2020 GECToRGrammaticalErrorCorrectio

From GM-RKB
Jump to navigation Jump to search

Subject Headings: GECToR; Grammatical Error Correction System, GEC Sequence Tagging System, LaserTagger, BERT Encoder.

Notes

Cited By

Quotes

Abstract

In this paper, we present a simple and efficient GEC sequence tagger using a Transformer encoder. Our system is pre-trained on synthetic data and then fine-tuned in two stages: first on errorful corpora, and second on a combination of errorful and error-free parallel corpora. We design custom token-level transformations to map input tokens to target corrections. Our best single-model/ensemble GEC tagger achieves an F0.5 of 65.3 / 66.5 on CONLL-2014 (test) and F0.5 of 72.4 / 73.6 on BEA-2019 (test). Its inference speed is up to 10 times as fast as a Transformer-based seq2seq GEC system. The code and [[trained model]]s are publicly available[1].

1. Introduction

Neural Machine Translation (NMT) - based approaches (Sennrich et al., 2016a) have become the preferred method for the task of Grammatical Error Correction (GEC)[2]. In this formulation, errorful sentences correspond to the source language, and error-free sentences correspond to the target language. Recently, Transformer-based (Vaswani et al., 2017) sequence-to-sequence (seq2seq) models have achieved state-of-the-art performance on standard GEC benchmarks (Bryant et al., 2019). Now the focus of research has shifted more towards generating synthetic data for pretraining the Transformer-NMT-based GEC systems (Grundkiewicz et al., 2019; Kiyono et al., 2019). NMTbased GEC systems suffer from several issues which make them inconvenient for real world deployment: (i) slow inference speed, (ii) demand for large amounts of training data and (iii) interpretability and explainability; they require additional functionality to explain corrections, e.g., grammatical error type classification (Bryant et al., 2017).

In this paper, we deal with the aforementioned issues by simplifying the task from sequence generation to sequence tagging. Our GEC sequence tagging system consists of three training stages: pretraining on synthetic data, fine-tuning on an errorful parallel corpus, and finally, fine-tuning on a combination of errorful and error-free parallel corpora.

Related Work

LaserTagger (Malmi et al., 2019) combines a BERT encoder with an autoregressive Transformer decoder to predict three main edit operations: keeping a token, deleting a token, and adding a phrase before a token. PIE (Awasthi et al., 2019) is an iterative sequence tagging GEC system that predicts token-level edit operations. While their approach is the most similar to ours, our work differs from theirs as described in our contributions below:

1. We develop custom g-transformations: token-level edits to perform (g)rammatical error corrections. Predicting g-transformations instead of regular tokens improves the generalization of our GEC sequence tagging system.
2. We decompose the fine-tuning stage into two stages: fine-tuning on errorful-only sentences and further fine-tuning on a small, high-quality dataset containing both errorful and error-free sentences.
3. We achieve superior performance by incorporating a pre-trained Transformer encoder in our GEC sequence tagging system. In our experiments, encoders from XLNet and RoBERTa outperform three other cutting-edge Transformer encoders (ALBERT, BERT, and GPT-2).

2. Datasets

Table 1 describes the finer details of datasets used for different training stages.

Dataset # sentences % errorful sentences Training stage
PIE-synthetic 9,000,000 100.0% I
Lang-8 947,344 52.5% II
NUCLE 56,958 38.0% II
FCE 34,490 62.4% II
W&I+LOCNESS 34,304 67.3% II, III
Table 1: Training datasets. Training stage I is pretraining on synthetic data. Training stages II and III are for fine-tuning.

Synthetic Data

For pretraining stage I, we use 9M parallel sentences with synthetically generated grammatical errors (Awasthi et al., 2019) [3].

Training Data

We use the following datasets for fine-tuning stages II and III: National University of Singapore Corpus of Learner English (NUCLE)[4] (Dahlmeier et al., 2013), Lang-8 Corpus of Learner English (Lang-8)[5] (Tajiri et al., 2012), FCE dataset[6] (Yannakoudakis et al., 2011), the publicly available part of the Cambridge Learner Corpus (Nicholls, 2003) and Write & Improve + LOCNESS Corpus (Bryant et al., 2019)[7].

Evaluation Data

We report results on CoNLL2014 test set (Ng et al., 2014) evaluated by official M2 scorer (Dahlmeier and Ng, 2012), and on BEA-2019 dev and test sets evaluated by ERRANT (Bryant et al., 2017).

