Input: Staminon Avoid some things like alcohol , cigarettes and caffeine .
Parse:
Avoid VB ROOT
+-- Staminon NNP nsubj
+-- things NNS dobj
| +-- some DT det
| +-- like IN prep
| +-- alcohol NN pobj
| +-- , , punct
| +-- cigarettes NNS conj
| +-- and CC cc
| +-- caffeine NN conj
+-- . . punct
Input: All of these listed merchandise is drugs which alter requires at least in same shape or form .
Parse:
drugs NNS ROOT
+-- All DT nsubj
| +-- of IN prep
Interesting. Looks like it just gives up on a Wingding post
Warcraft NNP ROOT
+-- Watch NNP nn
+-- 2016 CD appos
| +-- -LRB- -LRB- punct
| +-- -RRB- -RRB- punct
+-- movie NN dep
| +-- full JJ amod
+-- , , punct
+-- Online NNP appos
| +-- Watch NNP nn
| +-- Movie NNP nn
| +-- Warcraft NNP nn
| +-- -LRB- -LRB- punct
| +-- 2016 CD nn
| +-- -RRB- -RRB- punct
| +-- Full NNP nn
+-- Warcraft NNP appos
| +-- Watch NNP nn
| +-- 2016 CD appos
| +-- -LRB- -LRB- punct
| +-- -RRB- -RRB- punct
Looks like it tries to assign a verb as a root if it possible can, otherwise it falls back to a noun
Sentences with noun roots tend to be poorly formed
This could be very good for filtering out false positives on expletives.
Input: Because fuck you , thats why .
Parse:
thats NNS ROOT
+-- fuck NN dep
| +-- Because IN mark
| +-- you PRP dobj
+-- , , punct
+-- why WRB dep
+-- . . punct
Input: The word fuck was not till the 1900 's sex term call a poke .
Parse:
was VBD ROOT
+-- fuck NN nsubj
| +-- The DT det
| +-- word NN nn
+-- not RB neg
+-- call VBP advcl
| +-- till IN mark
| +-- term NN nsubj
| | +-- 1900 CD poss
| | | +-- the DT det
| | | +-- 's POS possessive
| | +-- sex NN nn
[ SmokeDetector ] URL in title, bad keyword in body, link at end of body, link following arrow in body, pattern-matching website in body: http://iqilicaqenagi.tumblr.com/ by user22734 on arduino.stackexchange.com (@NickGammon) …
[ SmokeDetector ] URL in title, bad keyword in body, bad keyword in title, link at end of body, pattern-matching website in body, pattern-matching website in title: http://musclebuildingbuy.com/dermaserre/ by Salvcan on security.stackexchange.com…
@Undo buggity bug: this search should not have tp's in that graph.
Also, were you thinking of using McParseface to help us parse spam? I'm thinking we could certainly use it to detect if an offensive answer is targeted at someone or just general.
@hichris123, You've been doing ML against posts, correct? Have you developed a list of stop words? (The most common words). I'm looking for the next level beyond (a, I, have, has, the) etc.
@SmokeDetector I fpu'd that from Tavern because when I checked his other account, it seems a legit answer. However, after searching the site on metasmoke, now I'm doubting if those accounts are in voting ring...
Now accepting donations to go toward an Nvidia Digits DevBox (Deep learning machine)...estimated cost: $15K. I promise to share with Smokey developers :) (Found via the PyImageSearch blog which has some decent image manipulation and ML tutorials)
@ArtOfCode, that's one of my issues too. I've seen a lot of good Linux tools, but my work environment uses Windows so that what my stack is on both the work machine and home machines so that I can move between the two fairly easily.
Ah. I'm planning on updating the desktop to 10 in the next month. I've been dragging my feet on the main computer I use. I've upgraded the other two machines.
@Ferrybig Honestly, I'm not worried about privacy. I trust that Microsoft won't share my email address without telling me; that's all I really care about.
I'm doing some experimentation with SyntaxNet (aka Parsey McParseface). You can follow my musings and result in Charcoal HQ if you so desire.
Based on some early results using very rudimentary ways to evaluate the boatloads of data I get from Parsey, I think there's a decent chance that somethin...
