> news articles are objective. HAHAHAHAHAHA
Just publishing an article is a subjective statement. Once you add the fact that virtually all large media conglomerates are owned by companies with corporate agendas outside of media and you will NEVER have objective mainstream news. Even a public broadcaster, like the BBC, has to have the larger interests and worldview of its democratic owner (the people of the United Kingdom) at heart. On 7/18/05, Matthew Hurst <[EMAIL PROTECTED]> wrote: > There are a number of companies that are currently competing in the marketing > intelligence space that have developed sentiment or polarity mining > systems (Intelliseek, > my employer, being one of them). The general buckets into which this > work falls include > > 1) affect analysis - grokking the emotional content of text > 2) polarity analysis - detecting author statements about favourable and > unfavourable conditions or opinions > 3) subjectivity analysis - are the statements subjective or objective? > > Ideally, journalism should be objective and so notions of affect (emotion) > ought > not to come into it unless reporting (objectively!) some other agent's > feelings. > > However, there are plenty of issues that are favourable/unfavourable. The > approach of a hurrican is unfavourable, the rescuing of a baby from a > crocodile > pit is favourable. This suggests thre requirements: > > 1) detection of favourable/unfavourable topics (bombs, murder, etc.) > 2) tracking of the development of stories (person rescued from kidnappers) > 3) the anlaysis of stories (rescue fro mburning building). > > #1 is probably reasonably well done with a keyword list and some knowledge of > an article being the first in a story arc (potentially to be continued > as the story develops). > #2 is pretty hard, as is > #3 > > I'd be interested in throwing our polarity system against the BBC news > feed to see > what happens when we ignore the assumption that news articles are > objective. If I have > the time to do this I'll report the results here... > > Matt Hurst > http://datamining.typepad.com > > > > On 7/18/05, Kim Plowright <[EMAIL PROTECTED]> wrote: > > >its effectively spam-assassin for news, but slightly different. > > > > Wow - that's a slightly terrifying concept: the ability to filter news > > according to your personal preferences so you only get 'good' news > > delivered to you... Very 1984. *Shudder* > > > > The mood would indeed only work effectively if it leant your preferences > > and filtered accordingly - in effect, the current system suggests that > > it is 'Bad news in the eyes of right thinking people' - it would look > > different if the intended audience was, say, Bond Villains. > > > > Thought - wonder if a system with user-based feedback loop would be > > effective at catching stories that are 'spun'? Ie, cross referencing a > > 'mood' against the yes/no votes of the users would yield some way of > > spotting editorial bias, or stories that are released to sweeten > > potentially damaging/worrying stories with palliative good news (not > > necessarily by the journalists, but maybe at source)? > > > > Have you thought about running a similar 'mood detector' through video > > transcripts, or ficiton? It could be a useful addition to a > > reccomendation engine? > > > > kim > > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] On Behalf Of Luke Dicken > > Sent: 18 July 2005 11:13 > > To: [email protected] > > Subject: RE: [backstage] Mood News 3 > > > > > > > A really clever site would ask you for your point of view (or > > > remember when > > > you disagree) and adjust accordingly. A complex task though... > > > > A feedback neural network should be able to solve it reasonably > > straight-forwardly (although you would need a large sample of data with > > which to train the network on a per-user basis, which could also give > > large overheads) - its effectively spam-assassin for news, but slightly > > different. > > > > > > - > > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, > > please visit > > http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > > > > > > - > > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please > > visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > > > > > - > Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please > visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html. > - Sent via the backstage.bbc.co.uk discussion group. To unsubscribe, please visit http://backstage.bbc.co.uk/archives/2005/01/mailing_list.html.
