Cross-lingual Named Entity Recognition via FastAlign: a Case Study

Hatami, Ali; Mitkov, Ruslan; Corpas, Gloria

doi:10.26615/978-954-452-071-7_010

Cited by 4 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With respect to obtaining cross-lingual annotation information, the basic concept of projecting label data in language pairs via word alignment has been discussed in various NLP contexts [47][48][49][50][51][52][53][54][55][56][57]. For medical use cases, little research exists [49], with focus on English and Chinese data and models.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

German Medical Named Entity Recognition Model and Data Set Creation Using Machine Translation and Word Alignment: Algorithm Development and Validation

Frei¹

2023

JMIR Form Res

View full text Add to dashboard Cite

Background Data mining in the field of medical data analysis often needs to rely solely on the processing of unstructured data to retrieve relevant data. For German natural language processing, few open medical neural named entity recognition (NER) models have been published before this work. A major issue can be attributed to the lack of German training data. Objective We developed a synthetic data set and a novel German medical NER model for public access to demonstrate the feasibility of our approach. In order to bypass legal restrictions due to potential data leaks through model analysis, we did not make use of internal, proprietary data sets, which is a frequent veto factor for data set publication. Methods The underlying German data set was retrieved by translation and word alignment of a public English data set. The data set served as a foundation for model training and evaluation. For demonstration purposes, our NER model follows a simple network architecture that is designed for low computational requirements. Results The obtained data set consisted of 8599 sentences including 30,233 annotations. The model achieved a class frequency–averaged F1 score of 0.82 on the test set after training across 7 different NER types. Artifacts in the synthesized data set with regard to translation and alignment induced by the proposed method were exposed. The annotation performance was evaluated on an external data set and measured in comparison with an existing baseline model that has been trained on a dedicated German data set in a traditional fashion. We discussed the drop in annotation performance on an external data set for our simple NER model. Our model is publicly available. Conclusions We demonstrated the feasibility of obtaining a data set and training a German medical NER model by the exclusive use of public training data through our suggested method. The discussion on the limitations of our approach includes ways to further mitigate remaining problems in future work.

show abstract

Section: Related Workmentioning

confidence: 99%

“…We projected the annotation labels onto the translated German sentences using the word-level mapping between the corresponding English and German sentence to obtain new annotation label indices in the German sentence. In nonmedical contexts, similar work on non-German target languages exists (eg, [57]).…”

Section: Overviewmentioning

confidence: 99%

German Medical Named Entity Recognition Model and Data Set Creation Using Machine Translation and Word Alignment: Algorithm Development and Validation

Frei¹

2023

JMIR Form Res

View full text Add to dashboard Cite

show abstract

“…The major NLP tasks relating to MWUs are (1) identifying and extracting MWUs from corpus data and disambiguating their internal syntax, and (2) interpreting MWUs. Steadily, the completion and evaluation of such tasks have fostered the celebration of dedicated workshops (Corpas Pastor et al 2016) and the publication of subsequent volumes (Mitkov et al 2018), and they have been pipelined with parsers and applications such as word alignment (Hatami, Mitkov and Corpas 2021).…”

Section: Introductionmentioning

confidence: 99%

A corpus-based analysis of mediation in EU multi-word organization names

Rodas¹

2022

Proceedings of the International Conference EUROPHRAS 2022 (Short Papers, Posters and MUMTTT Workshop Contributions)

View full text Add to dashboard Cite

This study aims at using Named Entity Recognition (NER) to extract a specific type of multi-word entity, that is, multi-word organization names (MWORGs), from an English-Spanish comparable corpus of European Parliament documents. Following a triadic, Peircean model of translation and grammar, we hypothesize that MWORGs are nominal constructions (or signs) which serve a semiotic function of mediation in EU translations (Stecconi 2009;Torres-Martínez 2022). Specific performance of the VIP-DeepPavlov NER system (Corpas Pastor 2021) with MWORGs is evaluated in terms of precision, recall, and F-1 scores. Relevant MWORGs are then annotated and analyzed from a contrastive, semi-constructional approach (Boas 2010) to determine how many of them are mediating, and under which schemata. Results predictably show that non-mediating constructions are prevalent in non-translated English (66 %), as mediating constructions are in translated Spanish (81 %). However, a surprising 34 % of the organization names in non-translated English are mediating; inversely, 19 % of the MWORGs in translated Spanish serve a non-mediating function. Seven different mediation schemes (blending, borrowing, translation, and further combinations of the three) where discovered among MWORGs, some of them languagepreferent. This reinforces our belief that names are largely disregarded semiotic hubs, and indeed a crucial piece in the understanding of (non-)translations and (non-)interpretations as construction-based grammars with a specific number of similar, different, and mediating rules in each language and textual typology.

show abstract