3. Token-Level Transformations

We developed custom token-level transformations $T(x_i)$ to recover the target text by applying them to the source tokens $\left(x_1 \ldots x_N\right)$. Transformations increase the coverage of grammatical error corrections for limited output vocabulary size for the most common grammatical errors, such as Spelling, Noun Number, Subject-Verb Agreement and Verb Form (Yuan, 2017, p.28).

The edit space which corresponds to our default tag vocabulary size = 5000 consists of 4971 basic transformations (token-independent KEEP, DELETE and 1167 token-dependent APPEND, 3802 REPLACE) and 29 token-independent g-transformations.

Basic transformations perform the most common token-level edit operations, such as: keep the current token unchanged (tag $KEEP), delete current token (tag $DELETE), append new token $t_1$ next to the current token $x_i$ (tag $APPEND.$t_1$) or replace the current token $x_i$ with another token $t_2$ (tag $REPLACE.$t_2$)

 g-transformations perform task-specific operations such as: change the case of the current token (CASE tags), merge the current token and the next token into a single one (MERGE tag]]s) and split the current token into two new tokens (SPLIT tags). Moreover, tags from NOUN NUMBER and VERB FORM transformations encode grammatical properties for tokens. For instance, these transformations include conversion of singular nouns to plurals and vice versa or even change the form of regular/irregular verbs to express a different number or tense.

To obtain the transformation suffix for the VERB FORM tag, we use the verb conjugation dictionary[8]. For convenience, it was converted into the following format: $token_0\dot token_1: tag_0 tag_1$ (e.g., go goes: $V\;B \cdot V\; BZ$). This means that there is a transition from $word_0$ and $word_1 to the respective tags. The transition is unidirectional, so if there exists a reverse transition, it is presented separately.

The experimental comparison of covering capabilities for our token-level transformations is in Table 2. All transformation types with examples are listed in Appendix, Table 9.

Tag vocab. size Transformations
Basic transf. All transf.
100 60.4% 79.7%
1000 76.4% 92.9%
5000 89.5% 98.1%
10000 93.5% 100.0%
Table 2: Share of covered grammatical errors in CoNLL-2014 for basic transformations only (KEEP, DELETE, APPEND, REPLACE) and for all transformations w.r.t. tag vocabulary’s size. In our work, we set the default tag vocabulary size = 5000 as a heuristical compromise between coverage and model size.

Preprocessing

To approach the task as a sequence tagging problem we need to convert each target sentence from training/evaluation sets into a sequence of tags where each tag is mapped to a single source token. Below is a brief description of our 3-step preprocessing algorithm for color-coded sentence pair from Table 3:

Step 1). Map each token from source sentence to subsequence of tokens from target sentence. [A ↦ A], [ten ↦ ten, -], [years ↦ year, -], [old ↦ old], [go ↦ goes, to], [school ↦ school, .].

For this purpose, we first detect the minimal spans of tokens which define differences between source tokens $\left(x_1 \ldots x_N \right)$ and target tokens $\left(y_1 \ldots y_M\right)$. Thus, such a span is a pair of selected source tokens and corresponding target tokens. We can’t use these span-based alignments, because we need to get tags on the token level. So then, for each [[source token $x_i$, $1 \leq i \leq N$ we search for best-fitting subsequence $\Upsilon_i = \left(y_{j1} \ldots y_{j2} \right)$, $1 \leq j_1 \leq j_2 \leq M$ of target tokens by minimizing the modified Levenshtein distance (which takes into account that successful g-transformation is equal to zero distance)

Step 2). For each mapping in the list, find token-level transformations which convert source token to the target subsequence: [A ↦ A]: $KEEP, [ten ↦ ten, -]: $KEEP, $MERGE HYPHEN, [years ↦ year, -]: $NOUN NUMBER SINGULAR, $MERGE HYPHEN], [old ↦ old]: $KEEP, [go ↦ goes, to]: $VERB FORM VB VBZ, $APPEND to, [school ↦ school, .]: $KEEP, $APPEND {.}].

Step 3). Leave only one transformation for each source token: A ⇔ $KEEP, ten ⇔ $MERGE HYPHEN, years ⇔ $NOUN NUMBER SINGULAR, old ⇔ $KEEP, go ⇔ $VERB FORM VB VBZ, school ⇔ $APPEND {.}.

The iterative sequence tagging approach adds a constraint because we can use only a single tag for each token. In case of multiple transformations we take the first transformation that is not a $KEEP tag. For more details, please, see the preprocessing script in our repository[9].

