FlexNER: A Flexible LSTM-CNN Stack Framework for Named Entity Recognition

Hongyin Zhu; Wenpeng Hu; Yi Zeng

REVIEW 4 major objections 5 minor 29 references

FlexNER: A Flexible LSTM-CNN Stack Framework for Named Entity Recognition

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Replacing a named entity in a training sentence with another entity of the same type, and fusing a baseline sub-network with one trained on such swaps, improves NER across five languages and two biomedical corpora without external…

desk verdict A credible NER augmentation scheme (in-dataset entity swapping plus bilateral fusion) whose English results are consistent, but the abstract's 'generally improve' overreaches: multilingual evidence is in an appendix and the fusion doesn't always help. read the letter →

arxiv 1908.05009 v1 pith:M5UII47B submitted 2019-08-14 cs.CL

Hongyin Zhu , Wenpeng Hu , Yi Zeng This is my paper · ORCID

classification cs.CL

keywords namedentityrecognitiondataaugmentationentity-contextdiversitybilateralnetworkLSTM-CNNmultilingualNERbiomedicalsequencelabeling

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that a named-entity recognizer gets better when the training set is enlarged by swapping an entity for another entity of the same type drawn from the same data, and when the original and swapped views are learned by two sub-networks whose outputs are fused. The paper calls the first idea entity-context diversity enhancement: it needs no external knowledge base or annotated resources beyond the entities already in the training set. The bilateral network trains a baseline sub-network on the original sentences, then an augmented sub-network on the dynamically swapped sentences, then retrains the fusion layer on original data. Reported results cover English, German, Spanish, Dutch, Chinese and two biomedical corpora, and the paper's repeated claim is that this strategy generally improves model performance. The reason it matters is portability: if correct, a domain or language needs no extra resources to get a better model.

What carries the argument

Two mechanisms do the work. The first is entity-context augmentation: sentence-centric augmentation (SCA) picks a sentence and swaps each entity slot with a different entity of the same type with probability p (0.5 to 0.9), while entity-centric augmentation (ECA) samples an entity and a sentence that contains that entity type and then performs the crossover. The second is the bilateral architecture: a baseline sub-network and an augmented sub-network, each optionally built from Bi-LSTM, CNN, or stack components, that are separately trained with frozen counterparts and finally fused by retraining a shared conditional random field on concatenated outputs. The separate training is essential because it lets each sub-network keep its own function; the paper reports that joint training on the same input produced worse results.

What would settle it

Take a biomedical corpus in which the same entity type splits into mutually exclusive subgroups (e.g., drug names that trigger opposing treatment decisions), run the bilateral model at the paper's highest swap rate p=0.9, and check whether its F1 stays above the no-augmentation baseline. The paper's own chemicals-corpus result—Augment 69.98 versus Bilateral 69.79—already shows fusion can hurt, so a dataset where even the baseline wins would refute the 'generally improve' claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that the missing signal in fixed training sets is entity-context diversity, and that a second sub-network can be taught that diversity through in-dataset entity swapping. The paper's strongest evidence is the comparison of Baseline+Augment against the baseline across all English network combinations in Table 1, with the best configuration reaching 91.47 versus the baseline's 91.00, and the improvement on the biomedical cell-line corpus from 78.78 to 84.86. In its own words, the strategy 'can generally improve model performance on different datasets.' The paper also reports that fusing two baseline sub-networks (Baseline+Baseline) changes almost nothing, which isolates the augmentation as the active ingredient, and that separate training of the two sub-networks beats joint training.

Load-bearing premise

The load-bearing premise is same-type entity interchangeability: any two entities with the same label are interchangeable in context, so a swapped sentence like 'Germany imported 47600 sheep from national tennis centre' still trains the model meaningfully; the paper explicitly calls quality control critical and defers it to future work.

