Finding pointers of discursively constructed news values in Danish journalism using computer-assisted methods

Within journalism, news values are commonly treated as a value system existing in the minds<br/>of journalists used for making judgements about the perceived newsworthiness of an event [1].<br/>As an alternative, discursive news values analysis (DNVA) analyzes how events are presented as<br/>newsworthy in discourse using various semiotic resources including language and images etc.<br/>[2]. Reserving news values for values in news actors and events, nine news values have been<br/>suggested: consonance, eliteness, impact, negativity, personalisation, proximity, superlativeness,<br/>timeliness, and unexpectedness [1]. Initially, DNVA was introduced as a discourse analytic<br/>perspective that could, for example, be an additional tool for (Multimodal) Critical Discourse<br/>Analysis [2], but DNVA has also been conducted by combining qualitative and quantitative<br/>methods. Using corpus analytic techniques and manual computer-aided annotation, pointers<br/>to each specific news value have been identified [1], [3], [4], in all cases stressing the need for<br/>combining qualitative and quantitative methods because of the role contextual factors play<br/>for conveying news values. With few exceptions [5], [6], [7], [8] on Chinese; [9] on Italian;<br/>[10], [11] on Iranian, DNVA has predominantly been applied to English material using corpus<br/>linguistic methods.<br/>In this work, we explore the potential for computer assisted DNVA to be applied to Danish<br/>journalistic material. Using techniques such as frequency analysis, cluster analysis, concordance analysis and statistical testing combined with in-depth qualitative analysis as devised by<br/>Javadinejad [4], we aim to find linguistic pointers of news values in Danish. Hypothesizing that<br/>different news values will be constructed in so-called ‘hopeful news’ as compared to stories<br/>that are not labeled as such, we will collect and analyze different data sets, such as a set of news<br/>stories indexed as ‘hopeful news’ and a set of ‘top stories’ published by the same news outlet,<br/>TV2, in the same time period.<br/>Doing this, we aim to present the first results on pointers used to construct specific news<br/>values in Danish and establish whether news values can be used as a distinguishing factor when it comes to news labeled as ‘hopeful’. This research contributes to research on the discursive<br/>manifestation of a trend within journalism, that of ‘constructive news’ where stories focus on solutions often in a positive, encouraging tone of ‘good news’ [12], and we also engage in the theoretical discussion of whether to include positivity as a news value in DNVA [1].<br/>As a next step, the same data sets will be subject to sentiment analysis [13] to explore relationships between the two measures as well as their applicability on ‘hopeful news’ specifically<br/>and top news stories more generally, i.e. stories that are not labelled with a specific topic or<br/>news story type, except from being considered important.<br/>Applicability will be addressed and discussed in terms of how our DNVA investigations could<br/>be incorporated into applications, including journalistic tools and recommender systems. For<br/>the former, integration into journalistic editor’s tools could facilitate, for example, evaluating<br/>how written content adheres to an organization’s DNVA doctrine, and thus help an organization<br/>in upholding its charter of values [14]. For the latter, incorporation of DNVA considerations<br/>within recommender systems could provide potential nuanced notions of similarity, to help<br/>derive recommendations aligned with user DNVA sensibilities, or to look to explicitly make<br/>recommendations that have high positive values (an important consideration for recommender<br/>systems as accountability, for what is getting recommended, is becoming increasingly significant<br/>[15]).<br/>The talk will present and reflect on preliminary results and methodological challenges of<br/>using computer-assisted methods for identifying context-sensitive linguistic systematicities.<br/>This includes the tradeoffs of constructing dictionaries using manual and automated methods<br/>and their generalizability on various domains, genres, and newspaper types (e.g., tabloid vs.<br/>omnibus) [13]. The work will also address obstacles and solutions when dealing with non-English materials, including the application of English-centric tools for Danish material.