Figure 4
The multi-panel visualisation presents bar charts of terms predicting different brand metrics with Feature on the y-axis and Coefficient on the x-axis, including Terms predicting brand defence with positive coefficients for terms such as e a, dice, best, great, devs, game, better, b f s, frostbite, and community, and negative coefficients for firearm, published, mistake, tired, awful, ok, care, l o l, haha, and deleted, Terms predicting brand positivity with positive terms like love, amazing, best, great, blast, favourite, fun, awesome, masterpiece, and perfect, and negative terms including fuck, trash, wont, worse, doesnt, dont, awful, wasnt, deleted, and worst, Terms predicting brand appraisal with positive terms such as weapon, vehicle, felt, gun, b f s, map, class, mechanic, compared, and different, and negative terms like dumb, saw, dead, dogshit, deleted, l m a o, thanks, i d k, post, and l o l, Terms predicting brand zest with positive terms best, love, amazing, awesome, favourite, blast, great, masterpiece, fun, and incredible, and negative terms think, isnt, people, wasnt, wouldnt, didnt, deleted, worst, i d k, and dont, Terms predicting virtual positive expression with positive terms love, amazing, awesome, best, epic, blast, absolutely, wow, gorgeous, and hell, and negative terms war, problem, old, didnt, bad, pretty, deleted, people, better, and dont, and Terms predicting brand knowledge with positive terms sale, vehicle, class, weapon, destruction, e a, map, different, variety, and premium, and negative terms worst, dead, anymore, deleted, i d k, hell, hate, wasnt, shit, and dont, collectively illustrating how language features influence different brand perception dimensions.Shapley Additive Explanations analysis for Battlefield (unigram)
Source: Authors’ own work