Open Information Extraction from Question-Answer Pairs

Bhutani, Nikita; Suhara, Yoshihiko; Tan, Wang-Chiew; Halevy, Alon; Jagadish, H. V.

doi:10.18653/v1/n19-1239

Cited by 18 publications

(17 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…(Fader et al, 2011;Mausam et al, 2012;Del Corro and Gemulla, 2013), most recent open IE research has focused on deep-neural-network-based supervised learning models. Such systems are typically based on bidirectional long short-term memory (BiLSTM) and are formulated for two categories: sequence labeling (Stanovsky et al, 2018;Sarhan and Spruit, 2019;Jia and Xiang, 2019) and sequence generation (Cui et al, 2018;Sun et al, 2018;Bhutani et al, 2019). The latter enables flexible extraction; however, it is more computationally expensive than the former.…”

Section: Introductionmentioning

confidence: 99%

Multiˆ2OIE: Multilingual Open Information Extraction Based on Multi-Head Attention with BERT

Ro¹,

Lee²,

Kang³

2020

Findings of the Association for Computational Linguistics: EMNLP 2020

View full text Add to dashboard Cite

In this paper, we propose Multi 2 OIE, which performs open information extraction (open IE) by combining BERT (Devlin et al., 2019) with multi-head attention blocks (Vaswani et al., 2017). Our model is a sequence-labeling system with an efficient and effective argument extraction method. We use a query, key, and value setting inspired by the Multimodal Transformer (Tsai et al., 2019) to replace the previously used bidirectional long short-term memory architecture with multihead attention. Multi 2 OIE outperforms existing sequence-labeling systems with high computational efficiency on two benchmark evaluation datasets, Re-OIE2016 and CaRB. Additionally, we apply the proposed method to multilingual open IE using multilingual BERT. Experimental results on new benchmark datasets introduced for two languages (Spanish and Portuguese) demonstrate that our model outperforms other multilingual systems without training data for the target languages.

show abstract

Section: Introductionmentioning

confidence: 99%

Multiˆ2OIE: Multilingual Open Information Extraction Based on Multi-Head Attention with BERT

Ro¹,

Lee²,

Kang³

2020

Findings of the Association for Computational Linguistics: EMNLP 2020

View full text Add to dashboard Cite

show abstract

“…To the best of our knowledge, there is no typed dependencies (grammatical relations) analysis available for Arabic from an architectural perspective; Stanford CoreNLP does not attempt to do everything. It is nothing more than a straightforward pipeline architecture [21] [22].…”

Section: Dependency Annotation Schemementioning

confidence: 99%

ADPBC: Arabic Dependency Parsing Based Corpora for Information Extraction

Mohamed¹,

Hussien²,

Mousa³

2021

IJITCS

View full text Add to dashboard Cite

There is a massive amount of different information and data in the World Wide Web, and the number of Arabic users and contents is widely increasing. Information extraction is an essential issue to access and sort the data on the web. In this regard, information extraction becomes a challenge, especially for languages, which have a complex morphology like Arabic. Consequently, the trend today is to build a new corpus that makes the information extraction easier and more precise. This paper presents Arabic linguistically analyzed corpus, including dependency relation. The collected data includes five fields; they are a sport, religious, weather, news and biomedical. The output is CoNLL universal lattice file format (CoNLL-UL). The corpus contains an index for the sentences and their linguistic meta-data to enable quick mining and search across the corpus. This corpus has seventeenth morphological annotations and eight features based on the identification of the textual structures help to recognize and understand the grammatical characteristics of the text and perform the dependency relation. The parsing and dependency process conducted by the universal dependency model and corrected manually. The results illustrated the enhancement in the dependency relation corpus. The designed Arabic corpus helps to quickly get linguistic annotations for a text and make the information Extraction techniques easy and clear to learn. The gotten results illustrated the average enhancement in the dependency relation corpus.

show abstract

“…Dalam pembuatan penelitian ini dgunakan beberapa tinjauan pustaka untuk membantu dalam proses penelitian [6], [7]. Tinjauan pustaka yang digunakan tidak sama persis dengan penelitian ini, tetapi memiliki prinsip pendekatan yang mirip.…”

Section: Rule-based Atau Statisticalunclassified

Information Extraction Berbasis Rule Untuk Soal Ujian

Soenardjo

Gunawan

2021

INSYST

View full text Add to dashboard Cite

Proses information extraction dapat dilakukan pada beberapa macam media, seperti artikel berita, tanya jawab dan sebagainya. Penelitian ini mencoba untuk melakukan information extraction pada media soal ujian yang dilengkapi dengan jawaban. Pendekatan pengolahan informasi yang dibahas dalam penelitian ini adalah information extraction berbasis rule. Informasi yang hendak digali adalah informasi data soal ujian beserta jawabannya. Inputan dalam penelitian ini pasangan file soal dan jawaban milik Cambridge. Ada beberapa mata pelajaran yang digunakan, yaitu Biologi, Matematika dan Ekonomi. Jenis soal yang digunakan juga ada beberapa macam, yaitu pilihan ganda dan esai. Hasil penelitian ini diharapkan bisa menjadi media pembelajaran. Penelitian dilakukan dengan menggunakan sebanyak 100 pasang data soal dan ujian. Sistem akan menerima 2 inputan file dengan format PDF. Kedua file ini merupakan pasangan soal dan jawaban. Proses yang diakukan adalah file akan dirubah menjadi 2, yaitu file HTML dan file PNG. File HTML mengandung semua teks soal dan file PNG mengandung semua gambar dari soal. Sistem akan mengambil teks dan gambar dari masing-masing soal dan jawaban berdasrkan rule yang sudah ditentukan. Penentuan rule dilakukan secara manual dengan mempelajari pola-pola data yang tedapat dalam tag HTML. Setelah proses ekstraksi, soal dan jawaban ini dipasangkan sesuai dengan nomor urutnya masing-masing. Pasangan soal dan jawaban ini kemudian akan disimpan ke dalam database. Dari hasil penelitian, tingkat akurasi yang didapatkan adalah sekitar 46%. Kendala utama yang dihadapi adalah format soal dan jawaban yang tidak strandar sehingga menimbulkan kesulitan dalam proses ekstraksi informasi.

show abstract

Open Information Extraction from Question-Answer Pairs

Cited by 18 publications

References 29 publications

Multiˆ2OIE: Multilingual Open Information Extraction Based on Multi-Head Attention with BERT

Multiˆ2OIE: Multilingual Open Information Extraction Based on Multi-Head Attention with BERT

ADPBC: Arabic Dependency Parsing Based Corpora for Information Extraction

Information Extraction Berbasis Rule Untuk Soal Ujian

Contact Info

Product

Resources

About