1 00:00:01,649 --> 00:00:09,760 So we looked at the sequence matcher class which returns the ratio method of this 2 00:00:09,760 --> 00:00:17,679 class. It returns this number, but now we need to see how to get the most similar 3 00:00:17,679 --> 00:00:25,689 word out of a list, out of the keys of a dictionary, so let's focus first on 4 00:00:25,689 --> 00:00:33,640 on a list on sequence. So a list is a sequence and from difflib import 5 00:00:33,640 --> 00:00:46,720 get_close_matches. If you do help get_close_matches you can see the 6 00:00:46,720 --> 00:00:52,900 arguments that you can pass to this method so the first one in the word so 7 00:00:52,900 --> 00:00:57,250 that will be the word that the user passes, and then you have a sequence 8 00:00:57,250 --> 00:01:01,960 of possibilities and so word is a sequence for which close matches are 9 00:01:01,960 --> 00:01:06,429 desired typically a string, so typically this would be a string that you want to 10 00:01:06,429 --> 00:01:12,520 compare and possibilities is a list of sequences or a list of strings 11 00:01:12,520 --> 00:01:21,390 you can say. So typically it's a list of strings. And then you have n equals to 3. 12 00:01:21,390 --> 00:01:27,039 That will define the number of the matches that you want the get close 13 00:01:27,039 --> 00:01:31,840 matches method to return. Let's say you pass a list of possibilities with six 14 00:01:31,840 --> 00:01:40,689 items and each item has that ratio so from 1 to 0, from 0 to 1 and this method 15 00:01:40,689 --> 00:01:45,840 will return the three most similar matches, the most similar strings to the 16 00:01:45,840 --> 00:01:52,630 word passed in here. You can change that argument value if you and you have the 17 00:01:52,630 --> 00:02:00,369 cut-off, so that's the ratio here. By default this will return all 18 00:02:00,369 --> 00:02:06,969 with those items that have a ratio of at least 0.6 or greater than that 19 00:02:06,969 --> 00:02:11,819 so under that the matches will be not included. 20 00:02:14,370 --> 00:02:23,350 Q2 end, then help and get let me try close matches. Let's see what we get. 21 00:02:23,350 --> 00:02:33,250 Let's say rain and then you have a list in square brackets of course. 22 00:02:33,250 --> 00:02:47,770 Let's say help, pyramid and rain. And I'll leave the n and the cut off as they are by 23 00:02:47,770 --> 00:02:54,010 default. So in this case you get your rain the reason you get rain is 24 00:02:54,010 --> 00:03:01,150 because the rain had a cut-off of greater than 0.6 or similar to that. If 25 00:03:01,150 --> 00:03:05,620 you're interested about the cutoff you can use the sequence matcher but 26 00:03:05,620 --> 00:03:12,010 we're not interested in that in the specific value. So rain was the 27 00:03:12,010 --> 00:03:17,020 only one who satisfied those conditions, so we get rain here as a 28 00:03:17,020 --> 00:03:24,580 list. So okay, we were able to get the close match out of a list, but how do we get 29 00:03:24,580 --> 00:03:31,030 the close match out of our dictionary keys? We still have the data variable here 30 00:03:31,030 --> 00:03:40,870 which I loaded previously here, up in this session. I'll load it from the JSON 31 00:03:40,870 --> 00:03:50,920 file so with a data.keys so there's a method of dictionaries which is called 32 00:03:50,920 --> 00:03:57,700 keys and with that you can get a list of all the keys of the dictionary, so we're 33 00:03:57,700 --> 00:04:01,780 talking about only the keys, not the values of the keys. So these are the 34 00:04:01,780 --> 00:04:07,780 words and each of them has a value associated with it and basically you can 35 00:04:07,780 --> 00:04:15,070 check the type of data.keys just see what it is. So it's like a dict_keys 36 00:04:15,070 --> 00:04:23,660 object but it behaves like a list. And what you can do then is apply the 37 00:04:23,660 --> 00:04:33,800 get matches method so you pass rain there to data.keys. Let's see 38 00:04:33,800 --> 00:04:42,880 what we get. So we go three matches. We got rain, train and rainy. 39 00:04:43,430 --> 00:04:52,379 You know if you pass an n of 5 you get more matches there, but it's important to 40 00:04:52,379 --> 00:05:01,319 note that the list here is ordered, so the first, the very first word is the one with a 41 00:05:01,319 --> 00:05:08,909 higher ratio of similarity. So we're only interested in that first word. How to get 42 00:05:08,909 --> 00:05:13,710 that? Well, you can ignore this number since you only want the first word and 43 00:05:13,710 --> 00:05:17,669 you can pass the zero there because you know this is a list and the first item 44 00:05:17,669 --> 00:05:24,680 has an index of zero and so you get rain for the rainn mistyped word. 45 00:05:24,680 --> 00:05:30,180 So basically we have the engine to get the most similar word out of a 46 00:05:30,180 --> 00:05:36,300 sequence now and all we have to do is to implement that functionality in our 47 00:05:36,300 --> 00:05:40,979 function in here. So first I would just like to let the user know through a 48 00:05:40,979 --> 00:05:46,830 message they have input a wrong word. So I just want to let the user know and 49 00:05:46,830 --> 00:05:51,360 so that they can run a program again so next time the input a new word for now. 50 00:05:51,360 --> 00:05:55,589 But later we'll do something more intelligent, so for now just think about 51 00:05:55,589 --> 00:06:01,409 this. Give out a message when you mention your suggestion with the correct word 52 00:06:01,409 --> 00:06:06,539 that the user may have be, may have had in mind. So let's do that in the next 53 00:06:06,539 --> 00:06:07,000 lecture, see you.