PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation

Papadopoulos, Dimitris N.; Papadakis, Nikolaos; Matsatsinis, Nikolaos F.

doi:10.18653/v1/2021.eacl-srw.4

Cited by 6 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Knowledge graph construction is a task that has been applied in text with different types of domains ( [6]) and languages ( [7]). Regarding the Greek language, there is no other work for Knowledge Graph construction from Greek unstructured documents, except [8]. In [8] they proposed to utilize transformer models for machine translation from Greek to English and backwards, in order to and apply existing triple extraction tools for English texts.…”

Section: Related Workmentioning

confidence: 99%

“…Regarding the Greek language, there is no other work for Knowledge Graph construction from Greek unstructured documents, except [8]. In [8] they proposed to utilize transformer models for machine translation from Greek to English and backwards, in order to and apply existing triple extraction tools for English texts. Yet, this approach do not achieve satisfactory results with the domain-specific vocabulary of government documents.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Towards Knowledge Graph Creation from Greek Governmental Documents

Georgoudi

Konstantinidis

Meditskos

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Documents contain textual information, which is of the utmost importance for all the organizations. Document management systems have been used to store vast amounts of unstructured textual data described with minimal metadata, a method that has several limitations. In order to convert hidden knowledge into machine-readable data with rich connections, this paper presents work in progress on the development of the first end-to-end guided approach to construct a Knowledge Graph from Greek government documents from the Greek open government portal. The resulted Knowledge graph consists a proof-of-concept graph, that illustrates the beneficial semantic relationships between the textual data.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Towards Knowledge Graph Creation from Greek Governmental Documents

Georgoudi

Konstantinidis

Meditskos

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Moreover, Ope-nIE systems built for other languages often work only for a single language due to their reliance on language-specific resources. For example, Bassa et al (2018); Rahat and Talebpour (2018); Romadhony et al (2018); Guarasci et al (2020); Papadopoulos et al (2021) focus on German, Persian, Indonesian, Italian, and Greek, respectively. Claro et al (2019) present the importance of and various challenges involved with building multilingual OpenIE systems.…”

Section: Related Workmentioning

confidence: 99%

Alignment-Augmented Consistent Translation for Multilingual Open Information Extraction

Kolluru¹,

Mohammed²,

Mittal³

et al. 2022

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

Progress with supervised Open Information Extraction (OpenIE) has been primarily limited to English due to the scarcity of training data in other languages. In this paper, we explore techniques to automatically convert English text for training OpenIE systems in other languages. We introduce the Alignment-Augmented Consistent Translation (AACTRANS) model to translate English sentences and their corresponding extractions consistently with each other -with no changes to vocabulary or semantic meaning which may result from independent translations. Using the data generated with AACTRANS, we train a novel two-stage generative OpenIE model, which we call GEN2OIE, that outputs for each sentence: 1) relations in the first stage and 2) all extractions containing the relation in the second stage. GEN2OIE increases relation coverage using a training data transformation technique that is generalizable to multiple languages, in contrast to existing models that use an English-specific training loss. Evaluations on 5 languages -Spanish, Portuguese, Chinese, Hindi and Telugu -show that the GEN2OIE with AACTRANS data outperforms prior systems by a margin of 6-25% F1. 1 * denotes equal contribution 1 Code and models released at github.com:dair-iitd/moie

show abstract

“…The model was trained on the Greek and English version of the combined SNLI and MultiNLI (Williams et al, 2018) corpora (AllNLI). We used the Englishto-Greek machine translation model by Papadopoulos et al, 2021 to create the Greek version of the AllNLI dataset. The trained model takes the premise-hypothesis pair as input and predicts one of the following labels for each case: "contradiction": c, "entailment": e or "neutral": n. The logits for each class are then converted to probabilities using the softmax function.…”

Section: Natural Language Inferencementioning

confidence: 99%

“…Each row of the dataset comprises a claim in free text, a list of evidence information including a URL to the Wikipedia page of the corresponding evidence and an annotated label (SUPPORTS, REFUTES, NOT ENOUGH INFO). We manually translated a subset of 150 claims from the FEVER validation set from English to Greek and populated the graph database with the content of the corresponding Wikipedia URLs, which was automatically translated into Greek (due to its size), using the NMT model by Papadopoulos et al, 2021. We report FarFetched's performance in terms of accuracy, precision, recall and F1-score on Table 1.…”

Section: End-to-end Performancementioning

confidence: 99%

Proceedings of the Third Workshop on Deep Learning for Low-Resource Natural Language Processing

2022

View full text Add to dashboard Cite

The lack of resources for languages in the Americas has proven to be a problem for the creation of digital systems such as machine translation, search engines, chat bots, and more. The scarceness of digital resources for a language causes a higher impact on populations where the language is spoken by millions of people. We introduce the first official large combined corpus for deep learning of an indigenous South American low-resource language spoken by millions called Quechua. Specifically, our curated corpus is created from text gathered from the southern region of Peru where a dialect of Quechua is spoken that has not traditionally been used for digital systems as a target dialect in the past. In order to make our work repeatable by others, we also offer a public, pre-trained, BERT model called Qu-BERT which is the largest linguistic model ever trained for any Quechua type, not just the southern region dialect. We furthermore test our corpus and its corresponding BERT model on two major tasks: (1) named-entity recognition (NER) and (2) part-of-speech (POS) tagging by using state-of-the-art techniques where we achieve results comparable to other work on higher-resource languages. In this article, we describe the methodology, challenges, and results from the creation of QuBERT which is on on par with other state-of-the-art multilingual models for natural language processing achieving between 71 and 74% F1 score on NER and 84-87% on POS tasks. ReferencesWillem FH Adelaar. 2004. The languages of the Andes.

show abstract

PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation

Cited by 6 publications

References 15 publications

Towards Knowledge Graph Creation from Greek Governmental Documents

Towards Knowledge Graph Creation from Greek Governmental Documents

Alignment-Augmented Consistent Translation for Multilingual Open Information Extraction

Proceedings of the Third Workshop on Deep Learning for Low-Resource Natural Language Processing

Contact Info

Product

Resources

About