UFRGSandLIF at SemEval-2016 Task 10: Rule-Based MWE Identification and Predominant-Supersense Tagging

Cordeiro, Silvio; Ramisch, Carlos; Villavicencio, Aline

doi:10.18653/v1/s16-1140

“…A comparable result was observed in the 2016 DiMSUM shared task (Schneider et al, 2014), in which a rule-based baseline was ranked second. This system extracted MWEs from the training corpus and then annotated them in the test corpus based on lemma/part-of-speech matching and heuristics such as allowing a limited number of intervening words (Cordeiro et al, 2016).…”

Section: Progress Potential In Seen Datamentioning

confidence: 99%

Without lexicons, multiword expression identification will never fly: A position statement

Savary¹,

Cordeiro²,

Ramisch³

2019

Proceedings of the Joint Workshop on Multiword Expressions and WordNet (MWE-WN 2019)

Self Cite

View full text Add to dashboard Cite

Because most multiword expressions (MWEs), especially verbal ones, are semantically non-compositional, their automatic identification in running text is a prerequisite for semantically-oriented downstream applications. However, recent developments, driven notably by the PARSEME shared task on automatic identification of verbal MWEs, show that this task is harder than related tasks, despite recent contributions both in multilingual corpus annotation and in computational models. In this paper, we analyse possible reasons for this state of affairs. They lie in the nature of the MWE phenomenon, as well as in its distributional properties. We also offer a comparative analysis of the state-of-the-art systems, which exhibit particularly strong sensitivity to unseen data. On this basis, we claim that, in order to make strong headway in MWE identification, the community should bend its mind into coupling identification of MWEs with their discovery, via syntactic MWE lexicons. Such lexicons need not necessarily achieve a linguistically complete modelling of MWEs' behavior, but they should provide minimal morphosyntactic information to cover some potential uses, so as to complement existing MWE-annotated corpora. We define requirements for such a minimal NLP-oriented lexicon, and we propose a roadmap for the MWE community driven by these requirements.

show abstract

“…The traditional CE techniques interpret any single and multiple token nominal chunk as a concept [32] or do a dictionary lookup, as, e.g., DBpedia Spotlight [5], which matches and links identified nominal chunks with DBpedia entries (6.6M entities, 13 billion RDF triples) 4 , based on the Apache OpenNLP 5 models for phrase chunking and named entity recognition (NER). Given the large coverage of DBpedia, the performance of DBpedia Spotlight is rather competitive.…”

Section: Related Workmentioning

confidence: 99%

“…Selection of n-grams as fragments of NP chunks that can form part of multiple token concepts. For this task, we formed the PoS-patterns based on Penn Treebank tagset 10 , which were inherited from the patterns for multiword expression detection introduced in [4] and expanded here resulting in the following set:…”

Section: Compilation Of the Training Corpusmentioning

confidence: 99%

Concept Extraction Using Pointer–Generator Networks and Distant Supervision for Data Augmentation

Shvets

¹

,

Wanner

²

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Concept extraction is crucial for a number of downstream applications. However, surprisingly enough, straightforward single token/nominal chunk-concept alignment or dictionary lookup techniques such as DBpedia Spotlight still prevail. We propose a generic opendomain OOV-oriented extractive model that is based on distant supervision of a pointer-generator network leveraging bidirectional LSTMs and a copy mechanism. The model has been trained on a large annotated corpus compiled specifically for this task from 250K Wikipedia pages, and tested on regular pages, where the pointers to other pages are considered as ground truth concepts. The outcome of the experiments shows that our model significantly outperforms standard techniques and, when used on top of DBpedia Spotlight, further improves its performance. The experiments furthermore show that the model can be readily ported to other datasets on which it equally achieves a state-of-the-art performance.

show abstract

“…Selection of n-grams as fragments of NP chunks that can form part of multiple token concepts. For this task, we formed the PoS-patterns based on Penn Treebank tagset 10 , which were inherited from the patterns for multiword expression detection introduced in [4] and expanded here resulting in the following set: P = {N N, J N, V N, N J, J J, V J, N of N, N of DT N, N of J, N of DT J, N of V, N of DT V, CD N, CD J}, where N stands for "noun", i.e., NN|NNS|NNP|NNPS, J stands for "adjective", i.e., JJ|JJR|JJS, V -"verb" but limited to VBD|VBG|VN, CD -"cardinal number", DT -"determiner", and "of" is an exact pronoun. Each pattern matches an n-gram with two open-class lexical items and at most two auxiliary tokens between them.…”

Section: Compilation Of the Training Corpusmentioning

confidence: 99%

Concept Extraction Using Pointer-Generator Networks

Shvets,

Wanner

2020

Preprint

View full text Add to dashboard Cite

Concept extraction is crucial for a number of downstream applications. However, surprisingly enough, straightforward single token/nominal chunk-concept alignment or dictionary lookup techniques such as DBpedia Spotlight still prevail. We propose a generic opendomain OOV-oriented extractive model that is based on distant supervision of a pointer-generator network leveraging bidirectional LSTMs and a copy mechanism. The model has been trained on a large annotated corpus compiled specifically for this task from 250K Wikipedia pages, and tested on regular pages, where the pointers to other pages are considered as ground truth concepts. The outcome of the experiments shows that our model significantly outperforms standard techniques and, when used on top of DBpedia Spotlight, further improves its performance. The experiments furthermore show that the model can be readily ported to other datasets on which it equally achieves a state-of-the-art performance.

show abstract

UFRGSandLIF at SemEval-2016 Task 10: Rule-Based MWE Identification and Predominant-Supersense Tagging

Cited by 10 publications

References 12 publications

Without lexicons, multiword expression identification will never fly: A position statement

Without lexicons, multiword expression identification will never fly: A position statement

Concept Extraction Using Pointer–Generator Networks and Distant Supervision for Data Augmentation

Concept Extraction Using Pointer-Generator Networks

Contact Info

Product

Resources

About