The SIGMORPHON 2016 Shared Task—Morphological Reinflection

Cotterell, Ryan; Kirov, Christo; Sylak-Glassman, John; Yarowsky, David; Eisner, Jason; Hulden, Mans

doi:10.18653/v1/w16-2002

Cited by 225 publications

(226 citation statements)

References 40 publications

Supporting

Mentioning

219

Contrasting

Unclassified

Order By: Relevance

“…Again, the later methods can be further classified depending on whether context of the current word is considered or not. Lemmatization without context (Cotterell et al, 2016;Nicolai and Kondrak, 2016) is closer to stemming and not the focus of the present work. It is noteworthy here that the supervised lemmatization methods do not try to classify the lemma of a given word form as it is infeasible due to having a large number of lemmas in a language.…”

Section: Related Workmentioning

confidence: 99%

Context Sensitive Lemmatization Using Two Successive Bidirectional Gated Recurrent Networks

Chakrabarty¹,

Pandit²,

Garain³

2017

Proceedings of the 55th Annual Meeting of the Association For Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

We introduce a composite deep neural network architecture for supervised and language independent context sensitive lemmatization. The proposed method considers the task as to identify the correct edit tree representing the transformation between a word-lemma pair. To find the lemma of a surface word, we exploit two successive bidirectional gated recurrent structures -the first one is used to extract the character level dependencies and the next one captures the contextual information of the given word. The key advantages of our model compared to the state-of-the-art lemmatizers such as Lemming and Morfette are -(i) it is independent of human decided features (ii) except the gold lemma, no other expensive morphological attribute is required for joint learning. We evaluate the lemmatizer on nine languages -Bengali, Catalan, Dutch, Hindi, Hungarian, Italian, Latin, Romanian and Spanish. It is found that except Bengali, the proposed method outperforms Lemming and Morfette on the other languages. To train the model on Bengali, we develop a gold lemma annotated dataset 1 (having 1, 702 sentences with a total of 20, 257 word tokens), which is an additional contribution of this work.

show abstract

Section: Related Workmentioning

confidence: 99%

Context Sensitive Lemmatization Using Two Successive Bidirectional Gated Recurrent Networks

Chakrabarty¹,

Pandit²,

Garain³

2017

Proceedings of the 55th Annual Meeting of the Association For Computational Linguistics (Volume 1: Long Papers)

View full text Add to dashboard Cite

show abstract

“…The accuracy of data extraction and label association for data from Wiktionary was verified according to the process described in Kirov et al (2016). However, verifying the full linguistic accuracy of the data was beyond the scope of preparation for the task, and errors that resulted from the original input of data by crowdsourced authors remained in some cases.…”

Section: Data Sources and Annotation Schemementioning

confidence: 99%

“…Closer to the other end, we find work that focuses on defining morphological models with limited lexicons that are then extended using raw text (Clément et al, 2004; Forsberg et al, 2006). The setting of the shared task on morphological reinflection (Cotterell et al, 2016), which provides a rich partly annotated training data set, encourages methods that are supervised.…”

Section: Related Workmentioning

confidence: 99%

“…Englishspeaking, Spanish-speaking). Kirov et al (2016) describe the process of extracting lemmas and inflected wordforms from Wiktionary, associating them with morphological labels from Wiktionary, and mapping those labels to a universalized annotation scheme for inflectional morphology called the UniMorph Schema (Sylak-Glassman et al, 2015b).…”

Section: Data Sources and Annotation Schemementioning

confidence: 99%

See 1 more Smart Citation

Proceedings of the 14th SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology

Elsner¹,

Kuebler²

2016

View full text Add to dashboard Cite

Welcome to the 14th SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology. The workshop aims to bring together researchers interested in applying computational techniques to problems in morphology, phonology, and phonetics. Our program this year highlights the ongoing and important interaction between work in computational linguistics and work in theoretical linguistics. We received 23 submissions and accepted 11.The volume of submissions made it necessary to recruit several additional reviewers. We'd like to thank all of these people for agreeing to review papers on what seemed like impossibly short notice.This year also marks the first SIGMORPHON shared task, on morphological reinflection. The shared task received 9 submissions, all of which were accepted, and greatly advanced the state of the art in this area.We thank all the authors, reviewers and organizers for their efforts on behalf of the community. AbstractThis paper conceptualizes speech prosody data mining and its potential application in data-driven phonology/phonetics research. We first conceptualize Speech Prosody Mining (SPM) in a time-series data mining framework. Specifically, we propose using efficient symbolic representations for speech prosody time-series similarity computation. We experiment with both symbolic and numeric representations and distance measures in a series of time-series classification and clustering experiments on a dataset of Mandarin tones. Evaluation results show that symbolic representation performs comparably with other representations at a reduced cost, which enables us to efficiently mine large speech prosody corpora while opening up to possibilities of using a wide range of algorithms that require discrete valued data. We discuss the potential of SPM using time-series mining techniques in future works. IntroductionCurrent investigations on the phonology of intonation and tones (or pitch accent) typically employ data-driven approaches by building research on top of manual annotations of a large amount of speech prosody data (for example, (Morén and Zsiga, 2006; Zsiga and Zec, 2013), and many others). Meanwhile, researchers are also limited by the amount of resources invested in such expensive endeavor of manual annotations. Given this paradox, we believe that this type of data driven approach in phonology-phonetics interface can benefit from tools that can efficiently index, query, classify, cluster, summarize, and discover meaningful prosodic patterns from a large speech prosody corpus.The data mining of f 0 1 (pitch) contour patterns from audio data has recently gained success in the domain of Music Information Retrieval (aka MIR, see (Gulati and Serra, 2014; Gulati et al., 2015; Ganguli, 2015) for examples). In contrast, the data mining of speech prosody f 0 data (here on referred to as Speech Prosody Mining (SPM) 2 ) is a less explored research topic (Raskinis and Kazlauskiene, 2013). Fundamentally, SPM in a large prosody corpus aims at discovering meaningful patterns in the f ...

show abstract

“…Results from the 2016 SIGMORPHON Shared Task on Morphological Reinflection (Cotterell et al, 2016) indicate that models based on recurrent neural networks can deliver high accuracies for reinflection. The winning system by Kann and Schütze (2016) achieved an average accuracy in excess of 95% when tested on 10 languages.…”

Section: Introductionmentioning

confidence: 99%

Data Augmentation for Morphological Reinflection

Silfverberg¹,

Wiemerslage²,

Liu³

et al. 2017

Proceedings of the CoNLL SIGMORPHON 2017 Shared Task: Universal Morphological Reinflection

View full text Add to dashboard Cite

This paper presents the submission of the Linguistics Department of the University of Colorado at Boulder for the 2017 CoNLL-SIGMORPHON Shared Task on Universal Morphological Reinflection. The system is implemented as an RNN Encoder-Decoder. It is specifically geared toward a low-resource setting. To this end, it employs data augmentation for counteracting overfitting and a copy symbol for processing characters unseen in the training data. The system is an ensemble of ten models combined using a weighted voting scheme. It delivers substantial improvement in accuracy compared to a non-neural baseline system in presence of varying amounts of training data.

show abstract

The SIGMORPHON 2016 Shared Task—Morphological Reinflection

Cited by 225 publications

References 40 publications

Context Sensitive Lemmatization Using Two Successive Bidirectional Gated Recurrent Networks

Context Sensitive Lemmatization Using Two Successive Bidirectional Gated Recurrent Networks

Proceedings of the 14th SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology

Data Augmentation for Morphological Reinflection

Contact Info

Product

Resources

About