Iteration # Sentence’s evolution # corr.
Orig. sent A ten years old boy go school -
Iteration 1 A ten-years old boy goes school 2
Iteration 2 A ten-year-old boy goes to school 5
Iteration 3 A ten-year-old boy goes to school. 6
Table 3: Example of iterative correction process where GEC tagging system is sequentially applied at each iteration. Cumulative number of corrections is given for each iteration. Corrections are in bold.

4. Tagging Model Architecture

Our GEC sequence tagging model is an encoder made up of pretrained BERT-like transformer stacked with two linear layers with softmax layers on the top. We always use cased pretrained transformers in their Base configurations. Tokenization depends on the particular transformer's design]]: BPE (Sennrich et al., 2016b) is used in RoBERTa, WordPiece (Schuster and Nakajima, 2012) in BERT and SentencePiece (Kudo and Richardson, 2018) in XLNet. To process the information at the token-level, we take the first subword per token from the encoder's representation, which is then forwarded to subsequent linear layers, which are responsible for error detection and error tagging, respectively

5. Iterative Sequence Tagging Approach

To correct the text, for each input token $x_i,\; 1 \leq i \leq N $from the source sequence $\left(x_1 \ldots x_N \right)$, we predict the tag-encoded token-level transformation $T\left(x_i\right)$ described in Section 3. These predicted tag-encoded transformations are then applied to the sentence to get the modified sentence.

Since some corrections in a sentence may depend on others, applying GEC sequence tagger only once may not be enough to fully correct the sentence. Therefore, we use the iterative correction approach from (Awasthi et al., 2019): we use the GEC sequence tagger to tag the now modified sequence, and apply the corresponding transformations on the new tags, which changes the sentence further (see an example in Table 3). Usually, the number of corrections decreases with each successive iteration, and most of the corrections are done during the first two iterations (Table 4). Limiting the number of iterations speeds up the overall pipeline while trading off qualitative performance.

Iteration# P R F0.5 # corr.
Iteration 1 72.3 38.6 61.5 787
Iteration 2 73.7 41.1 63.6 934
Iteration 3 74.0 41.5 64.0 956
Iteration 4 73.9 41.5 64.0 958
Table 4: Cumulative number of corrections and corresponding scores on CoNLL-2014 (test) w.r.t. number of iterations for our best single model.

6. Experiments

Training Stages

We have 3 training stages (details of data usage are in Table1):

 : I Pre-training on synthetic errorful sentences as in (Awasthi et al., 2019).

II Fine-tuning on errorful-only sentences.
III Fine-tuning on subset of errorful and error-free sentences as in (Kiyono et al., 2019).

We found that having two fine-tuning stages with and without error-free sentences is crucial for performance (Table 5).

Training stage # CoNLL-2014 (test) BEA-2019 (dev)
P R F0.5 P R F0.5
Stage I. 55.4 35.9 49.9 37.0 23.6 33.2
Stage II. 64.4 46.3 59.7 46.4 37.9 44.4
Stage III. 66.7 49.9 62.5 52.6 43.0 50.3
Inf. tweaks 77.5 40.2 65.3 66.0 33.8 55.5
Table 5: Performance of GECToR (XLNet) after each training stage and inference tweaks.

All our models were trained by Adam optimizer (Kingma and Ba, 2015) with default hyperparameters. Early stopping was used; stopping criteria was 3 epochs of 10K updates each without improvement. We set batch size=256 for pre-training stage (20 epochs) and batch size=128 for fine-tuning stages II and III (2-3 epochs each). We also observed that freezing the encoder's weights for the first 2 epochs on training stages I-II and using a batch size greater than 64 improves the convergence and leads to better GEC performance.

Encoders from Pretrained Transformers

We fine-tuned BERT (Devlin et al., 2019), RoBERTa (Liu et al., 2019), GPT-2 (Radford et al., 2019), XLNet (Yang et al., 2019), and ALBERT (Lan et al., 2019) with the same hyperparameters setup. We also added LSTM with randomly initialized embeddings (dim = 300) as a baseline. As follows from Table 6, encoders from fine-tuned Transformers significantly outperform LSTMs. BERT, RoBERTa and XLNet encoders perform better than GPT-2 and ALBERT, so we used them only in our next experiments. All models were trained out-of-the-box[10] which seems to not work well for GPT-2. We hypothesize that encoders from Transformers which were pretrained as a part of the entire encoder-decoder pipeline are less useful for GECToR.

