?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=PENDEKATAN+BARU+EKSTRAKSI+INFORMASI%0D%0ABIOMEDICAL+BIG+DATA+REPORT+DENGAN+BIOWORDVEC%0D%0AMENGGUNAKAN+MODEL+HYBRID+LONG+SHORT-TERM%0D%0AMEMORY+%E2%80%93+CONVOLUTIONAL+NEURAL+NETWORK%0D%0A%0D%0A(LSTM+%E2%80%93+CNN)&rft.creator=%09Dian+%2C+Kurniasari%09&rft.subject=500+ilmu+pengetahuan+alam+dan+matematika&rft.description=Peningkatan+angka+kematian+akibat+leukemia+telah+mendorong+pesatnya%0D%0Apertumbuhan+publikasi+mengenai+penyakit+ini.+Lonjakan+publikasi+tersebut%0D%0Aberdampak+signifikan+pada+peningkatan+literatur+biomedis%2C+yang+membuat+ekstraksi%0D%0Ainformasi+relevan+tentang+leukemia+secara+manual+semakin+menantang.+Hal+ini%0D%0Adikarenakan+penelitian+yang+ada+sebelumnya+umumnya+hanya+memperhitungkan%0D%0Akomponen+leksikal+dan+sintaksis+teks+tanpa+mempertimbangkan+makna%0D%0Asemantiknya.%0D%0ATujuan+dilakukannya+penelitian+ini+antara+lain+adalah+untuk+menemukan%0D%0Amodel+Deep+Learning+(DL)+terbaik+dalam+melakukan+ekstraksi+informasi+yang%0D%0Arelevan+secara+semantik+pada+sejumlah+besar+data+biomedis+yang+disebut%0D%0Abiomedical+big+data+report.+Semantic+Text+Similarity+(STS)+adalah+salah+satu+bidang%0D%0Apenelitian+penting+dalam+aplikasi+saat+ini+yang+terkait+dengan+analisis+semantik+teks.%0D%0AMetode+tersebut+memungkinkan+ekstraksi+informasi+dari+suatu+teks+menjadi+lebih%0D%0Abermakna+karena+melibatkan+penerapan+representasi+distribusi+kata-kata+atau%0D%0Asumber+eksternal+pengetahuan+semantik+terstruktur+seperti+word+embedding.%0D%0ANamun+perlu+diperhatikan+bahwa+word+embedding+yang+digunakan+harus+sesuai%0D%0Adengan+domain+penelitian.%0D%0APenelitian+ini+mengusulkan+penerapan+arsitektur+Siamese+Manhattan+pada%0D%0A%0D%0Amodel+DL%2C+yaitu+model+CNN%2C+LSTM%2C+hybrid+CNN-LSTM%2C+dan+hybrid+LSTM-%0D%0ACNN%2C+untuk+melakukan+analisis+semantik+teks+biomedis.+Teks+biomedis+yang%0D%0A%0D%0Amemiliki+makna+semantik+atau+berada+pada+konteks+yang+sama+direpresentasikan%0D%0Ake+dalam+bentuk+vektor+berdasarkan+word+embedding+khusus+domain+biomedis%2C%0D%0Ayaitu+BioWordVec.+Lebih+lanjut%2C+model+tersebut+dibangun+dan+dibandingkan%0D%0Aberdasarkan+jumlah+lapisan+tersembunyi+dan+metode+pelabelan+yang+digunakan.%0D%0AJumlah+lapisan+tersembunyi+yang+digunakan+adalah+dua+dan+tiga%2C+sedangkan+metode%0D%0Apelabelan+yang+digunakan+adalah+metode+Cosine+Similarity+(CS)+dan+metode+Word%0D%0AMover%E2%80%99s+Distance+(WMD).+Hasil+analisis+semantik+menunjukkan+bahwa+setiap%0D%0Akalimat+memiliki+makna+semantik+yang+identik+dengan+tingkat+similarity+1.%0D%0AHasil+tersebut+selanjutnya+menjadi+landasan+untuk+dilakukan+klasifikasi+teks%0D%0Asebagai+bentuk+aplikasi+langsung+dari+STS.