Di era digital, ulasan pelanggan memegang peranan penting dalam membentuk citra dan reputasi perusahaan termasuk maskapai penerbangan. Tingginya volume ulasan serta beragamnya gaya bahasa yang digunakan sering menyulitkan proses analisis secara manual. Penelitian ini bertujuan mengembangkan sistem peringkasan teks otomatis abstraktif pada ulasan maskapai penerbangan menggunakan metode Long Short-Term Memory (LSTM). Data penelitian diperoleh dari situs www.airlinequality.com. Model LSTM dengan arsitektur encoder-decoder dan mekanisme attention digunakan untuk menghasilkan ringkasan abstraktif. Proses pemodelan meliputi praproses teks, representasi kata menggunakan FastText, pelatihan model dengan fungsi loss sparse categorical crossentropy, serta optimisasi menggunakan algoritma Adam. Evaluasi dilakukan dengan metrik ROUGE-1 untuk menilai kualitas ringkasan. Hasil penelitian menunjukkan bahwa model LSTM mencapai nilai F1-score ROUGE-1 sebesar 0,9462, yang mencerminkan tingkat kesamaan yang tinggi antara ringkasan yang dihasilkan dan ringkasan referensi, baik dari segi ketepatan (precision) maupun cakupan informasi (recall). Sistem ini diharapkan dapat membantu industri penerbangan menganalisis dan memahami ulasan pelanggan, mempercepat pengambilan keputusan dalam peningkatan kualitas layanan, serta mengoptimalkan manajemen informasi dan perencanaan strategi bisnis maskapai.

In the digital era, customer reviews play an important role in shaping a company’s image and reputation including that of airlines. The high volume of reviews and the diversity of language styles used often make manual analysis challenging. This study aims to develop an abstractive automatic text summarization system for airline reviews using the Long Short-Term Memory (LSTM) method. The research data were obtained from the www.airlinequality.com website. An LSTM model with an encoder-decoder architecture and attention mechanism was employed to generate abstractive summaries. The modeling process included text preprocessing, word representation using FastText, model training with the sparse categorical crossentropy loss function, and optimization using the Adam algorithm. Evaluation was conducted using the ROUGE-1 metric to assess summary quality. The results show that the LSTM model achieved a ROUGE-1 F1-score of 0.9462, indicating a high level of similarity between the generated summaries and reference summaries in terms of both precision and recall. This system is expected to assist the airline industry in efficiently analyzing and understanding customer reviews, accelerating decision-making to improve service quality, and optimizing information management and business strategy planning.