[MUSIC] One of the most common types of classifiers is a so called linear classifier. So let's talk a little bit about that. The question here is how do we represent the classifier? So we start with some sentences, for example in a sentiment the mouse's case, the classifier. You get a prediction whether it is a positive sentence or a negative sentence. So, how does this classifier work? In the sentimental analysis you can imagine a simple kind of threshold classifier. Suppose that I take a sentence and somebody tells me these are all of the positive words, great, awesome, good amazing and so on. Here's a bunch of negative words. Bad, terrible, disgusting, food, and so on. And so what I can do is take the sentence and count how many positive words there are in a sentence and how many negative words are in the sentence, just count those. And then say if the number of positive words is higher than the number of negative words. Then we have a positive sentence but if they use more negative more words then you have a negative sentence. So for example, if the input sentence that we have is the sushi was great positive one, the food was awesome positive two, but the service was terrible negative one. You have two positives one negative and in the end was the positive wins and so you have a positive prediction. Now threshold classifiers have some limitations. Where does this list of positive and negative words actually come from? It has to magically come from somewhere and not just that, words have different degrees of positive and negativeness. So great is more positive than good. You wanna tune and figure out what's better great, good, amazing, is amazing better than great? Who knows? So how do we figure that out, how do we weigh different words? And single words might not be enough to make good classification. So good food, so the food was good is positive. The food was not good is actually negative. And so these are all issues that need to be addressed. The first two areas where the positive and negative words come from and how do you weigh them comes from learning a classifier and we're going to talk about next. The issue of good versus not good is addressed by using more complicated features than single words. And we're gonna talk about it towards the end of the module. So a linear classifier, instead of having a list of positive and negative words, actually takes all the words and adds weights to them. So, for example, good might have a weight of 1, great might have a weight of 1.5, awesome might have really big weight of 2.7. While bad might have a weight of -1, terrible may have a weight of -2.1, awful might be -3.3 cuz awful is really just awful. And this ways that really don't matter to sentiment. So things like the, we, where, restaurant, they appear both in positive and negative sentences so they get weight zero. Suppose somebody magically told you what the weight of each word were. We're going to talk about it in a little bit how those get learned by the classifier. But given those weight, how to figure out if the sentence is positive or negative. Here we use the idea of scoring. So for example take this sentence. The sushi was great, the food was awesome, but the service was terrible. Let's score that sentence. So we're gonna compute this, the score of the input sentence x. So in this case, you get from great, you get positive 1.2. From awesome, you get another 1.7 but from table, you get minus 2.1 right here, and so the grand total here is 2.9 minus 2.1, which is 0.8. And the key here is that since the score of the sentence is greater than zero, we're going to predict that it is a positive sentence. If the score was the opposite, if the score of x were less than zero, then we'd have predicted it's a negative sentence. So this is how a linear classifier works, if you know the weight of each word, and this is called a linear classifier because the output is basically the weighted sum of the input. Just weight, what features appears, what words appear in the input. So we're working for simple linear classifier to start out with. So in summary, given a sentence and given the weights for the sentence, what we do is compute the score, which is the weighted count of the words that appear in the sentence. And then we say if the score is greater than zero, we predict y-hat to be positive. While if the score is less than zero, we predict it to be negative. And that is a linear classifier. [MUSIC]
