or common, and then they are encoded in language; that is, they appear in grammar rules, user manuals, and spellings (Méndez García de Paredes 1999: 111).

4. Study methodology and data

The study of technology-mediated communication requires different methods of research and data collection. The use of large-scale quantitative data, known as big data, is extremely interesting, allowing for downloading of massive amounts of data and observing communicative or linguistic phenomena on internet platforms. De Benito and Estrada Arraz (2018) use phonic and morphosyntactic analysis, crossing it with geolocation, to extract data on variation. While this approach is undoubtedly interesting, it also has some limitations, such as the heterogeneity of collected interactions and of users, data that was not specifically selected, making it extremely difficult to contextualise the phenomena being studied.

Another approach is the compilation of data in a smaller corpus but that has an internal coherence allowing the observation of discursive practice of the Internet. This study follows this approach and has been carried out following the compilation and selection of a corpus of reviews of Spanish hotels chains in the major tourist destinations (Ministry of Commerce and Tourism, Government of Spain, 2019). The aim of this is to focus on the analysis on a corpus of reviews in Spanish, written by native Spanish-speaking reviewers, which is why Spanish-speaking destinations were chosen, ignoring other destinations for the time being.3 Meanwhile, instead of collecting only negative reviews, as was done by Vasquez (2011), reviews posted over a period of time were studied (October 2018-March 2019), obtaining a corpus of 120 reviews and 102 responses. In line with studies on stylistic features, such as those published by Sabater, Turney and Montero (2008) or Mancera Rueda (2016), the following study parameters were selected: (i) greetings and closings, (ii) T-V distinctions, (iii), politeness formulas, and (iv) typographical syntax and spelling. In order to do this, the full corpus was labelled and a manual and automatic analysis was carried out using Sketch Engine, calculating the appearance of the elements studied and the ratio of appearance per review. Parameters were selected not to be exhaustive but to ensure it was possible to observe interactions between review authors and hotels. The working hypothesis is that the corpus presents variation in style between the reviews and the responses, and that these differences are related to the communicative intentions of the participants on the platform. In the sections that follow, we present the results of the analysis, as well as a discussion on the differences found and their relationship with notions of the linguistic norm.

5. Analysis of style traits in reviews and responses

Table 1 shows the total amount of appearances and the ratio of greetings and farewells that appear in the hotel reviews and responses. These constitute forms of opening and closing of interactions that provide relevant information on the social and pragmatic intentions of participants and the relationship they aim to establish, since through them, they identify themselves as ‘social actors’ who play a role in the interaction and that show a certain degree of familiarity or distance with