Towards Computational Ecolinguistics: Natural Language Processing Methods and Resources for Linguistic Analysis in Environmental Contexts

Can linguistics,ecology,andNaturalLanguageProcessing(NLP)beinterwovento address thecurrentecologicalcrisisinameaningfulandsynergisticway?Thisdoctoral thesis aimstoaffirmativelyrespondtothisquestionbyaddressingasignificantgapinthe literature—namely,theabsenceofasystematic,interdisciplinaryanalysisofecological discourse. Adoptinganinter-andtransdisciplinaryapproach,thisworkproposesasetof methodological frameworksthatleverageNLPtoolsincombinationwiththeoretically grounded linguisticinsightstoexamineenvironmentalandclimatechangenarratives.At its core,thethesishighlightstheroleoflanguageandlinguisticstructuresinshaping meaning andframingecologicalissues,whileexploringhowNLPcanenabletheir large-scale,systematicanalysis. The dissertationbeginsbyestablishingthetheoreticalfoundationsandmotivationsof the study,followedbyacriticaloverviewofexistingintersectionsbetweenNLP,linguis- tics, andecology.Itthenintroducestwooriginalcorporadesignedformultilevelanalysis across differenttextgenres,alongsideexperimentalsetupsinvolvingbothtraditional classifiers andlargelanguagemodels. Subsequently,Ipresentanarrative-basedapproachusingtheCharacter-RoleFrame- worktoanalyzehowhumanandnon-humanentitiesarediscursivelyframedinenvi- ronmental texts.Thisisfollowedbyamultilingualecocriticalanalysisgroundedin lexicalandsyntacticalignmenttechniques,whichexamineshowecologicalconcepts are constructedacrossEnglish,Italian,andGerman.Finally,thethesisproposesa novelframeworkforevaluatinganthropocentricbiasinlargelanguagemodeloutputs cross-linguistically,providingafoundationformoreecologicallyinformedandcritically engagedNLPresearch.