Aspect Detection using Word and Char Embeddings with (Bi) LSTM and CRF

Augustyniak, Łukasz; Kajdanowicz, Tomasz; Kazienko, Przemysław

doi:10.1109/aike.2019.00016

Cited by 17 publications

(13 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Lapisan ini akan melakukan klasifikasi berita sesuai dengan inputan yang diterima dari lapisan sebelumnya. Pada lapisan ini terdiri dari LSTM yang tersusun dari gerbang (gates) yaitu forget gate, input gate, dan gerbang keluaran output gate yang berguna untuk mengontrol keseimbangan input yang akan dimasukan ke memory cell dan keseimbangan dari keadaan sebelumnya yang akan hapus [14]. Proses di balik lapisan Bi-LSTM adalah membagi unit menjadi dua bagian.…”

Section: Bi-lstm Layerunclassified

“…Proses di balik lapisan Bi-LSTM adalah membagi unit menjadi dua bagian. Pertama yaitu forward states yang bertanggung jawab memproses informasi berdasarkan arah waktu positif atau kedepan, sedangkan yang kedua yaitu backward states memproses informasi berdasarkan arah waktu negatif atau mundur [14]. Pada penelitian ini, parameter yang digunakan pada lapisan Bi-LSTM yaitu units sebanyak 64, 128, dan 256.…”

Section: Bi-lstm Layerunclassified

See 1 more Smart Citation

Model Klasifikasi Berita Palsu Menggunakan Bidirectional LSTM dan Word2vec sebagai Vektorisasi

Pakpahan¹,

Panjaitan²,

Amalia³

et al. 2022

JATISI

View full text Add to dashboard Cite

Secara umum, klasifikasi didefinisikan sebagai metode pembelajaran yang mengklasifikasikan data ke dalam label kelas. Proses klasifikasi tersebut dapat dilakukan pada data structured dan unstructured berdasarkan training data yang telah dilakukan. Penelitian ini melakukan klasifikasi menggunakan data berita yang memiliki label berita fakta dan berita palsu. Proses klasifikasi berita tersebut menggunakan teknik Bidirectional LSTM dan menggunakan arsitektur CBOW pada Word2vec sebagai vektorisasi kata dalam membangun model klasifikasi berita. Ada tiga parameter utama yang digunakan dalam penelitian tugas akhir yaitu embedding size, windows size, dan units bilstm. Optimasi kinerja model akan dilihat dari pengaruh dari ketiga parameter tersebut. Performa model yang dibangun diukur dengan menggunakan metrik evaluasi accuracy, recall, precision, f1- score dan computational time. Hasil penelitian menunjukan bahwa model dengan windows size 3, embedding size 200, dan units 128 memiliki performa terbaik untuk data judul dengan akurasi 79.18%. Sementara untuk data konten model dengan windows size 5, embedding size 300, dan units 256 memiliki performa terbaik dengan akurasi 92.8%.

show abstract

Section: Bi-lstm Layerunclassified

Model Klasifikasi Berita Palsu Menggunakan Bidirectional LSTM dan Word2vec sebagai Vektorisasi

Pakpahan¹,

Panjaitan²,

Amalia³

et al. 2022

JATISI

View full text Add to dashboard Cite

show abstract

“…LSTMs are a Recurrent Neural Networks (RNN) (Medsker and Jain, 2001) which have an internal memory that allows them to store the information learned during training. LSTMs are frequently used in the reverse dictionary task (Sherstinsky, 2018) and in word and sentence embeddings tasks in general (Augustyniak, Kajdanowicz and Kazienko, 2019;Liu et al, 2020), as they can learn long-term dependencies between existing words in the sentence and thus compute context representation vectors for each word. BiLSTM for its part, is a variant of LSTMs, it allows a bidirectional representation of words (Augustyniak, Kajdanowicz and Kazienko, 2019).…”

Section: Advanced Modelmentioning

confidence: 99%

BL.Research at SemEval-2022 Task 1: Deep networks for Reverse Dictionary using embeddings and LSTM autoencoders

Bendahman¹,

Breton²,

Nicolaieff³

et al. 2022

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)

View full text Add to dashboard Cite

This paper describes our two deep learning systems that competed at SemEval-2022 Task 1 "CODWOE: Comparing Dictionaries and WOrd Embeddings". We participated in the subtask for the reverse dictionary which consists in generating vectors from glosses. We use sequential models that integrate several neural networks, starting from Embeddings networks until the use of Dense networks, Bidirectional Long Short-Term Memory (BiLSTM) networks and LSTM networks. All glosses have been preprocessed in order to consider the best representation form of the meanings for all words that appears. We achieved very competitive results in reverse dictionary with a second position in English and French languages when using contextualized embeddings, and the same position for English, French and Spanish languages when using char embeddings. Our source code can be found at GitHub 1 .

show abstract

“…The representation may be extended with additional ontologies (Bloehdorn and Hotho, 2004) or WordNets (Scott and Matwin, 1998;Piasecki et al, 2009;Misiaszek et al, 2014;Janz et al, 2017;Kocoń et al, 2019b) and used with SVM (Razavi et al, 2010) or logistic regression models (Waseem and Hovy, 2016;Sahlgren et al, 2018;Kocoń et al, 2018;Kocoń and Maziarz, 2021). New methods often use word embeddings (Wiegand et al, 2018;Łukasz Augustyniak et al, 2021) (Wiegand et al, 2018; mixed with character embeddings (Augustyniak et al, 2019), together with deep neural networks, e.g. CNN (Zampieri et al, 2019a) or LSTM (Yenala et al, 2017).…”

Section: Related Workmentioning

confidence: 99%

Controversy and Conformity: from Generalized to Personalized Aggressiveness Detection

Kanclerz¹,

Figas²,

Gruza³

et al. 2021

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Confer

Self Cite

View full text Add to dashboard Cite

There is content such as hate speech, offensive, toxic or aggressive documents, which are perceived differently by their consumers. They are commonly identified using classifiers solely based on textual content that generalize pre-agreed meanings of difficult problems. Such models provide the same results for each user, which leads to high misclassification rate observable especially for contentious, aggressive documents. Both document controversy and user nonconformity require new solutions. Therefore, we propose novel personalized approaches that respect individual beliefs expressed by either user conformity-based measures or various embeddings of their previous text annotations. We found that only a few annotations of most controversial documents are enough for all our personalization methods to significantly outperform classic, generalized solutions. The more controversial the content, the greater the gain. The personalized solutions may be used to efficiently filter unwanted aggressive content in the way adjusted to a given person.

show abstract

Aspect Detection using Word and Char Embeddings with (Bi) LSTM and CRF

Cited by 17 publications

References 24 publications

Model Klasifikasi Berita Palsu Menggunakan Bidirectional LSTM dan Word2vec sebagai Vektorisasi

Model Klasifikasi Berita Palsu Menggunakan Bidirectional LSTM dan Word2vec sebagai Vektorisasi

BL.Research at SemEval-2022 Task 1: Deep networks for Reverse Dictionary using embeddings and LSTM autoencoders

Controversy and Conformity: from Generalized to Personalized Aggressiveness Detection

Contact Info

Product

Resources

About