So Many Words.
Let's do some more with words. R allows us to analyze the intent of the writer. Raspberry Pi allows us to simulate the reader.
R Programming Language
Natural Language Processing refers to the words in a document as "terms." Nothing mysterious about that. When you analyze a document, you want to count the terms. R allows you to create a matrix with terms on one axis and documents on the other. For example...
# Name of the columns are terms head(RT_DTmatrix$dimnames$Terms) [1] "aback" "abandon" "abanindra" "abanindranath" [5] "abash" "abat" # the rows are documents > head(RT_DTmatrix$dimnames$Docs) [1] "Chitra..a.Play.in.One.Act" [2] "Creative.Unity" [3] "Fruit.Gathering" [4] "Gitanjali" [5] "Glimpses.of.Bengal.Selected.from.the.Letters.of.Sir.Rabindranath.Tagore" [6] "Mashi..and.Other.Stories"
In the full document, there are 14,026 terms and 19 documents. Here's how to build a document/term matrix from the works of Rabindranath Tagore.
# NLP: create document vs terms.
# Load the tm package ---------------------------------
# install.packages("tm")
library(tm)
# start with the corpus developed in previous session
# RT_corpus contains the work of Rabindranath Tagore as found at project gutenberg
load(file = "Rabindranath_corpus.rdata")
# Term Document Matrix ----------------------------------------------------
# Term Document is terms on rows.
# Document Term is documents on rows
RT_TDmatrix <- TermDocumentMatrix(RT_corpus, control = list(
stopwords = TRUE,
removePunctuation = TRUE,
removeNumbers = TRUE,
stemming = TRUE))
# this tokenizes and places terms across top. Inspect dimnames of RT_TDmatrix
RT_DTmatrix <- DocumentTermMatrix(RT_corpus, control = list(
stopwords = TRUE,
removePunctuation = TRUE,
removeNumbers = TRUE,
stemming = TRUE))
inspect(RT_DTmatrix)
# this is a matrix - albeit object type = DocumentTermMatrx.
# with "inspect()" You can subset just like any other matrix
inspect(RT_DTmatrix[ 3 , "heart"]) # row 3, column "heart"
inspect(RT_DTmatrix[ , "heart"]) # all rows, column "heart"
inspect(RT_DTmatrix[ , c("heart", "eye", "sad")]) # all rows, multiple columns
Confused? - here's a video to explain document/term or term/document matrices...
With this matrix of words by documents, you can do some really interesting stuff. More to come next week.
Raspberry Pi
Speech is words come to life. Or in the case of a computer, the simulation of words come to life. Here's a segment from Raspberry Pi weekly where we hook up a speech synthesizer.
It's obviously a machine speaking - but if it reads poetry, is it inhuman? Before you debate the theory, here's some code to implement an example...
# feed a line of text to Emic
import serial
import random
import time
ser = serial.Serial(
port='/dev/ttyS0',
baudrate = 9600,
parity=serial.PARITY_NONE,
stopbits=serial.STOPBITS_ONE,
bytesize=serial.EIGHTBITS
)
def speakAString( theCommand, sayThis):
time.sleep(.25) # pause to give Emics a secod to reset
sayThisFormatted = theCommand + str(sayThis) + "\r" # Emics requires lines to end in return
# print("sending",sayThisFormatted)
ser.write(sayThisFormatted.encode())
sayThis = "nothing yet"
while (sayThis != "stop"):
# choose a random voice
theVoice = random.randint(0,8)
speakAString("N",theVoice)
print("Speaking with voice #", theVoice)
# choose a speed
theSpeed = random.randint(75,250) #set speaking rate. max = 600
speakAString("W",theSpeed)
print("Speaking Rate: ",theSpeed)
# choose a volume
theVolume = random.randint(-20,18) # Minimum is -48, but you can't hear it
speakAString("V", theVolume)
print("Volume: ", theVolume)
sayThis = input("Listening to you: ")
print("Speaking: ", sayThis)
speakAString("S", sayThis)
print("\n\n")
ser.close()
I'm on a mission to explore alternatives to user interface. Are monitors / keyboards our only option for communicating with a computer? I hope not - and here's an alternative.
What is this newsletter?
Every week I discuss the R language, Raspberry Pi or other R topics. Please share and subscribe!
#rstats #raspberrypi #r