On Wikipedia, articles are categorized to aid readers in navigating content efficiently. The manual creation of new categories can be laborious and time-intensive. To tackle this issue, we built language models to translate Wikipedia categories from English to Vietnamese with a dataset containing 15,000 English-Vietnamese category pairs. Subsequently, small to medium-scale Transformer pre-trained models with a sequence-to-sequence architecture were fine-tuned for category translation. The experiments revealed that OPUS-MT-en-vi surpassed other models, attaining the highest performance with a BLEU score of 0.73, despite its smaller model storage. We expect our paper to be an alternative solution for translation tasks with limited computer resources.
University of Helsinki（赫尔辛基大学）；Wikimedia Foundation（维基媒体基金会）
The quality and quantity of articles in each Wikipedia language varies greatly. Translating from another Wikipedia is a natural way to add more content, but the translation process is not properly supported in the software used by Wikipedia. Past computer-assisted translation tools built for Wikipedia are not commonly used. We created a tool that adapts to the specific needs of an open community and to the kind of content in Wikipedia. Qualitative and quantitative data indicates that the new tool helps users translate articles easier and faster.
We present Namesakes, a dataset of ambiguously named entities obtained from English-language Wikipedia and news articles. It consists of 58862 mentions of 4148 unique entities and their namesakes: 1000 mentions from news, 28843 from Wikipedia articles about the entity, and 29019 Wikipedia backlink mentions. Namesakes should be helpful in establishing challenging benchmarks for the task of named entity linking (NEL).
Wikipedia is a great source of general world knowledge which can guide NLP models better understand their motivation to make predictions. Structuring Wikipedia is the initial step towards this goal which can facilitate fine-grain classification of articles. In this work, we introduce the Shinra 5-Language Categorization Dataset (SHINRA-5LDS), a large multi-lingual and multi-labeled set of annotated Wikipedia articles in Japanese, English, French, German, and Farsi using Extended Named Entity (ENE) tag set. We evaluate the dataset using the best models provided for ENE label set classification …
FPT Technology Research Institute（FPT技术研究院）；FPT University（FPT大学）；Vietnam National University, Hanoi（河内国家大学）
The FPT.AI team participated in the SHINRA2020-ML subtask of the NTCIR-15 SHINRA task. This paper describes our method to solving the problem and discusses the official results. Our method focuses on learning cross-lingual representations, both on the word level and document level for page classification. We propose a three-stage approach including multilingual model pre-training, monolingual model fine-tuning and cross-lingual voting. Our system is able to achieve the best scores for 25 out of 30 languages; and its accuracy gaps to the best performing systems of the other five languages are r…
Accessing Web archives raises a number of issues caused by their temporal characteristics. Additional knowledge is needed to find and understand older texts. Especially entities mentioned in texts are subject to change. Most severe in terms of information retrieval are name changes. In order to find entities that have changed their name over time, search engines need to be aware of this evolution. We tackle this problem by analyzing Wikipedia in terms of entity evolutions mentioned in articles. We present statistical data on excerpts covering name changes, which will be used to discover simila…
Working with Web archives raises a number of issues caused by their temporal characteristics. Depending on the age of the content, additional knowledge might be needed to find and understand older texts. Especially facts about entities are subject to change. Most severe in terms of information retrieval are name changes. In order to find entities that have changed their name over time, search engines need to be aware of this evolution. We tackle this problem by analyzing Wikipedia in terms of entity evolutions mentioned in articles regardless the structural elements. We gathered statistics and…
Faculty of Humanities and Social Sciences, University of Zagreb（萨格勒布大学人文与社会科学学院）
This article presents the application of the Universal Named Entity framework to generate automatically annotated corpora. By using a workflow that extracts Wikipedia data and meta-data and DBpedia information, we generated an English dataset which is described and evaluated. Furthermore, we conducted a set of experiments to improve the annotations in terms of precision, recall, and F1-measure. The final dataset is available and the established workflow can be applied to any language with existing Wikipedia and DBpedia. As part of future research, we intend to continue improving the annotation…
