Media bias through collocations: a corpus-based study of Egyptian and Ethiopian news coverage of the Grand Ethiopian Renaissance Dam

The study investigates how Egyptian and Ethiopian online English-language newspapers linguistically represent the Grand Ethiopian Renaissance Dam (GERD). It situates news discourse as a powerful, non-neutral medium that shapes public understanding through lexical choices. Against the backdrop of long-standing Nile water-sharing tensions—Egypt invoking historical agreements from 1929 and 1959 and Ethiopia advocating equitable allocation—the paper argues that linguistic and ideological analyses of GERD media coverage are underexplored. The purpose is to compare lexical choices and detect bias by identifying highly salient keywords and their collocational patterns, interpreted through Critical Discourse Analysis (CDA) and corpus linguistics (CL). Significance stems from adopting a critical linguistic perspective, combining CDA with CL, and analyzing an extended period (2013–2020). The study addresses two research questions: (1) What lexical items collocate with the highest statistically salient GERD-related keywords in Egyptian and Ethiopian newspapers? (2) What do these collocations reveal about the stances of the two countries regarding GERD?

Prior media studies on GERD have focused largely on framing and proximity effects rather than language-centered analyses. Bealy (2014) identified frames in Ethiopian coverage (development, national image, right, victimhood, mutual benefit, war). El-Tawil (2018) found dominant conflict/problem frames and a theme of denial in Egyptian coverage. El Damanhoury (2023) showed geopolitical proximity influenced international outlets’ framing. Sayed El Ahl et al. (2023) reported national-interest-aligned framing across Egyptian, Ethiopian, and Sudanese outlets. Corpus-based linguistic studies are fewer: Elieba (2022) compared lexico-syntactic features across countries but did not deploy CDA; Elsoufy and Ibrahim (2022) examined themes/semantic domains without collocational analysis. Other GERD discourse studies (e.g., El Shazly & El Falaky, 2023; Siraw, 2023) analyzed official statements and speeches but not lexical choices in depth. This paper fills a gap by integrating CDA with CL to examine keyword collocations and their ideological implications over a longer timeframe with balanced corpora.

Design: A corpus-assisted CDA approach combining Fairclough’s three-dimensional model (text, discourse practice, social practice) and van Dijk’s ideological square with corpus linguistic tools for keyword and collocation analysis. Data collection: Two comparable online English-language news corpora (2013–2020) were compiled from six outlets (Egypt: Ahram Online; Egypt Independent; Daily News Egypt. Ethiopia: The Ethiopian Herald; Reporter Ethiopia; Walta Information). Articles were retrieved via Factiva using the query “Grand Ethiopian Renaissance Dam”, manually screened to retain hard news reports only (excluding editorials/op-eds), cleaned of metadata, and saved as plain text. The final dataset comprised 2655 articles and over 1,000,000 words, with each national corpus containing over 500,000 words. Software and statistics: Wordsmith Tools (v8) was used to extract keywords; Sketch Engine was used to produce word sketches and collocations. Keyness metrics followed Gabrielatos (2018): (a) Log-likelihood (LL; p<0.000001) to determine statistical significance; (b) Bayes Information Criterion (BIC) to strengthen evidence against the null hypothesis; (c) Log Ratio (Hardie, 2014) as the primary effect-size metric for sorting keywords. Equal-sized corpora were used to stabilize significance measures. Collocational salience was measured with LogDice (effect-size, corpus-size independent). Colligations (grammatical relations) and semantic preference/prosody were analyzed. Procedure: (1) Extract top 50 content keywords per corpus (excluding style words); (2) select five statistically salient, thematically relevant keywords per corpus; (3) identify collocates and grammatical patterns (subjects/objects/modifiers/conjunctions); (4) interpret semantic preferences and prosodies; (5) connect lexical patterns to ideological strategies (victimization, dramatization, attribution of agency, polarized description/self-glorification) and to broader socio-historical contexts.

