 So first thing I would like to say is thank you very much, as Alessandra was saying this at the end, I would like to start with this. Thank you very much for organizing the Coda Day and for inviting me. And yeah, it's a great opportunity to contribute to this event. So I'm going to talk about compositional analysis of tourism related content as the majority of you would know if you have seen the abstract and my bio. I have been in the tourism field from the very beginning of my career, but I started using compositional data analysis less than 10 years ago. In 2012 I did the Coda course, the week Coda course, and it was quite difficult at the beginning to publish or to start in publishing in tourism field journals, because at the beginning it was like journals or the researchers in the field were a bit afraid of this new methodology. And then while we manage, yeah, and today I can talk about tourism related content and how I applied the compositional data analysis I mean the methodology in the tourism field I use the methods, and I apply it to tourism, to the tourism field. I divided the presentation in three main applications, which are different studies and different publications. I have used in the first application, the distance in the second application, the by plots and in the last one, the man over so. Yeah, I'm using the different Coda tools to respond to different objectives. Everything I mean all these studies are related to analyzing content, okay, posted on social media, online platforms as it is the case of Airbnb or TripAdvisor or Facebook. And with some of the colleagues here in the University of Lleida with Estela Mariné and Eva Martín Fuentes. We started using compositional data analysis to complement the established methods to analyze content, to complement text mining and the expert Estela Mariné Roche is one of the experts in tourism in the tourism field to analyze the tourist destination. And we found an opportunity all together to use compositional data analysis to analyze content posted in this case on the internet online. The first application is that of and I mean that of published in Annals of Tourism Research, which is one of the top journals is First Decide, and we analyze here the destination image. With Airbnb reviews in the second application, we wanted to analyze the the complaints of hotels in Barcelona. In this case, we selected I'm going to give more details and just to introduce him a little bit. In this case, we wanted to analyze in for those hotels with evaluation of one or two out of five, five is is excellent and one is horrible. So those hotels with evaluation of one or two, what the users complained about. And finally, the third application is based on the content posted by hotels again, but in this case on the social media Facebook. And we wanted to see differences between the category, the, the auto class. We are with the first application with the first study. And we wanted to compare the tourist destination image, which is something that one of the colleagues worked a lot of for Spanish cities, Barcelona, Valencia, Madrid, and Sevilla. And we wanted considering that Airbnb is one of the most used peer to peer platforms accommodation platforms, and the experience that tourists that we when traveling. The experience we have when traveling in the accommodation really determines the satisfaction of the whole trip. Okay, so the idea is how tourists perceive the destination what is the image of the destination that tourists. When when posting a review on the Airbnb platform. Okay. And the data we use in this case are reviews Airbnb reviews so opinions of users in this for cities between 2010 and 2018. And we considered all reviews written in English of those listings accommodation listings with at least a minimum of number of reviews, which was 10. So, in the end, we came up with a lot of reviews. We can see here more than 300,000 reviews in Barcelona, 260,000 in Madrid, almost 130,000 in Sevilla, and almost 89,000 in in Valencia. And I would like to point something relevant here, which is, well the number of reviews for each city is not relevant. I mean, this is this not bias our analysis because in the end, considering all reviews written in English from 2010 to 2018. It's true that in 2010, and the number of reviews was really really small. But since we are going to process the text of these reviews, and we are going to or we classified categorized this text within eight destination image categories that the number of reviews was not important because in the end we had eight destination image categories, and we classified the main keywords which were extracted from the content analysis techniques. Yeah, we had percentages. So, for sure, we have proportions. Okay. So, yeah, this tourism destination image categories were already defined in a previous work by Marina Rods and Anton Clave. So, this was based on theory. In this case, we replaced very, very few count zeros. We consider we have count zeros because we are counting how much time, how much times a keyword appears in the image categories. Here we computed the CLR transformation to afterwards compute the distance between the pairs of cities. So we wanted to measure how different, how big or small, the difference between the perception of tourists of these four cities. So, if in Barcelona, tourists talk more about sports than in Madrid or in Sevilla, there is more content about the sea, for example, etc. So we had these pairs of, I mean, distances between pairs of cities and we had these six pairs of cities. So, here we have the results and we can see in this first part of the table, the percentages per category. Here we have the eight content destination image categories, foot and wine, intangible heritage, leisure and recreation, sports, tangible heritage, urban environment, etc. And we have out of the total key keywords extracted from the content techniques, which were classified, we have that in Sevilla proportionally, there is more content in the Airbnb reviews about foot and wine. And also there is more content about intangible heritage. In Barcelona, it's notably, this is related to the Olympic Games, because there is a lot of reviews talking about sports in Barcelona, and this difference between, I mean, this percentage is really high compared to the others. And also in Valencia, we have more keywords classified into sun, sea and sun, etc. Here we have, for example, in Madrid, there is more about urban environment. So, in this second part of the table, we can see the contributions to the itison distance for each of the pairs, and we are quantifying the difference, so we can see that the biggest difference is between Barcelona and Sevilla. So, tourists, posting or writing reviews on Airbnb, perceive these two cities as really different in terms of these image categories. And this category intangible heritage is the one which contributes more to distinguish, to differentiate between these cities. And also sports, which contributes a lot. And also we can talk about sun, sea and sun, which contributes to differentiate Sevilla and Valencia. And the smallest gap is between Barcelona and Madrid. So, tourists visiting Spain and visiting these two cities, they perceive Barcelona and Madrid as quite similar destinations in terms of this already defined tourism destination categories. When I do this presentation in non-coder audiences, I also like to show this table, which is the proportion subtracting, I mean are the results of proportion subtracting directly without considering the proportionality between, I