[R] Using associate words to search text .

θ ” yarmi1224 at hotmail.com
Sat May 27 04:18:38 CEST 2017

I am using the package "tm" for text-mining of CMP patents.
I use findAssocs() function to find the words which associated my dictionary (technical names).
Now, I want to use these associate words to search for which documents are contain.

For example:
From the result of findAssocs(gram_dtm, dictionary_word, 0.5).

 apparatus polishing             glycerin          method high apparatus comprising
                0.66                 0.54                 0.54                 0.53

It shows "apparatus" is associate with "apparatus polishing", "glycerin", " method high" and "apparatus comprising".
How do I use the set of words to do the following works?
1. Search which documents have appear?
2. The words frequency in each documents ?
3. How to save these documents?

My code:¡@
#Load the text mining package(s)

#Build Corpus
cluster1_df<- read.csv("cluster_1.csv",stringsAsFactors = F)
cluster1_combined <- cluster1_df[,c(3,4,5)]
corpus <- Corpus(DataframeSource(cluster1_combined))

#Pre-processing and tranforming the Corpus
myStopwords <- c(stopwords("english"), stopwords("SMART"),"claim")
corpus_tm <- tm_map(corpus, content_transformer(tolower))
corpus_tm <- tm_map(corpus_tm, removeWords, myStopwords)
corpus_tm <- tm_map(corpus_tm, removeNumbers)
corpus_tm <- tm_map(corpus_tm, removePunctuation)
corpus_tm <- tm_map(corpus_tm, stripWhitespace)
corpus_tm <- tm_map(corpus_tm, stemDocument)

BigramTokenizer <- function(x) NGramTokenizer(x, Weka_control(min = 1, max = 3))
gram_dtm <- DocumentTermMatrix(corpus_tm, control = list(tokenize = BigramTokenizer,
                                                         weighting = function(x) weightTfIdf(x, normalize = TRUE)))
gram_dtm <- removeSparseTerms(gram_dtm, 1-(5/length(corpus_tm)))

dictionary_word <- c("neutral", "abras", "particl", "acid", "apparatus", "back film", "basic", "carrier",
                     "chemic", "chromat","confoc", "clean", "cmp", "compens type", "compress",
                     "comsum", "control", "pressur", "dresser", "condition", "detect",
                     "flow","rate", "fractal", "groov", "hard", "improv type", "infrar",
                     "laser", "layer", "measur", "micro stuctur", "monitor",
                     "multi layer", "none-por", "nonwoven", "pad", "pad applic", "pad condit",
                     "pad materi", "pad properti", "pad structur", "ph","planet", "plate",
                     "plat", "ratio", "polish head", "polish system", "polym", "polyurethan",
                     "porous", "process"," paramet", "path", "time", "recoveri", "speed",
                     "rough", "scatter", "semiconductor", "sensor", "signal", "singl layer",
                     "slurri", "flow rate",  "stirrer", "slurri suppli",
                     "temperatur", "weight percentag","wt", "storag cmp", "stylus profil", "substrat cmp",
                     "thick", "transfer robot", "ultrason", "urethan", "wafer cassett", "wafer transfer",
                     "white light interferomet", "youngs modulus")

onto_assocs<- findAssocs(gram_dtm, dictionary_word, 0.5)

	[[alternative HTML version deleted]]

More information about the R-help mailing list