Artificial Intelligence with Java

[Giovanni R.]

I'd like to create an application using Java where users give a phrase in input and the application says if it's a /negative/ or /positive/ phrase.

Where should I start?
Has AI already solved this kind of problems?

Thanks.

 

[Josip Almasi]

I'd like to create an application using Java where users give a phrase in input and the application says if it's a /negative/ or /positive/ phrase.

Where should I start?

Here, I guess:
http://www.alicebot.org/downloads/programs.html

Has AI already solved this kind of problems?

Responding to user input - yes
Analysing phrases - yes.
Recognizing negative/positive phrases - not to my knowledge.

Regards...

 

[Jeff Higgins]

I'd like to create an application using Java where users give a phrase in input and the application says if it's a /negative/ or /positive/ phrase.

Where should I start?
Has AI already solved this kind of problems?

Thanks.

A priori polarity shouldn't be too hard.
Contextual polarity maybe harder.
Google Sentiment+Analysis.
An extended example:
<http://alias-i.com/lingpipe/demos/tutorial/sentiment/read-me.html>

 

[Stefan Ram]

I'd like to create an application using Java where users give
a phrase in input and the application says if it's a
/negative/ or /positive/ phrase.

Is the following sentence negative or positive and why?

»Nella predisposizione dei turni di ferie, nei limiti delle
esigenze di servizio, viene data precedenza ai lavoratori
disabili rientranti nelle categorie di cui alla legge
12.03.1999 n. 68; per il restante personale si tiene conto
delle richieste degli interessati in rapporto alla loro
situazione familiare e alla loro anzianità di servizio.«

 

[Jeff Higgins]

A priori polarity shouldn't be too hard.
Contextual polarity maybe harder.
Google Sentiment+Analysis.
An extended example:
<http://alias-i.com/lingpipe/demos/tutorial/sentiment/read-me.html>

Online sentiment detector with Java library.
<http://sentistrength.wlv.ac.uk/>

 

[Jeff Higgins]

Online sentiment detector with Java library.
<http://sentistrength.wlv.ac.uk/>

The text 'It has been real and it has been fun but it has not been real
fun.'
has positive strength 2 and negative strength -1

 

[Jeff Higgins]

A priori polarity shouldn't be too hard.
Contextual polarity maybe harder.
Google Sentiment+Analysis.
An extended example:
<http://alias-i.com/lingpipe/demos/tutorial/sentiment/read-me.html>

Memento for me. "Recognizing Contextual Polarity in Phrase-Level
Sentiment Analysis
"
<http://www.cs.pitt.edu/~wiebe/pubs/papers/emnlp05polarity.pdf>

 

[Sarah Connor]

I'd like to create an application using Java where users give a phrase
in input and the application says if it's a /negative/ or /positive/
phrase.

Where should I start?
Has AI already solved this kind of problems?

Consider that the same phrase, "good job!", can mean it was a good job or
it was a terrible job depending on the context. (Ex.: In Star Wars Attack
of the Clones when Anakin and Padme have been captured and are chained up
next to Obi-Wan, he asks what they're doing here. Anakin says they came
to rescue him. Cue "good job!".)

The day AI can reliably tell when "good job!" is positive or negative is
the day we can all stop worrying about keeping *our* jobs -- one way or
another.

 

[Jeff Higgins]

The phrase you want to feed into a search engine is "natural language
processing".


Probably (or to put it another way -- I doubt whether AI is necessary):
depending on what you mean by negative or positive: for instance is "<some
name> is a <some insult>" negative (because of the insult) or positive (because
it makes a positive statement) ?

As an exercise, consider what part of the problem /wouldn't/ be solved by
testing whether the phrase contained the word "not" (including contractions).

Memento for me. "Contextual Valence Shifters"
<http://aaaipress.org/Papers/Symposia/Spring/2004/SS-04-07/SS04-07-020.pdf>

 

[Jeff Higgins]

Memento for me. "Contextual Valence Shifters"
<http://aaaipress.org/Papers/Symposia/Spring/2004/SS-04-07/SS04-07-020.pdf>

"Learning Extraction Patterns for Subjective Expressions"
<http://www.cs.utah.edu/~riloff/pdfs/emnlp03.pdf>

 

[Eric Sosman]

[...]
As an exercise, consider what part of the problem /wouldn't/ be solved by
testing whether the phrase contained the word "not" (including contractions).

Ain't nuthin' no better nor whut he said.

 

[Jeff Higgins]

Ain't nuthin' no better nor whut he said.

Aspects of English Negation
Yoko Iyeiri
John Benjamins Publishing, 2005

When Voices Clash: A Study in Literary Pragmatics
Jacob Mey
Walter de Gruyter, 1999

 

[Jeff Higgins]

I ain't gwine to tell no mo' 'cause I ain't to make
statement and testify 'bout sumpin' I ain' know about.

 

[Jeff Higgins]

Ain't got no cigarettes.

We don't need no education.

 

[Jeff Higgins]

We don't need no education.

But back to the original topic.

"The Polarity Items Bibliography"
<http://www.english-linguistics.de/sfb441/a5/pib/XML2HTML/list.html>

No reply back yet, but maybe the OP has
discovered something about web search engines.

 

[Andreas Leitgeb]

Ain't got no cigarettes.

Definitely positive (for your health) ;-)