Encoder CoNLL-2014 (test) BEA-2019 (dev)
P R F0.5 P R F0.5
LSTM 51.6 15.3 35.0 - - -
ALBERT 59.5 31.0 50.3 43.8 22.3 36.7
BERT 65.6 36.9 56.8 48.3 29.0 42.6
GPT-2 61.0 6.3 22.2 44.5 5.0 17.2
RoBERTa 67.5 38.3 58.6 50.3 30.5 44.5
XLNet 64.6 42.6 58.5 47.1 34.2 43.8
Table 6: Varying encoders from pretrained Transformers in our sequence labeling system. Training was done on data from training stage II only.

Tweaking the Inference

We forced the model to perform more precise corrections by introducing two inference hyperparameters (see Appendix, Table 11), hyperparameter values were found by random search on BEA-dev.

First, we added a permanent positive confidence bias to the probability of $KEEP tag which is responsible for not changing the source token. Second, we added a sentence-level minimum error probability threshold for the output of the error detection layer. This increased precision by trading off recall and achieved better F0.5 scores (Table 5).

Finally, our best single-model, GECToR (XLNet) achieves F0.5 = 65.3 on CoNLL-2014 (test) and F0.5 = 72.4 on BEA-2019 (test). Best ensemble model, GECToR (BERT + RoBERTa + XLNet) where we simply average output probabilities from 3 single models achieves F0.5 = 66.5 on CoNLL2014 (test) and F0.5 = 73.6 on BEA-2019 (test), correspondingly (Table 7).

GEC system Ens. CoNLL-2014 (test) BEA-2019 (test)
P R F0.5 P R F0.5
Zhao et al.(2019)   67.7 40.6 59.8 − − −
Awasthi et al. (2019)   66.1 43.0 59.7 − − −
Kiyono et al. (2019)   67.9 44.1 61.3 65.5 59.4 64.2
Zhao et al. (2019) ✓ 74.1 36.3 61.3 − − −
Awasthi et al. (2019) ✓ 68.3 43.2 61.2 − − −
Kiyono et al. (2019) ✓ 72.4 46.1 65.0 74.7 56.7 70.2
Kantor et al. (2019) ✓ − − − 78.3 58.0 73.2
GECToR (BERT)   72.1 42.0 63.0 71.5 55.7 67.6
GECToR (RoBERTa)   73.9 41.5 64.0 77.2 55.1 71.5
GECToR (XLNet)   77.5 40.1 65.3 79.2 53.9 72.4
GECToR(RoBERTa+ XLNet) ✓ 76.6 42.3 66.0 79.4 57.2 73.7
GECToR(BERT+RoBERTa+XLNet) ✓ 78.2 41.5 66.5 78.9 58.2 73.6
Table 7: Comparison of single models and ensembles. The M2 score for CoNLL-2014 (test) and ERRANT for the BEA-2019 (test) are reported. In ensembles we simply average output probabilities from single models.

Speed Comparison

We measured the model's average inference time on NVIDIA Tesla V100 on batch size 128. For sequence tagging we don't need to predict corrections one-by-one as in autoregressive transformer decoders, so inference is naturally parallelizable and therefore runs many times faster. Our sequence tagger's inference speed is up to 10 times as fast as the state-of-the-art Transformer from Zhao et al. (2019), beam size=12 (Table 8).

GEC system Time (sec)
Transformer-NMT, beam size = 12 4.35
Transformer-NMT, beam size = 4 1.25
Transformer-NMT, beam size = 1 0.71
GECToR (XLNet), 5 iterations 0.40
GECToR (XLNet), 1 iteration 0.20
Table 8: Inference time for NVIDIA Tesla V100 on CoNLL-2014 (test), single model, batch size=128.

7. Conclusions

We show that a faster, simpler, and more efficient GEC system can be developed using a sequence tagging approach, an encoder from a pretrained Transformer, custom transformations and 3-stage training.

Our best single-model/ensemble GEC tagger achieves an F0.5 of 65.3/66.5 on CoNLL-2014 (test) and F0.5 of 72.4/73.6 on BEA-2019 (test). We achieve state-of-the-art results for the GEC task with an inference speed up to 10 times as fast as Transformer-based seq2seq systems

8. Acknowledgements

This research was supported by Grammarly. We thank our colleagues Vipul Raheja, Oleksiy Syvokon, Andrey Gryshchuk and our ex-colleague Maria Nadejde who provided insight and expertise that greatly helped to make this paper better. We would also like to show our gratitude to Abhijeet Awasthi and Roman Grundkiewicz for their support in providing data and answering related questions. We also thank 3 anonymous reviewers for their contribution.