Editorial extensions

If this is right

The same augmentation recipe transfers across five languages and two biomedical corpora without per-language or per-domain feature engineering.
The cell-line corpus result, a 6.08-point F1 gain over the baseline, suggests low-resource biomedical datasets benefit most, an important case because expert annotation is expensive.
The data augmentation operations run only during training, so users never need to store or ship an expanded corpus.
At least 64 layer-stack and sub-network combinations are possible, letting the framework adapt to different data sizes and input types without a new design.
The separate training schedule, which lets each sub-network keep a distinct function, is why the bilateral model beats joint training in the paper's experiments.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: Beyond the paper's experiments, the same in-dataset swap idea should apply to other entity-annotated sequence labeling tasks such as slot filling or event-argument extraction, where type constraints play a similar role.
Editorial inference: The chemicals-corpus result, where augmentation alone (69.98) beats the bilateral fusion (69.79), suggests that noisy swaps can dilute the fused representation; a testable extension is a confidence gate that down-weights implausible swapped sentences before they reach the augmented sub-network.
Editorial inference: The hand-picked swap probability p (0.5 to 0.9) hints that the optimal noise level varies by dataset, so an adaptive schedule that raises p while held-out F1 improves could remove per-dataset tuning.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a solid, modest empirical paper on NER data augmentation. The core idea—swap entities with same-type entities from the training set to increase entity-context diversity, then fuse a baseline and an augmented sub-network—is simple and resource-free, and the English results are credible. The abstract's 'generally improve' overclaims: the multilingual evidence lives in an appendix we haven't seen, and the fusion doesn't always help (SCAI). Worth a serious referee, but it needs revision.

The genuinely new piece is the in-dataset entity swapping (SCA and ECA) plus the bilateral training scheme. It is a clean alternative to distant supervision or external KBs, and it can be dropped into any LSTM-CNN-CRF pipeline. The English table is internally consistent: all eight Baseline+Augment combinations beat the baseline, by 0.2–0.5 F1, and the runs were repeated three times with standard deviations. The GELLUS gain is striking: 78.78 to 84.86 F1. The paper is also honest about the main weakness: random same-type swaps can produce nonsense, and quality control is deferred.

The soft spots are proportionate. The headline 'generally improve' is not established: Table 2 shows the bilateral model on SCAI chemicals (69.79) doing worse than the Augment-only subnetwork (69.98), so the fusion is not universally beneficial. In English, the Augment-only model is worse than Baseline in two of six configurations, meaning the augmentation alone is not consistently helpful—the consistent improvement comes only in the fused model, which is a specific architectural choice rather than a demonstrated general principle. The multilingual results are relegated to an appendix, and the augmentation probability p is hand-tuned per dataset (0.5 to 0.9), which is hyperparameter tuning, not a single universal recipe. There is no control for mere oversampling or for corrupted labels, so the active ingredient isn't fully isolated.

Who this is for: anyone working on low-resource NER or data augmentation for sequence labeling. It's a useful empirical data point, not a new state of the art. I'd send it to peer review with the expectation of revision: move the multilingual results into the main text (or honestly re-scope the claim), add error bars to Table 2, report the p values per dataset, and ideally run a corrupted-label control. The central English results support the method's promise; the claims just need to match the evidence.

Referee Report

4 major / 5 minor

Summary. The paper proposes FlexNER, a named entity recognition framework that combines LSTM/CNN stack sub-networks in a bilateral architecture. Its main novelty is a data augmentation scheme that swaps same-type entities inside the training set using only in-dataset labels, without external resources. The authors report experiments on English CoNLL-2003 in Table 1 and on two biomedical corpora (SCAI chemicals and GELLUS cell lines) in Table 2, and they claim in the abstract that the strategy 'can generally improve model performance on different datasets'. The manuscript also mentions experiments in German, Spanish, Dutch, and Chinese, but those results are deferred to a supplementary appendix.

Significance. If the central claim holds, the paper offers a simple, resource-free augmentation method and a flexible network-combination framework that could transfer across languages and domains. The contribution is strengthened by the use of held-out test sets, repeated runs (three times) on the English data, and a publicly cited code repository; there is no fitted-constant circularity because augmentation uses only training-set labels and evaluation is empirical. The English bilateral (Baseline+Augment) model improves over Baseline in all six reported network configurations, and the GELLUS result is substantial (84.86 vs. 78.78). However, the generality claim is currently only weakly evidenced in the main text, and the mechanism behind the gains is not isolated from confounds such as extra training data or the two-network training schedule.

major comments (4)

