Telmo, I think you left out the main question: WHERE FROM did those texts generate (criticized by you only as for their technical content (different cultures and linguistic backgrounds).
As for the "general problem of AI"? did we ever come to a conclusion how to identify "INTELLIGENCE"? I am inclined to go back to the - - linguistic - origin (Lat) as reading (lego) the meaning(s) INTER, the hidden ones, not the straight vocabulary hit only. And I apply this not only to words proper I include paragraphs, even total contents to be understood even metaphorically, if you like, as 'close-enough' meaning of the written words. Anyway a mind-work above the )materialistic?) human thinking. Then we can start fabricating the 'machine-based' ARTIFICIAL. John M On Tue, Feb 9, 2016 at 12:18 PM, Telmo Menezes <te...@telmomenezes.com> wrote: > Hi Samiya, > > We have to be careful. This uses a technique usually referred to as > "sentiment analysis" and sometimes as "opinion mining". There is extensive > research on using it for things like election forecasting, and the results > are not exactly encouraging... > > The idea is very interesting in itself, but the current methods are quite > limited. The common approaches are: > > 1) Using a dictionary where every word is annotated by humans in terms of > a score for each base emotion, do a lookup for the entire text and present > the final summation; > > 2) Using machine learning to train a model to recognize emotions taking > into account n-grams, instead of a single word. > > The first method is very naif, many words have quite different emotional > valencies depending on context. It also fails to detect sarcasm and other > complexities of human language. > > The second method could in principle work much better, but it requires a > large corpus of text annotated by emotional valencies. Such corpora exist > for specific applications, but models trained that way tend to not work > when you deviate too much from the context of the training data. Religious > texts are most likely too far away from any useful training corpora. > > Worse still, we are comparing translations from vastly different cultures > and linguistic backgrounds. > > Some people suspect (me included) that producing a reliable sentiment > analysis algorithm requires solving the general problem of AI. > > Best, > Telmo. > > On Tue, Feb 9, 2016 at 6:01 PM, Samiya Illias <samiyaill...@gmail.com> > wrote: > >> Bible, Quran and Violence >> Software uses scripture to show what text analysis can do: >> http://m.toledoblade.com/Religion/2016/02/06/The-Bible-the-Qur-an-and-violence-computerized.html >> >> >> Samiya >> >> -- >> You received this message because you are subscribed to the Google Groups >> "Everything List" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to everything-list+unsubscr...@googlegroups.com. >> To post to this group, send email to everything-list@googlegroups.com. >> Visit this group at https://groups.google.com/group/everything-list. >> For more options, visit https://groups.google.com/d/optout. >> > > -- > You received this message because you are subscribed to the Google Groups > "Everything List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to everything-list+unsubscr...@googlegroups.com. > To post to this group, send email to everything-list@googlegroups.com. > Visit this group at https://groups.google.com/group/everything-list. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscr...@googlegroups.com. To post to this group, send email to everything-list@googlegroups.com. Visit this group at https://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.
