Text Only | Text with Attachments

QB64.org Forum

Active Forums => Programs => Topic started by: SpriggsySpriggs on July 22, 2020, 09:29:34 am

Title: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 22, 2020, 09:29:34 am
Use this demo here to type a line of text and send it to the sentim API machine learning AI to see if a phrase is negative, neutral, or positive in connotation. Many thanks to Luke for his assistance in understanding cURL strings.
This demo is very limited. The output from the AI actually contains more data than "negative", "positive", or "neutral". It even contains how negative or how positive something is. You could pass back that value and add more IF statements or whatever that could detect whether something is very nice or very mean.

Here are some screenshots:
 
 


And the source code:
Code: QB64: [Select]
  3.  
  5. DIM result AS STRING
  6.  
  7. LINE INPUT "Please input what text you would like the API to check: ", text
  8.  
  9. result = CheckConnotation(text)
  10. IF result = "negative" THEN
  11.     PRINT "That's not a nice thing to say."
  12. ELSEIF result = "positive" THEN
  13.     PRINT "Positivity is key."
  14. ELSEIF result = "neutral" THEN
  15.     PRINT "Meh."
  17.     PRINT "Well, that seems to not work. Check your connection or print the string that is being SHELLed to make sure the syntax isn't awry."
  19.  
  21.  
  22. FUNCTION CheckConnotation$ (Text AS STRING)
  23.     DIM sentim AS STRING
  24.     SHELL _HIDE "curl -XPOST -H " + CHR$(34) + "Accept: application/json" + CHR$(34) + " -H " + CHR$(34) + "Content-type: application/json" + CHR$(34) + " -d " + CHR$(34) + "{\" + CHR$(34) + "text\" + CHR$(34) + ": \" + CHR$(34) + Text$ + "\" + CHR$(34) + "}" + CHR$(34) + " " + CHR$(34) + "https://sentim-api.herokuapp.com/api/v1/" + CHR$(34) + " -o sentimapirequest"
  25.  
  26.     IF _FILEEXISTS("sentimapirequest") THEN
  27.         C = FREEFILE
  28.         OPEN "sentimapirequest" FOR BINARY AS #C
  29.         IF LOF(C) <> 0 THEN
  30.             sentim = SPACE$(LOF(C))
  31.             GET #C, , sentim
  32.             CLOSE #C
  33.             sentim = MID$(sentim, INSTR(sentim, "type" + CHR$(34) + ":") + 7)
  34.             sentim = LEFT$(sentim, INSTR(sentim, CHR$(34)) - 1)
  35.             CheckConnotation = sentim
  36.             KILL "sentimapirequest"
  37.             EXIT FUNCTION
  38.         ELSE
  39.             CLOSE #C
  40.             CheckConnotation = ""
  41.             KILL "sentimapirequest"
  42.             EXIT FUNCTION
  43.         END IF
  44.     ELSE
  45.         CheckConnotation = ""
  46.         EXIT FUNCTION
  47.     END IF
Title: Re: Check connotation of language with machine learning API!
Post by: loudar on July 22, 2020, 02:48:41 pm
Got this. What's wrong with it?

  [ This attachment cannot be displayed inline in 'Print Page' view ]  
Title: Re: Check connotation of language with machine learning API!
Post by: bplus on July 22, 2020, 03:17:08 pm
Worked for me, a little mod:
Code: QB64: [Select]
  3.  
  5. DIM result AS STRING
  6.  
  8.     LINE INPUT "Please input what text you would like the API to check: ", text
  9.     IF text = "" THEN EXIT DO
  10.     result = CheckConnotation(text)
  11.     IF result = "negative" THEN
  12.         PRINT "That's not a nice thing to say."
  13.     ELSEIF result = "positive" THEN
  14.         PRINT "Positivity is key."
  15.     ELSEIF result = "neutral" THEN
  16.         PRINT "Meh."
  17.     ELSE
  18.         PRINT "Well, that seems to not work. Check your connection or print the string that is being SHELLed to make sure the syntax isn't awry."
  19.     END IF
  20.     PRINT: PRINT " 5 sec delay..."
  21.     _DELAY 5
  23.  
  24. FUNCTION CheckConnotation$ (Text AS STRING)
  25.     DIM sentim AS STRING
  26.     SHELL _HIDE "curl -XPOST -H " + CHR$(34) + "Accept: application/json" + CHR$(34) + " -H " + CHR$(34) + "Content-type: application/json" + CHR$(34) + " -d " + CHR$(34) + "{\" + CHR$(34) + "text\" + CHR$(34) + ": \" + CHR$(34) + Text$ + "\" + CHR$(34) + "}" + CHR$(34) + " " + CHR$(34) + "https://sentim-api.herokuapp.com/api/v1/" + CHR$(34) + " -o sentimapirequest"
  27.  
  28.     IF _FILEEXISTS("sentimapirequest") THEN
  29.         C = FREEFILE
  30.         OPEN "sentimapirequest" FOR BINARY AS #C
  31.         IF LOF(C) <> 0 THEN
  32.             sentim = SPACE$(LOF(C))
  33.             GET #C, , sentim
  34.             CLOSE #C
  35.             sentim = MID$(sentim, INSTR(sentim, "type" + CHR$(34) + ":") + 7)
  36.             sentim = LEFT$(sentim, INSTR(sentim, CHR$(34)) - 1)
  37.             CheckConnotation = sentim
  38.             KILL "sentimapirequest"
  39.             EXIT FUNCTION
  40.         ELSE
  41.             CLOSE #C
  42.             CheckConnotation = ""
  43.             KILL "sentimapirequest"
  44.             EXIT FUNCTION
  45.         END IF
  46.     ELSE
  47.         CheckConnotation = ""
  48.         EXIT FUNCTION
  49.     END IF
  51.  
  52.  

