We have now seen that vector space operations did not work too well regarding the predictive accuracy of our model. In this section, we will attempt a technique which is very different and is closer to the semantic parsing model we mentioned at the beginning of this chapter. We will try sentiment analysis.
We will not only take into account the words in a text, but we will also take into account shifters (that is, negators, amplifiers, de-amplifiers, and adversative conjunctions). A negator flips the sign of a polarized word (for example, I do not like it). An amplifier increases the impact of a polarized word (for example, I really like it.). A de-amplifier reduces the impact of a polarized word (for example, I hardly like it). An adversative conjunction overrules the previous clause containing a polarized word (for example, I like it but...