Week 6: Enhancing my Sentiment Analysis Model
Krish S -
Hey folks! I know that last week I said that the sentiment analysis code will be done before this week’s post. I wasn’t lying, but I’ve also decided to tweak some things. After I finished the basic sentiment analysis model, I wasn’t entirely satisfied for two reasons: 1) there’s still around a month left before we need to be done with our research and so I don’t want to sit around and do nothing all day and 2) my sentiment analysis model doesn’t capture the full biases present in the film reviews. For those two reasons, I have decided to add to my code and make it just a little more complex.
By ‘just a little more complex,’ I mean two things. The first is making the sentiment analysis differentiate between gender. So, If female actors are more frequently described using appearance-focused or emotional words, while male actors are described with strength and leadership terms, that suggests bias, and I would want my sentiment analysis model accounting for that when assigning quantitative scores. The second is inputting code that tracks frequency analysis. Now, although film reviews are centered around the actor’s performance and the role of the character that the actor played, they are not entirely about the actor. There are other things talked about as well, like the director, supporting actors/characters, etc. However, that is another area of bias that could be looked into. I could focus on how frequent a male actor is described versus how frequent a female actor is described in their respective film reviews. Thus, I want my code to be able to decrease the sentiment score for an actor if he/she is talked about less frequently. Now, I am not a computer science wizard, so this might take a maximum of two weeks to do. I’ll keep you updated on my progress of adding both of these things to my code.
Comments:
All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.