"meh"  LOL
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 22, 2020, 04:59:05 pm
Quote from: loudar on July 22, 2020, 02:48:41 pm
Got this. What's wrong with it?
@loudar I'm not taking into account things like apostrophes or question marks. I'm new to JSON formats still and command prompt is still weird about characters and delimiters. You might have to escape the character. Nope, that's not it because bplus didn't change my function at all and was able to use both apostrophes and question marks. Maybe it's a connection issue? Try visiting the site in your browser and see if you can reach it. https://sentim-api.herokuapp.com (https://sentim-api.herokuapp.com)
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 22, 2020, 05:07:38 pm
Quote from: bplus on July 22, 2020, 03:17:08 pm
"meh"  LOL
@bplus Glad you liked it haha
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 22, 2020, 09:37:38 pm
Perhaps this weekend I'll look into using the "polarity" value in the result JSON that shows just how positive or negative a statement is. Can make for more flexibility.
Title: Re: Check connotation of language with machine learning API!
Post by: Ashish on July 22, 2020, 11:39:53 pm
Nice... @SpriggsySpriggs
Quote
Please input what text you would like the API to check: you should kill yourself
Meh.
How this can be neutral? ROFL
Title: Re: Check connotation of language with machine learning API!
Post by: SMcNeill on July 23, 2020, 04:28:18 am
Quote from: Ashish on July 22, 2020, 11:39:53 pm
Nice...How this can be neutral? ROFL

A lot of these type programs are first generation apps.  All they can do is look at the surface of what is written, and attempt to rate things based solely on the words used.  Much like profanity filters, they lack the capacity to understand user context.

Two friends talking:

Friend One: "Man, I can't believe I failed my class!  I thought the exam was tomorrow, not today.  I completely missed it."   
Friend Two (sarcastically): "You're such a genius!"

Those statements would be neutral for friend one, but positive for friend two.  Normally, calling someone a genius is positive reinforcement, but in this case it's sarcastic.  These programs react to words used, not the context of which the words are used.  They're cute to play with, and maybe someday they'll inspire someone to go the extra mile and sort out the nuances of subtle context, but as they are now, I wouldn't depend on them for any sort of diagnosis for my mental state or anything.  (And I have seen some people who attempt to do that with journals which grade their writings.  "You are 87% positive in your daily writings...")
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 23, 2020, 08:03:51 am
Quote from: Ashish on July 22, 2020, 11:39:53 pm
Nice... @SpriggsySpriggsHow this can be neutral? ROFL
@Ashish That's because I'm only grabbing the final result from the JSON file. There is more information like polarity that shows just how much it sways either way. I just chose neutral, negative, or positive as the final answer.
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 23, 2020, 08:13:44 am
Quote from: SMcNeill on July 23, 2020, 04:28:18 am
A lot of these type programs are first generation apps.  All they can do is look at the surface of what is written, and attempt to rate things based solely on the words used.  Much like profanity filters, they lack the capacity to understand user context.
@SMcNeill ,
This particular API is supposedly machine learning so as time goes on I reckon it will become more accurate in determining meaning of words. Similar to Cleverbot in how it uses user input and responses to generate more fluid dialog and as time goes on it has more natural responses.

This one is probably not used nearly as much as the webpage for it is rather sparse in both documentation as well as design quality. The profanity filter only recognized words and filtered. The more text you send to this API in a string the better the result. Short sentences like "I hate you" or "I love you" are cut and dry whereas a paragraph requires more scanning to get more context. The website example shows using a paragraph as the passed string and returning back a value. It can analyze many sentences at once.
Title: Re: Check connotation of language with machine learning API!
Post by: loudar on July 23, 2020, 12:16:00 pm
Quote from: SpriggsySpriggs on July 22, 2020, 04:59:05 pm
@loudar I'm not taking into account things like apostrophes or question marks. I'm new to JSON formats still and command prompt is still weird about characters and delimiters. You might have to escape the character. Nope, that's not it because bplus didn't change my function at all and was able to use both apostrophes and question marks. Maybe it's a connection issue? Try visiting the site in your browser and see if you can reach it. https://sentim-api.herokuapp.com (https://sentim-api.herokuapp.com)

Site works, but very slow.
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 23, 2020, 12:55:42 pm
@loudar Then it is just your connection. It's probably just timing out.
Title: Re: Check connotation of language with machine learning API!
Post by: bplus on July 23, 2020, 04:56:33 pm
Quote
This one is probably not used nearly as much as the webpage for it is rather sparse in both documentation as well as design quality. The profanity filter only recognized words and filtered. The more text you send to this API in a string the better the result. Short sentences like "I hate you" or "I love you" are cut and dry whereas a paragraph requires more scanning to get more context. The website example shows using a paragraph as the passed string and returning back a value. It can analyze many sentences at once.

Can't do that without feedback and it's not getting that from us.
Title: Re: Check connotation of language with machine learning API!
Post by: SpriggsySpriggs on July 23, 2020, 06:12:33 pm
@bplus Oh well. I'm not really going to be using this in any real world applications. It's mostly to show usage of CURL and APIs. The only APIs I plan on using elsewhere are the Weather API and all the data validation APIs.
Text Only | Text with Attachments