The original sentiment labels on the datasets (top left) vs. the generated labels using TriLex corresponding to the original label categories (top right).
figure
posted on 2025-04-17, 17:36 authored by Abdulrahman Alharbi, Rafaa Aljurbua, Shelly Gupta, Zoran ObradovicThe bottom figure represents the sentiment distribution using TriLex considering three classes.
History
Usage metrics
Categories
Keywords
short texts posesonline customer reviewsapply weighted averaging2 %&# 8211novel unsupervised approachgenerally short naturedynamic threshold derivedimprove sentiment labelingweak label basedsocial media platformspaper proposes trilexunsupervised sentiment analysisdynamic naturesentiment analysisweak labelsnew labelfusion approachsentiment predictionsentiment expressionxlink ">wide rangeuser sentimentstextual contentstrong labelsresults demonstraterecent yearsmultiple lexiconmajority votesmajority votelstm modelslogistic regressionlabeled dataf1 scoreexisting fusionefficiency issuedaily livescontext limitationschallenge duebenchmark datasets8 %.
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC