Sports journalism, artificial intelligence, and quality

Despite its relevance in contemporary media, sports journalism has not been characterized as being an object of analysis, far from the treatment that other areas of the media industry usually receive (English, 2016).

At a technological level, sports journalism has known how to innovate, incorporating advances such as data journalism into its chronicles and retransmissions (Rojas, 2014; Horky and Pelka, 2017) or the inclusion of 360º video to the usual narrative of journalistic coverage. (Martínez and Torrado, 2017, p. 154). Not surprisingly, as Boyle and Haynes (2009) point out, one aspect that has traditionally characterized the culture of sports fans has been the constant need for speed in the transmission of information, something that the incorporation of these technologies favors (p. 183). This need for sports- related content has not only involved the media but also sports institutions, which have also deployed different strategies on social networks to strengthen their bond with fans (Thompson et al., 2017).

In 2010, artificial intelligence projects linked to the production of sports news began to be developed (Bunz, 2010). In the Spanish sphere, it has been incorporated into the newsrooms for documentation or the preparation of chronicles, such as those carried out by BeSoccer (Segarra-Saavedra, Cristòfol, and Martínez-Sala, 2019) or AnaFut in El Confidencial (Rojas and Toural, 2019). In both cases, the opportunity that the use of this technology represents for the future of journalism is highlighted, in an area in which data analysis is so standardized.

Along the same lines, Galily (2018) defends that artificial intelligence applied to sports journalism may involve the incorporation of a greater number of professionals, destined for the generation, collection, and management of data to feed the generation processes of artificial intelligence. Technological innovations also imply the alteration of the traditional role of the sports journalist, with the professional taking on the role of moderator rather than mediator (Perreault and Bell, 2020) or focusing on the search for journalistic stories about the competition or its protagonists (Rojas, 2019).

Traditionally, sports journalism has been accompanied by a perception of a lack of seriousness in contrast to that of other types of fields (English, 2017). In research carried out in Chile, Scherman and Mellado (2019) detected that sports journalism uses fewer sources of information, integrates fewer points of view, and presents less verifiable information to the public. Similar results were found in the research carried out in Spain and Mexico by Márquez-Ramírez and Rojas (2017) on the informative treatment of one of the biggest corruption cases in world soccer, the so-called FIFA Gate. In the Spanish case, this research found that the level of proactivity of El País doubled that of the newspaper Marca. However, changes in professional practices are being seen throughout the world (Weedon et al., 2016; Goikoetxea and Ramírez, 2020), which could result in the search for models more focused on quality.

Among the characteristics that sports language brings together is passion (Naranjo de Arcos, 2011, p. 243; Loaiza, 2018, p. 221) or the habitual use of warlike or epic phrases, especially visible in some genres, such as the case of the journalistic chronicles (Arroyo and García, 2012; Martínez, 2018), where from an early moment many of the usual techniques in the so-called new journalism were tried out (Sánchez and Armañanzas, 2009).

The chronicle is usually understood as one of the most relevant genres of sports journalism, coming to be described as its “queen” (Sobrados, 2009, p. 82), with an outstanding role in sports such as soccer. A genre that stands out for its richness, which is prolific in the use of literary resources (Dauncey and Cooke, 2020) and rhetorical figures, such as metaphors (Kovljanin, 2018), anaphora (Quintero, 2015), hyperboles (Quintero and Hernández, 2019), or personifications (López, 2019).

It is a creative story where the author’s style is appreciated (Sobrados, 2009, p. 90) and which drinks from the hybridization of current journalism, in which the chronicler, beyond informing, aims to delight the reader (Román, 2015). A genre considered an “escape” by Marín (2000), which gives “free rein, without exaggeration, to the cultural and literary virtues of sports journalism.” The nature of the chronicle also makes it possible to stimulate subjectivity and attachment to certain values as a mechanism for the cohesion of readers and hobbies (Naranjo de Arcos, 2011, p. 347).

