Peringkasan Teks Otomatis adalah salah satu bidang Information Retrieval (IR) yang menawarkan solusi untuk mencari informasi penting dalam suatu dokumen, dengan adanya peringkasan teks maka dokumen yang tersimpan didalam database tidak akan overload, waktu retrieval akan semakin cepat. Sejauh ini banyak implementasi tentang Peringkasan Teks Otomatis tetapi hasil nilai akurasinya rendah. Sehingga dalam penelitian ini dikembangkan aplikasi Peringkasan Teks Otomatis Berbahasa Indonesia Berbasis Natural Language Processing dan Jaringan Syaraf Tiruan. Metode Single Layer Perceptron (SLP ) merupakan metode pembelajaran terbimbing pertama dalam JST. SLP ini dipilih karena salah satu bentuk JST yang sederhana. Nilai threshold SLP bisa dilatih sehingga sistem secara otomatis menentukan kalimat penting mana yang akan dijadikan ringkasan. Perhitungan bobot term berdasar prinsip Tf*Idf kemudian dilakukan 6 klasifikasi NLP yakni Bias, Nominal, Verbal, Adjectiv, Adverb dan Numerik. Hasilnya diklasifikasikan menjadi kalimat penting dan tidak penting. Proses evaluasi dihitung berdasarkan teks yang diringkas oleh ahli Tata Bahasa Indonesia menggunakan Majority Opinion. Dari hasil penelitian menunjukkan nilai rata-rata Precision sebesar 70.4 % , nilai rata-rata Recall sebesar 91,5 % , nilai rata-rata F-Measures sebesar 78,8 % dan nilai Accuracy sebesar 68,3 %.

Automatic Text Summarization (ATS) is one of the fields of Information Retrieval (IR) offers solutions to find important information in a document, with the summarization of text then the documents stored in the database will not overload, faster retrieval time. So far many ATS methods are implements but still have low accuracy so Automatic Text Summarization application Speak Indonesia-based Natural Language Processing and Neural Networks (ANN) is developed in this research. Method of Single Layer Perceptron (SLP) is a method of learning the first social interactions in Neural Networks. The SLP was chosen because it is a simple. Threshold value is this SLP trained so that the system automatically determines the important sentence which would be used as a summary. Calculation of the term weighting based on the principle of the Tf * Idf then conducted 6 classifications NLP i.e. Bias, Nominal, Verbal, Adjectiv, Nominal2/Adverb and numerically. The result classified into two patterns, i.e. the sentence is important and not important. The evaluation process is calculated based on the summarized text by the experts of Indonesian Grammer uses the Majority Opinion. The results showed Precision average value of 70.4%, an average value of 91.5% Recall, the average value of the F-Measures by 78.8% and amounted to 68.3% Accuracy value.

Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal Baca Jurnal