+Model+klasifikasi+teks+dibangun+dan%0D%0Adibandingkan+berdasarkan+dua+skema+pembagian+data%2C+yaitu+train-test+split+dan+k-%0D%0Afold+Cross+Validation.+Masalah+ketidakseimbangan+kelas+yang+muncul+selama%0D%0A%0D%0Aproses+klasifikasi+kemudian+diatasi+melalui+prosedur+resampling+menggunakan%0D%0Akombinasi+metode+Random+Undersampling+dan+Random+Oversampling.+Sama%0D%0Aseperti+tahap+sebelumnya+yaitu+STS%2C+tahap+klasifikasi+teks+juga+menerapkan%0D%0ABioWordVec+sebagai+metode+representasi+kata.%0D%0ASecara+keseluruhan%2C+model+hybrid+LSTM+%E2%80%93+CNN+yang+diusulkan+untuk%0D%0Aekstraksi+informasi+memiliki+performa+yang+lebih+baik+dibandingkan+model+CNN%2C%0D%0ALSTM%2C+dan+hybrid+CNN+%E2%80%93+LSTM+dengan+nilai+akurasi+mencapai+100%25+pada+tugas%0D%0ASTS+dan+mencapai+99%25+untuk+tugas+klasifikasi+teks.+Dengan+demikian%2C+dapat%0D%0Adisimpulkan+bahwa+model+DL+terbaik+untuk+melakukan+ekstraksi+informasi+pada%0D%0Apenelitian+ini+adalah+model+hybrid+LSTM+%E2%80%93+CNN+dengan+implementasi+word%0D%0Aembedding+khusus+domain+biomedis%2C+yaitu+BioWordVec.%0D%0AKata+Kunci%3A+Klasifikasi+Teks%2C+Semantic+Text+Similarity%2C+BioWordVec%2C+Hybrid%0D%0A%0D%0ALSTM+%E2%80%93+CNN.%0D%0AThe+rise+in+death+rates+associated+with+leukemia+has+fueled+the+rapid+expansion+of%0D%0Apublications+focused+on+this+disease.+The+rise+in+the+number+of+publications+has%0D%0Asubstantially+affected+the+growth+of+biomedical+literature%2C+making+it+more%0D%0Achallenging+to+manually+extract+pertinent+information+concerning+leukemia.+That+is%0D%0Abecause+prior+studies+typically+focused+solely+on+the+lexical+and+syntactic+aspects%0D%0Aof+the+text%2C+neglecting+its+semantic+significance.%0D%0AThis+study+aims+to+identify+the+most+effective+Deep+Learning+(DL)+model%0D%0Afor+extracting+semantically+significant+information+from+a+substantial+volume+of%0D%0Abiomedical+data+called+the+Biomedical+Big+Data+Report.+Semantic+Text+Similarity%0D%0A(STS)+is+a+crucial+field+of+research+in+contemporary+applications+that+deal+with+the%0D%0Asemantic+analysis+of+texts.+This+approach+enhances+extracting+information+from+a%0D%0Atext+by+utilizing+a+distributed+model+of+words+or+an+external+source+of+organized%0D%0Asemantic+knowledge%2C+such+as+word+embedding.+Nevertheless%2C+it+is+essential+to%0D%0Aacknowledge+that+word+embedding+must+suit+the+specific+research+field.%0D%0AThis+study+suggests+implementing+the+Siamese+Manhattan+architecture+in%0D%0ADeep+Learning+(DL)+models%2C+namely+Convolutional+Neural+Networks+(CNN)%2C+Long%0D%0AShort-Term+Memory+(LSTM)+networks%2C+hybrid+CNN-LSTM+models%2C+and+hybrid%0D%0ALSTM-CNN+models%2C+to+do+semantic+analysis+on+biomedical+text.