The group tries to predict the market's reaction towards possible sentiments contained in news. The study is meaningful because news definitely influences people's daily life, the business, and the stock market. The group uses Thompson Reuters' data, which, I agree, is a reliable source and unlikely to be corrupted. I also like the group's practice of using a dictionary and other trustful resources to determine the sentiments of words. Sentiment can be quite subjective and arbitrary and a professional, respectful authority of language is necessary.
The group uses three steps to develop three models. It is clear to me that the group is doing very organized work and improving its model is a progressive way. The group realized that even for words that are all positive or negative, they will have possibly different weights. The also realized an important fact that when two or more words combined, like "not good," the phrase may have a very different meaning than each individual word.
I do not have an ORIE background so it's hard for me to comment about the technical details of the group's approach. However, assigning a weight to each word and using a multi-gram method seem to be right approaches to tackle the problem.
One thing that confuses me is that the group tries to predict the market's reaction to a news in real time. At first I thought the purpose of this project is to evaluate a whole news' sentimentality. But reading from Section 8 it seems that the group wants to predict the news' sentiment when it is still being typed out? I do not quite understand this part and the group can do a better job explains that.
The group tries to predict the market's reaction towards possible sentiments contained in news. The study is meaningful because news definitely influences people's daily life, the business, and the stock market. The group uses Thompson Reuters' data, which, I agree, is a reliable source and unlikely to be corrupted. I also like the group's practice of using a dictionary and other trustful resources to determine the sentiments of words. Sentiment can be quite subjective and arbitrary and a professional, respectful authority of language is necessary.
The group uses three steps to develop three models. It is clear to me that the group is doing very organized work and improving its model is a progressive way. The group realized that even for words that are all positive or negative, they will have possibly different weights. The also realized an important fact that when two or more words combined, like "not good," the phrase may have a very different meaning than each individual word.
I do not have an ORIE background so it's hard for me to comment about the technical details of the group's approach. However, assigning a weight to each word and using a multi-gram method seem to be right approaches to tackle the problem.
One thing that confuses me is that the group tries to predict the market's reaction to a news in real time. At first I thought the purpose of this project is to evaluate a whole news' sentimentality. But reading from Section 8 it seems that the group wants to predict the news' sentiment when it is still being typed out? I do not quite understand this part and the group can do a better job explains that.