Footnotes

  1. ↑ https://github.com/grammarly/gector
  2. ↑ http://nlpprogress.com/english/grammatical_error_correction.html
  3. ↑ https://github.com/awasthiabhijeet/PIE/tree/master/errorify
  4. ↑ https://www.comp.nus.edu.sg/~nlp/corpora.html
  5. ↑ https://sites.google.com/site/naistlang8corpora
  6. ↑ https://ilexir.co.uk/datasets/index.html
  7. ↑ https://www.cl.cam.ac.uk/research/nl/bea2019st/data/wi+locness_v2.1.bea19.tar.gz
  8. ↑ https://github.com/gutfeeling/word_forms/blob/master/word_forms/en-verbs.txt
  9. ↑ https://github.com/grammarly/gector
  10. ↑ https://huggingface.co/transformers/

A. Appendix

id Core transformation Transformation suffix Tag Example
basic-1 KEEP ∅ $KEEP . . . many people want to travel during the summer . . .
basic-2 DELETE ∅ $DELETE . . . not sure if you are {you ⇒ ∅} gifting . . .
basic-3 REPLACE a $REPLACE_a . . . the bride wears {the ⇒ a} white dress . . .
. . . . . . . . . . . . . . .
basic-3804 REPLACE cause $REPLACE_cause . . . hope it does not {make ⇒ cause} any trouble . . .
basic-3805 APPEND for $APPEND_for . . . he is {waiting ⇒ waiting for} your reply . . .
. . . . . . . . . . . . . . .
basic-4971 APPEND know $APPEND_know . . . I {don’t ⇒ don’t know} which to choose. . .
g-1 CASE CAPITAL $CASE_CAPITAL . . . surveillance is on the {internet ⇒ Internet} . . .
g-2 CASE CAPITAL_1 $CASE_CAPITAL_1 . . . I want to buy an {iphone ⇒ iPhone} . . .
g-3 CASE LOWER $CASE_LOWER . . . advancement in {Medical ⇒ medical} technology . . .
g-4 CASE UPPER $CASE_UPPER . . . the {it ⇒ IT} department is concerned that. . .
g-5 MERGE SPACE $MERGE_SPACE . . . insert a special kind of gene {in to ⇒ into} the cell . . .
g-6 MERGE HYPHEN $MERGE_HYPHEN . . . and needs {in depth ⇒ in-depth} search . . .
g-7 SPLIT HYPHEN $SPLIT_HYPHEN . . . support us for a {long-run ⇒ long run} . . .
g-8 NOUN_NUMBER SINGULAR $NOUN_NUMBER_SINGULAR . . . a place to live for their {citizen ⇒ citizens}
g-9 NOUN_NUMBER PLURAL $NOUN_NUMBER_PLURAL . . . carrier of this {diseases ⇒ disease} . . .
g-10 VERB FORM VB_VBZ $VERB_FORM_VB_VBZ . . . going through this {make ⇒ makes} me feel . . .
g-11 VERB FORM VB_VBN $VERB_FORM_VB_VBN . . . to discuss what {happen ⇒ happened} in fall . . .
g-12 VERB FORM VB_VBD $VERB_FORM_VB_VBD . . . she sighed and {draw ⇒ drew} her . . .
g-13 VERB FORM VB_VBG $VERB_FORM_VB_VBG . . . shown success in {prevent ⇒ preventing} such . . .
g-14 VERB FORM VB_VBZ $VERB_FORM_VB_VBZ . . . a small percentage of people {goes ⇒ go} by bike . . .
g-15 VERB FORM VBZ_VBN $VERB_FORM_VBZ_VBN . . . development has {pushes ⇒ pushed} countries to . . .
g-16 VERB FORM VBZ_VBD $VERB_FORM_VBZ_VBD . . . he {drinks ⇒ drank} a lot of beer last night . . .
g-17 VERB FORM VBZ_VBG $VERB_FORM_VBZ_VBG . . . couldn’t stop {thinks ⇒ thinking} about it . . .
g-18 VERB FORM VBN_VB $VERB_FORM_VBN_ VB . . . going to {depended ⇒ depend} on who is hiring . . .
g-19 VERB FORM VBN_VBZ $VERB_FORM_VBN_ VBZ . . . yet he goes and {eaten ⇒ eats} more melons . . .
g-20 VERB FORM VBN_VBD $VERB_FORM_VBN_ VBD . . . he {driven ⇒ drove} to the bus stop and . . .
g-21 VERB FORM VBN_VBG $VERB_FORM_VBN_ VBG . . . don’t want you fainting and {broken ⇒ breaking} . . .
g-22 VERB FORM VBD_VB $VERB_FORM_VBD_ VB . . . each of these items will {fell ⇒ fall} in price . . .
g-23 VERB FORM VBD_VBZ $VERB_FORM_VBD_ VBZ . . . the lake {froze ⇒ freezes} every year . . .
g-24 VERB FORM VBD_VBN $VERB_FORM_VBD_VBN . . . he has been went {went ⇒ gone} since last week . . .
g-25 VERB FORM VBD_VBG $VERB_FORM_VBD_VBG . . . talked her into {gave ⇒ giving} me the whole day . . .
g-26 VERB FORM VBG_VB $VERB_FORM_VBG_VB . . . free time, I just {enjoying ⇒ enjoy} being outdoors . . .
g-27 VERB FORM VBG_VBZ $VERB_FORM_VBG_ VBZ . . . there still {existing ⇒ exists} many inevitable factors . . .
g-28 VERB FORM VBG_VBN $VERB_FORM_VBG_VBN . . . people are afraid of being {tracking ⇒ tracked} . . .
g-29 VERB FORM VBG_VBD $VERB_FORM_VBG_ VBD . . . there was no {mistook ⇒ mistaking} his sincerity . . .
Table 9: List of token-level transformations (section 3). We denote a tag which defines a token-level transformation as concatenation of two parts: a core transformation and a transformation suffix.

