REDES NEURAIS / INTELIGÊNCIA ARTIFICIAL
TRABALHO 2 – APRENDIZADO DE MÁQUINA
O objetivo do Trabalho 2 é desenvolver um sistema de análise de sentimento para o
reconhecimento de opiniões positivas e negativas em avaliações de filmes.
Exemplo de avaliação positiva:
“I gave this film my rare 10 stars. When I first began watching it and realized it would
not be a film with a strong plot line I almost turned it off. I am very glad I didn't. This
is a character driven film, a true story, which revolves mainly around the life of Rachel
"
Nanny" Crosby, a strong, beautiful (inside and out)Black woman and how she touched
the lives of so many in the community of Lackawanna. Highly interesting not only its
strong characterizations of Nanny and the people who lived at her boardinghouse, but
also it gives us a look at what life and community were like for African Americans in
the 1950's, prior to integration, and the good and bad sides of segregation and how it
ultimately affected and changed the Black community. In addition to excellent
performances by all members of the cast, there is some fine singing and dancing from
that era.”
Exemplo de avaliação negativa:
“I have not seen many low budget films I must admit, but this is the worst movie ever
probably, the main character the old man talked like, he had a lobotomy and lost the
power to speak more than one word every 5 seconds, a 5 year old could act better. The
story had the most awful plot, and well the army guy had put what he thought was army
like and then just went over the top, I only watched it to laugh at how bad it was, and
hoped it was leading onto the real movie. I can’t believe it was under the 2 night rental
thing at blockbusters, instead of a please take this for free and get it out of our sight. I
think there was one semi decent actor other than the woman, I think the only thing OK
with the budget was the make-up, but they show every important scene of the film in the
beginning music bit. Awful simply awful.”
O seu sistema deve ser capaz de reconhecer automaticamente avaliações positivas e
negativas de acordo com o texto fornecido pelo avaliador.
Para essa tarefa, você tem a disposição um conjunto de 50 mil exemplos de avaliações
de filmes extraídos do IMDB (25 mil avaliações positivas e 25 mil avaliações
negativas). Para o desenvolvimento do sistema, você deve utilizar um método de
aprendizado de máquina supervisionado, visto que temos um conjunto rotulado de
exemplos para treinamento (avaliações de filmes). Para isso, você deve seguir os
seguintes passos:
1
) Definir quais serão os atributos que serão usados para descrever os exemplos de
treinamento.