WEBVTT 00:01.140 --> 00:05.590 Heisel, let us have a look at the solution for this particular problem. 00:06.150 --> 00:12.670 So first of all, we have imported the required libraries and we have imported the data set as well. 00:13.110 --> 00:20.850 So the dataset contains two columns, one is revealed and the other one is the sentiment related to 00:20.850 --> 00:20.990 it. 00:21.480 --> 00:25.310 Now, the sentiment is really written as positive and negative. 00:25.320 --> 00:28.620 We need to convert that into a form that is 081. 00:31.100 --> 00:41.360 So what we can see here is that there are total 50000 rows of data when we talk about review the year 00:41.870 --> 00:50.560 around forty nine thousand five hundred unique reviews and the top review, the top number of reviews, 00:50.570 --> 00:57.340 is this a love to be and something and the sentiment related to the top one is positive. 00:58.980 --> 01:01.240 But that sentiment is positive. 01:03.360 --> 01:06.760 And as you can see, the frequency of this is 25000. 01:06.780 --> 01:10.200 So you can see that there are equal distribution of the needle. 01:10.220 --> 01:11.850 So this is not an imbalance data. 01:12.100 --> 01:13.540 It is completely balanced. 01:13.560 --> 01:18.660 It doesn't know he's using the beautiful hope here. 01:19.200 --> 01:21.240 You can use any other password. 01:21.240 --> 01:22.860 You can use any other method. 01:23.790 --> 01:28.170 What I have done here is I have used the esteemable about. 01:30.220 --> 01:33.590 This will read the file in alleged human form. 01:34.780 --> 01:44.260 So what I have done is I will be removing all this bracket from my text and after removing that, I 01:44.260 --> 01:52.690 will be removing all the noisy text and I will apply in this particular function 01:55.210 --> 01:58.030 on this data. 01:59.200 --> 02:07.010 So I will apply the noise and I will employ this remove between square brackets and everything. 02:07.360 --> 02:09.870 So this is my final I have. 02:11.710 --> 02:18.010 And later, I have also removed especially characters from my review because I don't want to keep any 02:18.010 --> 02:19.120 special categories. 02:20.260 --> 02:27.750 Next, what we do is we will be using labor by Malaysia, which will basically create lives for us. 02:28.300 --> 02:31.480 So it will give positive negative labels to us. 02:34.680 --> 02:43.110 So we have used for transport because we want to simply fix this by labor, by analyze it and apply 02:43.110 --> 02:43.380 it. 02:43.410 --> 02:45.540 We don't want to keep it for further use. 02:45.720 --> 02:52.590 If you want to keep it for the use, you will use it and then you will use transport for each transformation 02:52.590 --> 02:53.300 you want to have. 02:54.970 --> 02:58.450 Next, what we do is we get the sentiments. 03:01.380 --> 03:14.060 So 40000 sentiments into brain sentiment and then spectacular views, then the next 10000 in the best 03:14.220 --> 03:14.940 sentiment. 03:16.200 --> 03:20.870 This is basically my data splitting endorphin status. 03:22.940 --> 03:31.280 Next, what I do is I created this vinyl vectorized, you could use any other vectorized account representivity 03:31.920 --> 03:33.270 you can use in your office. 03:34.280 --> 03:37.210 What we do is we use the video victimiser. 03:37.250 --> 03:39.740 The minimum document frequency is 20. 03:40.040 --> 03:42.770 Maximum document frequency is zero point five. 03:43.010 --> 03:43.850 That means 50. 03:44.020 --> 03:52.130 Some document like this, it marks a word can occur in 50 percent on the entire use. 03:53.630 --> 03:55.440 Makes this Ingram rich. 03:56.270 --> 03:59.060 So this is the range which I want to keep. 03:59.090 --> 04:02.200 So I will be involved in parallel. 04:02.210 --> 04:03.740 So I'm a single world. 04:06.060 --> 04:15.700 Next, the greed of the transfer of this particular the effect ricin on the brain. 04:16.920 --> 04:23.730 So on the training review we use to transform and we use transform all the best of uses, but so we 04:23.730 --> 04:25.170 do the transformations. 