Although the academy has addressed these issues from various fields, there has been no proliferation of works that simultaneously address sports journalism, artificial intelligence, and quality, so this research aims to provide knowledge on the subject.

Use of sources

In the selected chronicles, both those made by algorithms and those written by humans, no documentary sources appear. The immediacy of digital media encourages chronicles to be published imminently and there is no time for meditation to consult other types of sources, archives, or sources of a match, such as the referee’s minutes, to delve deeper into some causes of the expulsion of players and other reasons.

The use of personal sources in the selected chronicles prepared with artificial intelligence or in those written by journalists is not usual. In this section, only one personal source appears in the Cádiz- Valencia match, and it is related to a precedent (ID-15). In this case, the only text that has this type of font has human intervention and retrieves sports news from throughout the week, something that artificial intelligence does not use in its texts.

Narrative structure

The narrative structure is significant if the data is crossed with the type of writing of the chronicle (AI/ Human). The use of statistics means that the artificial intelligence of Narrativa has a mixed distribution of the text in 100% of its writings, where important elements appear at the beginning of the text (inverted pyramid), but later it has paragraphs that are chronologically described. Furthermore, it concludes most of them with important data such as warnings and the general classification of La Liga.

On the other hand, the type of structure changes substantially when the text is written by an editor. Of the seven texts corresponding to the first division, five of them are written chronologically, 1 uses the inverted pyramid, and one has a mixed organization. On the other hand, 100% of the second division chronicles are written chronologically. Therefore, considering the global corpus of texts written using the human factor, 85.71% are written chronologically, 7.14% use the inverted pyramid, and the other 7.14% present a mixed narrative.

Style

The style of the journalist is what will determine the greatest differences between the texts written by artificial intelligence and those that are thought by a journalist. The evaluative adjectives, the attribution verbs (in the texts in which personal sources appear), and the rhetorical figures make up an interpretive addition that enriches the text and that is a common feature in the pieces written by humans.

The quantification of evaluative adjectives is an important element of interpretation in sports chronicles. If all the adjectives of the pieces elaborated by algorithms are added, the number of eight is reached, 4.47% of the adjectives counted, and an average of 0.57% per piece. A value well below the adjectives used in journalists’ writings where the sum shoots up to 171, which represents 95.53% of the whole, and an average of 12.21 per publication.

It is also significant that the change of category supposes a substantial increase in the adjectives made by journalists. While in the first category 101 are counted, in the second it is reduced to 70. The average number of evaluative adjectives in chronicles of the highest category is 14.42 and represents 59.06% of the attributions of this type of writing. However, the evaluative presence in the silver division is reduced with an average of ten adjectives per publication of journalists and 40.94% of the use of adjectives. The length of the chronicle, the greater or lesser degree of bias, and proximity (local or national environment) influence when using more assessment in the text. Some of the adjectives that can be read throughout the texts prepared by editors are “lethal” (ID-22), “unleashed” (ID-23), “debatable” (ID-24), “bittersweet” (ID -20) “shameless” (ID-16), “disastrous” (ID-19), and “sharp” (ID-28), among others. On the other hand, in those made by algorithms, the one that is repeated the most is “positive” (ID-13), and the adjectives “correct” and “disputed” also appear (ID-10).

Attribution verbs are directly related to personal sources. In this selection no significant attribution verbs appear, only in the Valencia-Cádiz match can one of them be read. Specifically, “announced” (ID-15) is the verb used for the use of a personal source. In this case, the text is written by a journalist. It is a resource that is not used too much in the selection of the global corpus, but that is indicative of the use of knowledge and monitoring of the topic.