- Corpus scope: 2655 hard news articles (2013–2020) totaling over 1M words; two balanced corpora (>500k words each). - Egyptian corpus top salient keywords: crisis, fears, affect, negotiations, interests. Patterns indicate a threat/victimization framing of GERD. • crisis: frequent modifiers (economic, political, internal, acute, ongoing; dam/GERD/Nile); verbs resolve (LogDice 11.31), solve (10.8), end (9.7), escalate (9.0). Links GERD to a multifaceted national crisis/conflict. • fears: strong collocates overpopulation (12.9), voice (12.5), raise (10.9), dispel (10.7). Prosody of panic/dramatization; fears are repeatedly voiced and stoked. • negotiations: collocates include resume (11.9), continue (11.2), stall (10.3), suspend (8.6); as subject: fail (10.0), falter (8.5). Prosody of failure, often attributing blame to Ethiopia. • affect: modifiers negatively (13.2), adversely (10.5), severely (8.4); objects share (12.9), supply (10.2), flow (10.0), quota (8.8). Grammatical pattern casts GERD/dam as agent affecting Egypt’s water/security. • interests: modifiers common (12.5), mutual (11.3), national/strategic; verbs harm (11.7), safeguard (11.4), protect (10.8). Mix of threat to Egypt’s interests and appeals to shared/common interests. - Ethiopian corpus top salient keywords: utilization, growth, nation, power, negotiation. Patterns emphasize development, rights, and national self-glorification. • power: modifiers electric (12.3), hydroelectric (10.1); verbs generate (12.4), export (10.2), supply (10.1), produce (10.0). Positive prosody linking GERD to energy and economic benefits. • utilization: modifiers equitable (12.4), reasonable (12.1), fair (12.0), sustainable (7.9), Nile (7.0); verbs ensure (11.1), share (9.6). Emphasizes equitable/reasonable use; “unfair” marks negative view of prior allocations. • growth: modifiers economic (11.9), rapid/fast/double-digit; verbs register (11.5), stimulate (10.4), sustain (9.5). Positive development framing. • nation: modifiers African (10.5), populous (8.8), sovereign (7.9); verbs enable (10.0), build (as subject 9.7). National self-glorification and legitimacy claims. • negotiation: modifiers tripartite/trilateral/AU-led/peaceful; verbs resume (11.1), continue (10.9), mediate (9.2), facilitate (8.5); as subject: fail (9.7). Frames Ethiopia as pursuing diplomacy; negotiation failures attributed to Egypt’s stance. - Cross-corpus: Both reference negotiations, but differ in attribution of failure and framing: Egyptian coverage stresses crisis, fear, and negative effects; Ethiopian coverage emphasizes development, equitable utilization, and national agency/rights. These patterns align with van Dijk’s ideological square (positive in-group/negative out-group) and CDA findings of polarized media bias.

The findings directly address the research questions by demonstrating that highly salient keywords and their collocational environments systematically differ across Egyptian and Ethiopian media. Egyptian outlets predominantly embed GERD within a crisis/ threat narrative (victimization and dramatization), casting Ethiopia as the agent responsible for failed negotiations and harmful outcomes to Egypt’s water security and interests. Ethiopian outlets foreground development, equitable utilization, and national progress (self-glorification and positive in-group description), often portraying Egypt as monopolizing the Nile and obstructing fair resource use. These lexical choices function as discursive strategies of polarization consistent with van Dijk’s ideological square (attribution of agency to the out-group for negative acts; detailed positive descriptions of the in-group). The synergy of CDA with corpus methods shows how recurrent collocations and prosodies scaffold ideological positions beyond isolated examples, contributing to media discourse studies by revealing systematic, statistically grounded patterns of bias.

The study shows that Egyptian and Ethiopian English-language news about GERD exhibit polarized lexical patterns: Egyptian reports concentrate on crisis, fears, negative impacts, and threatened interests, while Ethiopian reports accentuate power generation, growth, equitable utilization, and national capability. Both invoke negotiations but diverge in attributing responsibility for impasses. By combining CDA with corpus tools, the paper uncovers how collocations and semantic prosodies sustain ideological stances aligned with national interests. This work contributes to the literature by shifting from frame inventories to lexical-collocational evidence across an extended period and balanced corpora. Future research could expand multilingual scope (including Arabic), explore additional media genres (editorials, opinion, broadcast), and analyze newsroom practices and news values shaping objectivity versus bias.

- Genre scope: Focused on hard news; editorials/opinions and other media forms (e.g., TV talk shows) were not analyzed. - Language scope: Egyptian data limited to English; comparison with Arabic-language Egyptian coverage and other languages is needed. - Media-process analysis: Did not deeply examine newsroom standards, news values, or production practices affecting representation. - Framing breadth: While lexical-collocational analysis was in-depth, broader multimodal and narrative dimensions were beyond scope.