04:25.170 --> 04:28.410 We have converted them into the idea of vectors. 04:30.060 --> 04:34.260 Now you can see the details. 04:34.330 --> 04:43.140 The the shape of the review is 40000 and 60000, 356 and 60000, 356. 04:43.500 --> 04:47.190 This means that you have around 60000. 04:49.760 --> 05:00.410 Can't read the fine yet because of the size of the BFI, there is a 60000, three, 56 columns and 40000 05:00.710 --> 05:01.090 euros. 05:01.380 --> 05:07.810 Now, again, this is in the form of a sparse matrix, because when we use any DFAT electrodes that 05:07.820 --> 05:10.730 are found, the president will give us an sportsplex. 05:12.360 --> 05:18.150 Now we will be applying logistic regression, you can apply any of other algorithm you want, you can 05:18.150 --> 05:21.790 use maybe you can use SVM, it's completely up to you. 05:22.140 --> 05:23.750 This also performs limit. 05:24.030 --> 05:25.470 We can use any algorithm. 05:26.040 --> 05:29.400 So I'm simply applying logistic regression name for thing. 05:29.400 --> 05:34.020 And then I'm finding out the models for it comes out with zero point nine four. 05:34.290 --> 05:42.420 Now, this is the very simplistic implementation and the model score on the test comes out to be 90 05:42.420 --> 05:42.960 percent. 05:45.550 --> 05:54.280 Next, I think the confusion matrix for this and the final report of which shows that it is performing 05:54.280 --> 06:01.500 pretty well for us next, what we're doing is we are converting these features into coefficient. 06:01.690 --> 06:04.960 So these are two different features that we have. 06:04.960 --> 06:14.230 And we have also got the best words, which we have in the descriptions in these columns. 06:14.530 --> 06:17.870 So they are great, excellent, perfect, wonderful best. 06:17.890 --> 06:22.550 So these are the most exciting, positive words which we have in our vehicles. 06:23.500 --> 06:29.480 The most negative words are by far the most awful, most boring. 06:30.760 --> 06:41.950 Now we are all getting the part of speech of the words and the positive and negative words. 06:42.880 --> 06:49.840 So we have basically got the words and sort of them got the features. 06:49.840 --> 06:51.220 That is the column names. 06:51.700 --> 06:54.970 And we have simply got the details from this. 06:55.540 --> 06:58.280 And we have also created a word cloud here. 06:58.280 --> 07:01.930 And this is the word cloud, which I have the idea from the word to do what you. 07:04.670 --> 07:08.840 From all the positive and all the negative thoughts you have, just get in there then all the words 07:08.840 --> 07:15.260 here and we have been counting all the words because word cloud, you can create an entire sentence 07:15.260 --> 07:16.730 on a continuous text. 07:17.570 --> 07:19.400 So we have created a word cloud. 07:19.460 --> 07:27.410 Actually, here you can see how this entire text is present and what are the most upsetting words in 07:27.410 --> 07:28.850 the positive reviews. 07:28.850 --> 07:37.370 So in the positive review, the most appealing one that excellent will save best read. 07:38.060 --> 07:40.130 Brilliant, wonderful phoniest. 07:40.640 --> 07:43.220 See, these are the good ones which you find in the reviews. 07:43.710 --> 07:52.040 When we talk about the bad reviews or the negative reviews, you can see boring waste, terrible wars, 07:52.040 --> 07:59.690 ridiculous, dull by boring, awful, poor quality. 07:59.930 --> 08:01.760 These are the words which are getting heilig. 08:02.520 --> 08:08.000 You can see what are the major words which are actually giving this result to us. 08:10.000 --> 08:11.740 So this is the implementation. 08:12.160 --> 08:18.730 This is a very simple implementation, and you can use many different methods for it and this can be 08:18.730 --> 08:21.900 up to your choice and how you want to implement. 08:21.910 --> 08:28.090 This is one of the most simplistic implementation which we have hidden text classification. 08:28.390 --> 08:28.780 Thank you.