The evolution of named entities affects exploration and retrieval tasks in digital libraries. An information retrieval system that is aware of name changes can actively support users in finding former occurrences of evolved entities. However, current structured knowledge bases, such as DBpedia or Freebase, do not provide enough information about evolutions, even though the data is available on their resources, like Wikipedia. Our \emph{Evolution Base} prototype will demonstrate how excerpts describing name evolutions can be identified on these websites with a promising precision. The descripti…
The text generated on social media platforms is essentially a mixed lingual text. The mixing of language in any form produces considerable amount of difficulty in language processing systems. Moreover, the advancements in language processing research depends upon the availability of standard corpora. The development of mixed lingual Indian Named Entity Recognition (NER) systems are facing obstacles due to unavailability of the standard evaluation corpora. Such corpora may be of mixed lingual nature in which text is written using multiple languages predominantly using a single script only. The …
University of Amsterdam（阿姆斯特丹大学）；Informatics Institute（信息学研究所）
Neural Machine Translation (NMT) is an open vocabulary problem. As a result, dealing with the words not occurring during training (a.k.a. out-of-vocabulary (OOV) words) have long been a fundamental challenge for NMT systems. The predominant method to tackle this problem is Byte Pair Encoding (BPE) which splits words, including OOV words, into sub-word segments. BPE has achieved impressive results for a wide range of translation tasks in terms of automatic evaluation metrics. While it is often assumed that by using BPE, NMT systems are capable of handling OOV words, the effectiveness of BPE in …
Fraunhofer IAIS（弗劳恩霍夫智能分析与信息系统研究所）；Zerotha Research；Cerence GmbH；University of Bonn（波恩大学）；Goldman Sachs（高盛）
Pretrained Transformer models have emerged as state-of-the-art approaches that learn contextual information from text to improve the performance of several NLP tasks. These models, albeit powerful, still require specialized knowledge in specific scenarios. In this paper, we argue that context derived from a knowledge graph (in our case: Wikidata) provides enough signals to inform pretrained transformer models and improve their performance for named entity disambiguation (NED) on Wikidata KG. We further hypothesize that our proposed KG context can be standardized for Wikipedia, and we evaluate …
Los Alamos National Laboratory（洛斯阿拉莫斯国家实验室）；The University of Utah（犹他大学）；The University of Iowa（爱荷华大学）
Traditional disease surveillance systems suffer from several disadvantages, including reporting lags and antiquated technology, that have caused a movement towards internet-based disease surveillance systems. Internet systems are particularly attractive for disease outbreaks because they can provide data in near real-time and can be verified by individuals around the globe. However, most existing systems have focused on disease monitoring and do not provide a data repository for policy makers or researchers. In order to fill this gap, we analyzed Wikipedia article content. We demonstrate how a…
University of Amsterdam（阿姆斯特丹大学）；Facebook AI Research（脸书人工智能研究实验室）；ENS, PSL University（巴黎高等师范学院、PSL大学）；Inria（法国国家信息与自动化研究所）；University College London（伦敦大学学院）
Entities are at the center of how we represent and aggregate knowledge. For instance, Encyclopedias such as Wikipedia are structured by entities (e.g., one per Wikipedia article). The ability to retrieve such entities given a query is fundamental for knowledge-intensive tasks such as entity linking and open-domain question answering. Current approaches can be understood as classifiers among atomic labels, one for each entity. Their weight vectors are dense entity representations produced by encoding entity meta information such as their descriptions. This approach has several shortcomings: (i)…
Transliteration is a task of translating named entities from a language to another, based on phonetic similarity. The task has embraced deep learning approaches in recent years, yet, most ignore the phonetic features of the involved languages. In this work, we incorporate phonetic information into neural networks in two ways: we synthesize extra data using forward and back-translation but in a phonetic manner; and we pre-train models on a phonetic task before learning transliteration. Our experiments include three language pairs and six directions, namely English to and from Chinese, Hebrew an…
Recent progress in neural machine translation (NMT) has made it possible to translate successfully between monolingual language pairs where large parallel data exist, with pre-trained models improving performance even further. Although there exists work on translating in code-mixed settings (where one of the pairs includes text from two or more languages), it is still unclear what recent success in NMT and language modeling exactly means for translating code-mixed text. We investigate one such context, namely MT from code-mixed Modern Standard Arabic and Egyptian Arabic (MSAEA) into English. W…