mean the proportionality within the tourism destination categories. So, more or less, we can see that here we have a big difference, here we have another big difference, which was more or less the same case with the contributions and considering the proportionality. But in the end, yeah, in this table I'm not able to interpret it because I don't know how to do it. Because, for example, there is five times more of content, sun, sea and sun in Valencia than in Sevilla, or we have four times more or whatever. I mean, we are not considering the proportionality, right? And I like to show how to compute differences when we have proportions or percentages. So, this is the first application. Let's go to the second application. And in this case, well, we wanted to analyze the auto-reviews content and in TripAdvisor, I mean based on reviews on TripAdvisor, and more particularly to relate complaint topics to one another as well as to distinguish auto clusters based on major complaint topics. So, in this case, we are talking about complaints and we selected those hotels with a lower valuation on TripAdvisor. And we selected, we had a random selection of three, sorry, 31,000 auto-reviews of the city of Barcelona, which were downloaded, well, five years ago. And, well, we, to have a representative number of hotels and reviews which are auto-active in the economic terms or in the tourism activity terms, we selected those hotels which had at least 150 reviews. And finally, we got 50 hotels. And then, out of all of these reviews, we randomly selected 50 reviews of each 50 hotels. So, in the end, we had 2,500 reviews which were analyzed. And we did, in this case, we did the classification of reviews and counting, I mean, within the review, we wanted to count the topic of the complaint. So, this was done manually in this case. Okay, and we were to people and a student helping us to do this task. So, based on the contents of the reviews, we defined seven topics, which were these topics were cleanliness, were staff, services, environment, the facilities of the hotel, services offered by the hotel and other topics. In this case, we had some zeroes because first we defined actually eight topics, eight complaint topics, but then we realized that there were two topics which were quite similar. And when classifying the reviews into the seven topics, eight topics at the beginning, we realized that there was a bit of confusion and finally we decided to mix two of them. After doing that, we had very few zeroes and we replaced them as count zeroes. We computed the CLR transformation and finally we had, we just depicted the biplots and computed the cluster analysis. So here we can see that, for example, these autos here received more complaints in proportionality about services here that these ones. We have a number of hotels, which have more complaints about the cleanliness and the facilities, these perhaps more about environments and this one a little bit more of stuff. And here we have the cluster analysis, and we finally got three clusters. The first cluster, the red one, it's a bigger one, and here I included the centers of its topic per cluster. So here we have in the first cluster, we can see that complaints about facilities and complaints about environment are the ones which predominates here in the first cluster. In the second cluster, we have also facilities and then services. So in these hotels, autos belonging to the second cluster receive more complaints about these two topics. And then we have a very small cluster, including all these green dots, autos, and here we have mainly complaints about environment and then other topics. So let's go to the last application. And in this study, we wanted to see the strategy of autos posting content on the social media Facebook. And we wanted to see if autos talk more about the destination where they are located, or autos talk more about the service they offer as individual firms as autos. So comparing a little bit the content posted on Facebook. In this case, we considered those hotels from Barcelona and Madrid again, but those hotels having an active profile on Facebook and an active profile on Facebook, we considered those hotels publishing or posting at least three times per month, which is quite very few posts, but we were searching about how to consider a profile active in social media and finally we came with this. So we considered those hotels with a Facebook account and being active. And out of these hotels, we analyzed the last 25 posts published since a determined date in which we collected the data. And again we're manually classified into destination content categories so we had a composition related to destination or tourism destination content. And we had a second composition which was auto content categories. So let's see the different categories here we had heritage, urban nature and sports and gastronomy. And this was for the I mean this is the composition for tourism destination content and that of auto content we selected or we wanted to see the content about rooms if the hotels were posting content about rooms. About other facilities or about restaurants. Okay. And finally, this six location is here we are comparing the contents of hotels in the numerator versus the contents related to the tourism destination. Yes. Five minutes now. Right. Yes. Okay, actually, I have my clock here. Three minutes actually right. Okay. So, yeah, I'm just finalizing. And I'm going to comment a little bit this table. We computed a man over. And we can see here and the, the first and the second, the fourth and the fifth look ratios were statistically significant. And we can see, for example, here, and that's four and five style tells, talk more about the heritage, because these are positive. And we had heritage. Let me come back. We have heritage and the first look ratio we had heritage over the other contents. So this, I mean, highest class of auto posted content on Facebook related to heritage. So three style tells talk more about urban environment, more than the other classes. When we talk about when it has to do with the content about auto services, we can see here that five style tells talk more about other facilities not drums because yeah rooms it's like, well, it's compulsory because we are talking about hotels but proportionally five style tells post more content on Facebook related to other, I mean, the restaurant and the other facilities like the spa, the gym, the lobby or some meeting rooms, etc. And finally, considering the sixth location that of auto services versus tourism destination, we can, we can see here that five style hotels talk more about the hotel. While three style hotels talk more about the destination so it's it's like, it seems like five style hotels wants the clients to be at the hotel. Why, because they offer services, different services, which are high quality services, while three style hotels post more content about what to do at the destination so they want clients not to be out at the hotel but visiting the city. We did the same analysis for Barcelona and Madrid, and in the end, just to finalize in the end, we can see that hotels from from Barcelona. We can talk more about the, the hotel, while hotels in Madrid talk more about the destination. Okay, so, again, we can, we can see some, some, some differences. And, well, this is all. Thank you very much. I will be happy to discuss any, anything, anything else with with you related to to these analysis I'm sorry for the time perhaps I have been rushing a little bit. Hopefully, everything is clear.