10.6 LDA output
Now we are ready see the words associated with each of the 20 topics. Recall that we used stemming, which means some of the words will be difficult to read. We print first 10 words for each topic using terms() function.
terms(lda_model, 10)## Topic 1 Topic 2 Topic 3 Topic 4 Topic 5 Topic 6
## [1,] "christma" "googl" "speaker" "internet" "memori" "stick"
## [2,] "wife" "store" "sound" "email" "card" "box"
## [3,] "daughter" "android" "dot" "web" "black" "stream"
## [4,] "enjoy" "instal" "qualiti" "surf" "storag" "cabl"
## [5,] "happi" "limit" "tap" "brows" "friday" "remot"
## [6,] "mom" "appl" "bluetooth" "check" "expand" "roku"
## [7,] "mother" "expect" "listen" "basic" "download" "faster"
## [8,] "son" "download" "portabl" "want" "space" "fast"
## [9,] "like" "load" "voic" "video" "movi" "netflix"
## [10,] "birthday" "function" "connect" "access" "add" "cut"
## Topic 7 Topic 8 Topic 9 Topic 10 Topic 11 Topic 12
## [1,] "prime" "turn" "version" "charg" "batteri" "smart"
## [2,] "free" "issu" "model" "connect" "life" "control"
## [3,] "movi" "star" "upgrad" "wifi" "long" "voic"
## [4,] "account" "reason" "paperwhit" "problem" "charg" "turn"
## [5,] "access" "bit" "older" "charger" "hour" "hous"
## [6,] "video" "review" "featur" "issu" "week" "room"
## [7,] "stream" "figur" "origin" "plug" "last" "command"
## [8,] "member" "problem" "reader" "unit" "day" "skill"
## [9,] "netflix" "button" "improv" "power" "hold" "autom"
## [10,] "content" "open" "generat" "replac" "longer" "integr"
## Topic 13 Topic 14 Topic 15 Topic 16 Topic 17 Topic 18
## [1,] "weather" "day" "parent" "qualiti" "reader" "size"
## [2,] "question" "sale" "control" "pictur" "paperwhit" "small"
## [3,] "list" "order" "daughter" "nice" "white" "cover"
## [4,] "answer" "review" "child" "camera" "night" "carri"
## [5,] "news" "decid" "son" "excel" "paper" "fit"
## [6,] "fun" "store" "children" "clear" "eye" "case"
## [7,] "shop" "servic" "case" "pretti" "bright" "hold"
## [8,] "listen" "want" "learn" "high" "librari" "nice"
## [9,] "timer" "month" "download" "fast" "adjust" "hand"
## [10,] "joke" "custom" "age" "amaz" "dark" "travel"
## Topic 19 Topic 20
## [1,] "friend" "ipad"
## [2,] "famili" "money"
## [3,] "user" "worth"
## [4,] "enjoy" "cost"
## [5,] "learn" "spend"
## [6,] "item" "expens"
## [7,] "fun" "extra"
## [8,] "entertain" "pay"
## [9,] "technolog" "mini"
## [10,] "member" "compar"LDA did a fairly good job of picking topics. For instance, Topic 1 is all about celebrations and festivities. Topic 2 seems to be about Google App Store, Android, and download speeds of the apps. Topic 3 is about speaker quality, Topic 4 is about Internet, Topic 5 is about memory and storage, and so on.
Topic 20 looks to be about iPad and how it is worth the money. We can expect that whenever this topic showed up, the reviewer probably rated the corresponding Amazon product lower. We will perform this analysis next.