The use of literary resources is an indicator that serves to differentiate between the human factor and artificial intelligence. When it comes to quantifying the numerical quantity of the sample: little (0-5), average (6-10), and a lot (>11), it can be seen that the texts written by humans predominantly include figures of speech. In this sense, no text elaborated by algorithms exceeds the three uses of rhetorical figures. If the total number of rhetorical figures that appear in the 14 texts written with artificial intelligence is added, nine are counted and the average per text would be 0.64, therefore, its interval of use is between zero and five (little). On the other hand, the texts written by humans incorporate 193 figures of speech, so the average is 13.14 resources in writing and the use would be placed in the range of >11 (a lot). Regarding the global corpus, the number of rhetorical figures used in texts produced by artificial intelligence only represents 4.66%, so those written by journalists represent the remaining 95.34%.

If broken down by categories, 131 rhetorical figures of the texts written by journalists correspond to the division of honor of Spanish football, with an average of 18.71 (a lot) per text and they appear in 71.20% of texts of this nature. In those related to the lower category, the number of figures of speech used is limited to 53, which means an average of 7.57 per text (6-10, average use) and 28.80% of the total number of used literary resources. Some examples of rhetorical figures that appear in texts written by humans are: anaphora such as “and chew and chew” (ID-18), personifications such as “Granada CF is already smiling” (ID-16), hyperboles such as “the stadium and the celebration of the pichichi exploded” (ID-19), similes like “Pedrosa who arrived like a motorcycle” (ID-19), metaphors like “they turned the grass of the Metropolitano into a labyrinth of mirrors” (ID-18), and metonyms like “they were locked with a yellow to Merino” (ID-20), among other resources and examples.

Word frequency

The style of the journalist is what will determine the greatest differences between the texts written by artificial intelligence and those that are thought by a journalist. The evaluative adjectives, the attribution verbs (in the texts in which personal sources appear), and the rhetorical figures make up an interpretive addition that enriches the text and that is a common feature in the pieces written by humans.

When establishing a word frequency, those terms that are empty of content or that, in this context, do not add value to the study, such as conjunctions or links, nationalities and names of football players, cities, and football stadiums, have been ruled out, among others. It must be taken into account that the same chronicles have been analyzed, some written by artificial intelligence (14) and others by journalists (14). Words with three or fewer letters have also been discarded.

The word frequency of the global corpus of texts written by artificial intelligence shows that many of its concepts are used very repeatedly in the journalistic texts created by the Narrativa software. This is the case of the terms “team/s”, which is repeated 61 times and occupies 1.40% of all the words used in these journalistic pieces; “match/es”, with a repetition of 53 times and 1.21% frequency; point/s, with 49 uses and 1.12%; and “part”, with 48 repetitions and 1.10%, which is the last word that exceeds 1% presence in journalistic texts written using artificial intelligence. Therefore, the greater use of some words prevails over others and there is no balance in the use of terms.

Source: Own elaboration with the Nvivo 11 tool.

Quite the opposite happens with texts written through human labor. Although there are words that are the most used due to the very need to narrate a football match, their frequency is much more widely distributed. The two words that appear most often in the texts written by journalists, “minute/s” and “first/s”, are used in 0.45% of the 14 sports chronicles, on a total of 40 occasions. It is followed by “ball” and “match/s”, both with 36 uses and occupying 0.40% of the corpus.

Source: Own elaboration with the Nvivo 11 tool.

The words that coincide in the list of the 20 most used are: “team/s”, “match/es”, “point/s”, “part”, “minute/s”, “local/s”, and “arrived”. Therefore, of the 20 most frequent words of each of the writing methods, only these seven are identical. On the other hand, the most repeated terms that appear in the texts written with artificial intelligence and that do not fall within the most used by the pieces written by journalists are: “result”, “marker”, “against”, “visitor”, “set”, “three”, “stadium”, “win/s”, “match”, “front”, “card/s”, “last/s”, and “home”. In the case of the words that appear exclusively in texts with authorship, they are: “first/s”, “ball”, “game”, “area”, “little”, “centre/s”, “out /s”, “new”, “first”, “yellow/s”, “second”, “final/s”, and “had”.

Table 3 : The 20 most used words in texts written with artificial intelligence and by journalists. Own elaboration.

Source: Own elaboration.