+Biomedical%0D%0Alanguage+with+semantic+significance+or+in+the+same+context+is+transformed+into%0D%0Avector+representation+using+word+embedding+techniques+specifically+designed+for%0D%0Athe+biomedical+field%2C+known+as+BioWordVec.+In+addition%2C+the+models+are%0D%0Aconstructed+and+evaluated+based+on+the+number+of+hidden+layers+and+labelling%0D%0Atechniques+employed.+Two+to+three+hidden+layers+are+utilised%2C+along+with+the+Cosine%0D%0ASimilarity+(CS)+and+Word+Mover's+Distance+(WMD)+tagging+methods.+The+findings%0D%0Aof+the+semantic+analysis+indicate+that+every+sentence+has+the+same+semantic%0D%0Ameaning%2C+with+a+similarity+level+of+1.%0D%0AThese+results+are+the+foundation+for+text+classification%2C+which+directly%0D%0Aimplements+STS.+Text+classification+models+are+constructed+and+evaluated+using%0D%0Atwo+data+partitioning+methods%3A+train-test+split+and+k-fold+Cross+Validation.+The+class%0D%0Aimbalance+issue+during+the+classification+process+is+addressed+using+a+resampling%0D%0Atechnique+that+combines+Random+Undersampling+and+Random+Oversampling%0D%0A%0D%0Aapproaches.+Like+the+previous+step%2C+STS%2C+the+text+categorization+stage+utilizes%0D%0ABioWordVec+to+represent+words.%0D%0AThe+hybrid+LSTM+%E2%80%93+CNN+model+outperforms+the+CNN%2C+LSTM%2C+and+hybrid%0D%0ACNN+%E2%80%93+LSTM+models+in+information+extraction%2C+achieving+accuracy+rates+of+100%25%0D%0Aon+the+STS+task+and+99%25+on+the+text+classification+task.+Thus%2C+it+can+be+concluded%0D%0Athat+the+best+DL+model+for+extracting+information+in+this+research+is+a+hybrid+LSTM%0D%0A%E2%80%93+CNN+model+with+the+implementation+of+word+embedding+specifically+for+the%0D%0Abiomedical+domain%2C+namely+BioWordVec.%0D%0AKeywords%3A+Text+Classification%2C+Semantic+Text+Similarity%2C+BioWordVec%2C+Hybrid%0D%0A%0D%0ALSTM+%E2%80%93+CNN.&rft.publisher=FAKULTAS+MATEMATIKA+DAN+ILMU+PENGETAHUAN+ALAM+&rft.date=2024-03-22&rft.type=Disertasi&rft.type=NonPeerReviewed&rft.format=text&rft.identifier=http%3A%2F%2Fdigilib.unila.ac.id%2F83781%2F1%2F1.%2520ABSTRAK%2520-%2520Dian%2520Kurniasari.pdf&rft.format=text&rft.identifier=http%3A%2F%2Fdigilib.unila.ac.id%2F83781%2F3%2F2.%2520DISERTASI%2520FULL%2520-%2520Dian%2520Kurniasari.pdf&rft.format=text&rft.identifier=http%3A%2F%2Fdigilib.unila.ac.id%2F83781%2F2%2F3.%2520DISERTASI%2520FULL%2520TANPA%2520BAB%2520PEMBAHASDian%2520Kurniasari.pdf&rft.identifier=+++Dian+%2C+Kurniasari+++(2024)+PENDEKATAN+BARU+EKSTRAKSI+INFORMASI+BIOMEDICAL+BIG+DATA+REPORT+DENGAN+BIOWORDVEC+MENGGUNAKAN+MODEL+HYBRID+LONG+SHORT-TERM+MEMORY+%E2%80%93+CONVOLUTIONAL+NEURAL+NETWORK+(LSTM+%E2%80%93+CNN).++%5BDisertasi%5D+++++&rft.relation=http%3A%2F%2Fdigilib.unila.ac.id%2F83781%2F