Training stage # CoNLL-2014 (test) BEA-2019 (dev)
P R F0.5 P R F0.5
Stage I. 57.8 33.0 50.2 40.8 22.1 34.9
Stage II. 68.1 42.6 60.8 51.6 33.8 46.7
Stage III. 68.8 47.1 63.0 54.2 41.0 50.9
Inf. tweaks 73.9 41.5 64.0 62.3 35.1 54.0
Table 10: Performance of GECToR (RoBERTa) after each training stage and inference tweaks. Results are given in addition to results for our best single model, GECToR (XLNet) which are given in Table 5.

System name Confidence bias Minimum error probability
GECToR (BERT) 0.10 0.41
GECToR (RoBERTa) 0.20 0.50
GECToR (XLNet) 0.35 0.66
GECToR (RoBERTa + XLNet) 0.24 0.45
GECToR (BERT + RoBERTa + XLNet) 0.16 0.40
Table 11: Inference tweaking values which were found by random search on BEA-dev.

References

2019a

2019b

2019c

2019d

2019e

2019f

2019g

2019h

2019i

2019j

2019k

2019l

2018

2017a

2017b

2017c

  • (Yuan, 2017) ⇒ Zheng Yuan (2017). “Grammatical Error Correction in Non-Native English". Technical report, University of Cambridge, Computer Laboratory.

2016a

2016b

2015

2014

2013

2012a

2012b

2012c

2011

2003

BibTeX

@inproceedings{2020_GECToRGrammaticalErrorCorrectio,
  author    = {Kostiantyn Omelianchuk and
               Vitaliy Atrasevych and
               Artem N. Chernodub and
               Oleksandr Skurzhanskyi},
  editor    = {Jill Burstein and
               Ekaterina Kochmar and
               Claudia Leacock and
               Nitin Madnani and
               Ildiko Pilan and
               Helen Yannakoudakis and
               Torsten Zesch},
  title     = {GECToR - Grammatical Error Correction: Tag, Not Rewrite},
  booktitle = {Proceedings of the Fifteenth Workshop on Innovative Use of NLP for
               Building Educational Applications (BEA@ACL 2020)},
  pages     = {163--170},
  publisher = {Association for Computational Linguistics},
  year      = {2020},
  url       = {https://doi.org/10.18653/v1/2020.bea-1.16},
  doi       = {10.18653/v1/2020.bea-1.16},
}


 AuthorvolumeDate ValuetitletypejournaltitleUrldoinoteyear
2020 GECToRGrammaticalErrorCorrectioKostiantyn Omelianchuk
Vitaliy Atrasevych
Artem N. Chernodub
Oleksandr Skurzhanskyi		GECToR - Grammatical Error Correction: Tag, Not Rewrite2015
2020
Retrieved from "http://www.gabormelli.com/RKB/index.php?title=2020_GECToRGrammaticalErrorCorrectio&oldid=885328"