In this list, a metonymy can be seen in the texts written by journalists where they use “yellow” without specifying that it refers to “card/s”. The term “yellow” appears among the 20 most used, while “card/s” is among the most repeated. However, in the writing of journalistic pieces with artificial intelligence, the opposite happens and “card/s” is the most frequent term and “yellow” does not appear among the most used.

Word frequency can also be broken down by division (first with seven texts and second with ano- ther seven) and by elaboration (artificial intelligence and human intervention). The five most repeated words in first division texts written by journalists are: “ball” (24, 0.46%), “minute/s” (24, 0.46%), “match/es” (23, 0.44%), “team/s” (16, 031%), and “arrived” (15, 0.29%). On the other hand, if the first division journalistic pieces are created by artificial intelligence, the frequency would be as follows: “team/s” (32, 1.69%), “point/s” (26, 1.37%), “match/es” (24, 1.27%), “part” (21, 1.11%), and “marker” (19, 1%).

The frequency of words in the silver category of Spanish football also has its particularities in the use of terms. In this way, the most repeated terms in the second division texts elaborated by editors are: “game” (18, 0.48%), “minute/s” (16, 0.42%), “team/s” (13, 0.34%), “match/es” (13, 0.34%), and “ball” (12, 0.32%). For the second division texts made using the Narrativa software, the terms with the most uses are: “team/s” (29, 1.17%), “match/es” (29, 1.17%), “part” (27, 1.09%), “point/s” (23, 0.93%), and “result” (17, 0.69%).

The texts that were written for the first division and second division, but with their own authorship, coincide in 4 of the five most used words and differ only in one term, “arrived” for the gold category and “game” for the second. This same analysis carried out in writings elaborated with artificial intelligence has a coincidence of four of the five most frequent words. It only varies in one word in the first division (“marker”) and another in the second division (“result”).

There is one more difference if you compare the first division texts written by human intervention and artificial intelligence. Of the five most frequent words, three coincide and differ in two: “ball” and “arrived” for texts with authorship and “point/s” and “marker” for those pieces created by algorithms. For its part, for the second division, the coincidence is only two of the five words. Therefore, there are three words with more usability depending on the procedure when writing the information: “game”, “minute/s”, and “ball” in the texts written by journalists, and “part”, “point/s”, and “result” in the pieces created by the software.

Content

Headline

Artificial intelligence works with data and can be seen in the headlines of its journalistic texts. All its writings use titles of an informative type, with a simple structure and with the result of the encounter as a common denominator (in most of them). Instead, headline types vary if they are written by journalists

and depending on the category. In the first division, four would be informative, two appellatives, and one

expressive. On the other hand, in the second division, six of them are informative and one is appellative.

Presence of different points of view and bias

Regarding the presence of one or more points of view, there is complete homogeneity in the results obtained and all the texts selected from the corpus only have a single point of view.

Bias is a common feature in sports chronicles. While in the texts elaborated with artificial intelligence the bias is not appreciated, in those that are written by humans this characteristic predominates. Of the seven matches, six of them do have a bias, while the Mallorca-Levante match, published on the website Ultimahora.es, is written objectively towards the two teams. In this sense, in the second division, the trend is repeated in the writings created by journalists, there are six that do have it, and one of them that does not have this feature, specifically the Valladolid-Alcorcón match, published in the online version of the newspaper Marca.

Understanding

Another of the points analyzed is the comprehension of the texts. All the writings are understandable, both those that are made by artificial intelligence and those that are written by the journalist. One of the main characteristics of journalism is clarity and this element is fulfilled in all the selected pieces.

Emphasis

The emphasis that the journalist brings to the sports chronicles serves to determine the intervention of the editor in the text. In this case, 100% of the pieces produced by algorithms have a factual emphasis, that is, they are based on facts. On the other hand, the emphasis is on opinion in 12 of the human- written texts, while another emphasizes opinion and speculation together. The remaining piece presents a factual emphasis.