#
# }
# }
#
# #Making a data.table of the last four words of each text
# valid.DT <- data.table(lapply(valid.last.4, unlist))
# valid.DT <- valid.DT[, c("first", "second", "third", "temp") := tstrsplit(as.character(V1), ",", fixed=TRUE)]
# valid.DT[, c("fourth") := tstrsplit(temp, ")", fixed=TRUE)]
# valid.DT[, first := gsub("c\\(", "", first)]
# valid.DT[,c("V1","temp"):=NULL]
# valid.DT[, first := gsub("\\\"", "", first)]
# valid.DT[, second := gsub(" \\\"", "", second)]
# valid.DT[, second := gsub("\\\"", "", second)]
# valid.DT[, third := gsub(" \\\"", "", third)]
# valid.DT[, third := gsub("\\\"", "", third)]
# valid.DT[, fourth := gsub(" \\\"", "", fourth)]
# valid.DT[, fourth := gsub("\\\"", "", fourth)]
# #Back off functions
# uni.back.fun <- function(prev.words, k){
#
# uni.words <- fread("uni.words.60.pkn.csv")
#
# uni <- uni.words[!(last %in% prev.words)]
# new.k <- min(k, nrow(uni))
# pred.words <- uni$last[1:new.k]
# pred.probs <- uni$p.kn[1:new.k]
#
# preds <- data.table(return.words = pred.words, return.probs = pred.probs )
#
# remove(uni.words)
#
# return(preds)
# }
#
# bi.back.fun <- function(word3, prev.words, k){
#
# bi.words <- fread("bi.words.60.pkn.csv")
#
# bi.exist <- any(!(is.na( bi.words[first==word3][!(last %in% prev.words)])))
#
# if (bi.exist){
#
# bi <- bi.words[first==word3][!(last %in% prev.words)]
# new.k <- min(k, nrow(bi))
# pred.words <- bi$last[1:new.k]
# pred.probs <- bi$p.kn[1:new.k]
#
# # If backing off
# uni.back <- uni.back.fun(c(prev.words, pred.words), k)
# pred.words <- c(pred.words, uni.back$return.words)
# pred.probs <- c(pred.probs, uni.back$return.probs)
#
# } else {
#
# uni.back <- uni.back.fun(prev.words, k)
# pred.words <- uni.back$return.words
# pred.probs <- uni.back$return.probs
#
# }
#
# preds = data.table(return.words = pred.words, return.probs = pred.probs)
#
# remove(bi.words)
#
# return(preds)
# }
#
# tri.back.fun <- function(word2, word3, prev.words, k) {
#
# tri.words <- fread("tri.words.60.pkn.csv")
#
# tri.exist <- any(!(is.na( tri.words[first==word2 & second==word3][!(last %in% prev.words)])))
#
# if (tri.exist){
#
# tri <- tri.words[first==word2 & second==word3][!(last %in% prev.words)]
# new.k <- min(k, nrow(tri))
# pred.words <- tri$last[1:new.k]
# pred.probs <- tri$p.kn[1:new.k]
#
#
# # If backing off
# bi.back <- bi.back.fun(word3, c(prev.words, pred.words), k)
# pred.words <- c(pred.words, bi.back$return.words)
# pred.probs <- c(pred.probs, bi.back$return.probs)
#
# } else {
#
# bi.back <- bi.back.fun(word3, prev.words, k)
# pred.words <- bi.back$return.words
# pred.probs <- bi.back$return.probs
#
# }
#
# preds = data.table(return.words = pred.words, return.probs = pred.probs)
#
# remove(tri.words)
#
# return(preds)
# }
#
# quadr.back.fun <- function(word1, word2, word3, prev.words, k) {
#
# quadr.words <- fread("quadri.words.60.pkn.csv")
#
# quadr.exist <- any(!(is.na( quadr.words[first==word1 & second==word2 & third==word3])))
#
# if (quadr.exist){
#
# quadr <- quadr.words[first==word1 & second==word2 & third==word3]
# new.k <- min(k, nrow(quadr))
# pred.words <- quadr$last[1:new.k]
# pred.probs <- quadr$p.kn[1:new.k]
#
# # If backing off
# tri.back <- tri.back.fun(word2,word3, c(prev.words, pred.words), k)
# pred.words <- c(pred.words, tri.back$return.words)
# pred.probs <- c(pred.probs, tri.back$return.probs)
#
# } else {
#
# tri.back <- tri.back.fun(word2,word3, prev.words=0, k)
# pred.words <- tri.back$return.words
# pred.probs <- tri.back$return.probs
#
# }
#
# preds <- data.table(return.words = pred.words, return.probs = pred.probs)
#
# remove(quadr.words)
#
# return(preds)
#
# }
#Runs words 1, 2, and 3 of the last four words of each text through the KN back off probability functions to build
#a table of next wrd predictions
#Searches each list for word 4 from the testing data and if found prints TRUE, if not found prints FALSE.
# preds <- vector(mode="list")
# time.60.3 <- system.time({
# for (i in 1:length(valid.DT[[1]])){
# #for (i in 1:20){
#
# preds[i] <- quadr.back.fun(valid.DT[[1]][i], valid.DT[[2]][i], valid.DT[[3]][i], prev.words=0, k=25)
# preds[i] <- valid.DT[[4]][i] %in% unlist(preds[i])
# }
# })
# for (i in 1:length(preds)){
# true.pred <- length(which(preds =="TRUE"))
# }
#How many of the textx received an accurate prediction?
#valid.length <- nrow(valid.DT)
summary.preds <- data.frame(Model=c("Unfiltered", "Stop.Words.Filtered", "Profanity.Filtered", "Both.Filtered"), Sys.Time=c(time.60.3[2],time.60.3.1[2],time.60.3.2[2],time.60.3.3[2]), Percent.Correct=c((true.pred/valid.length)*100,(true.pred.1/valid.length)*100,(true.pred.2/valid.length)*100,(true.pred.3/valid.length)*100))
save(summary.preds, file="summary.preds")