[Abstract and Section 4.2] The abstract's load-bearing claim that the strategy 'can generally improve model performance on different datasets' is not supported by the main text for multilingual data. Section 4.2 lists CoNLL-2002/2003 German, Spanish, Dutch, and Chinese datasets, but Table 1 reports only English results, and no multilingual numbers appear in the main text. The readers are told to consult a supplementary appendix that was not part of this review. The authors should either include the multilingual results in the paper or explicitly restrict the claim to the datasets actually reported.
[Section 3.3 and Table 1] The proposed bilateral fusion is not isolated as the source of improvement. Table 1 shows that Augment alone is worse than Baseline in two of the six configurations (90.87 vs. 91.00 for the character-LSTM baseline, and 89.67 vs. 90.15 for stack-b), and Baseline+Baseline is essentially flat. The only consistently improving row is Baseline+Augment, which differs from Baseline in two ways: the augmented training data and the two-sub-network training schedule. A controlled comparison is needed, for example, training a single network on the same amount of augmented data, training on oversampled original data, or training with corrupted-label augmented data. Without such controls, the paper's claim that entity-context diversity is the active ingredient is not yet established.
[Section 3.2 and Table 2] The authors' own quality-control caveat undermines the label-preserving assumption of same-type entity swapping. Section 3.2 acknowledges that random swaps can produce implausible examples such as 'Germany imported 47600 sheep from national tennis centre' and defers quality control to future work. This is not merely a presentation concern: in Table 2, the SCAI bilateral model (69.79) is worse than the Augment-only sub-network (69.98), which is consistent with noisy augmented data hurting the fused model. The paper should report the per-dataset augmentation probability p, the amount of augmented data used, and ideally a sensitivity analysis or a filtering mechanism for implausible swaps.
[Table 2] Table 2 reports only point estimates for the biomedical corpora, with no variance or number of runs, unlike Table 1, which reports three-run means and standard deviations. The SCAI difference between Bilateral (69.79) and Augment (69.98) is small, and without error bars it is unclear whether the bilateral architecture helps or hurts on this corpus. This matters because the paper's generality claim includes biomedical domains.

minor comments (5)

[Section 4.3] In the sentence about GELLUS, 'F1 sore' should be 'F1 score'.
[Section 3.2] Step 1 of the SCA algorithm says 'we needs to convert label sequence into complete entities'; the verb should be 'need'.
[Section 3.2] Step 3 of SCA writes 'a(i,k)∼ Binomial(...)'; the distribution name appears to be a typo for 'Binomial' or a different sampling scheme, and should be clarified.
[Section 4.1 and Table 2] The paper does not state the evaluation metric for Table 2 or the exact training/validation split for SCAI and GELLUS; adding this information would improve reproducibility.
[Section 3.4] The separate training procedure is described in text but no pseudocode or algorithm box is provided; a concise algorithmic listing would make the freezing/masking steps easier to follow.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: FlexNER's claims are empirical, evaluated on held-out test sets, with no fitted parameter renamed as a prediction.

full rationale

FlexNER is an empirical systems paper rather than a derivation. The central proposal is an in-dataset data augmentation procedure (same-type entity swapping) plus a two-subnetwork training schedule. Augmented sentences are generated from training-set entity glossaries and evaluated on held-out CoNLL, SCAI, and GELLUS test splits, so no fitted quantity is renamed as a prediction. The per-dataset replacement probability p is hyperparameter tuning, not a loaded parameter that forces the reported F1 differences. The paper even reports configurations where augmentation does not help (English character-LSTM Augment 90.87 vs Baseline 91.00; SCAI Bilateral 69.79 vs Augment-only 69.98), which shows the results are not forced by construction. The 'context pattern entailment' premise is an empirical assumption about same-type entity interchangeability, and the paper explicitly acknowledges its failure mode ("Quality control is critical when faced with data in a specific domain and is reserved for future work"), but acknowledging an assumption is not circular reasoning. Self-citations (e.g., refs. 7, 8, 9, 28, 29) appear only in related-work or future-work contexts and do not supply load-bearing evidence for the main empirical claims. No equation reduces to an input definition, no uniqueness theorem is imported from the authors' prior work, no ansatz is smuggled in via citation, and no known result is merely renamed. The contribution is a set of controlled empirical comparisons against reproduced baselines and external systems, so the derivation chain is self-contained with respect to circularity.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central result rests on one domain assumption (same-class entity interchangeability) that the paper itself flags as imperfect, plus standard ML machinery used as-is. The only fitted input is the augmentation probability p, tuned per dataset over 0.5 to 0.9 without a stated criterion; training hyperparameters are not reported in the main text. The bilateral network and augmentation schemes are architecture choices, not invented entities with independent falsifiable handles, so invented_entities is empty.

free parameters (2)

SCA augmentation probability p = 0.5 to 0.9, chosen per dataset
The Bernoulli lighting probability for entity slots is hand-picked in the range 0.5 to 0.9 according to each dataset (§3.2 step 2). The exact value per dataset is not given in the main text, so the augmentation strength is an unmeasured free parameter of the central result.
training hyperparameters (learning rate, hidden dims, epochs) = not reported in main text
The paper does not enumerate learning rate, hidden sizes, or epoch counts in the main text; the appendix is referenced but not provided here, so these settings are unspecified inputs to the reported numbers.

assumptions (3)

domain assumption Same-class entity interchangeability preserves label validity (context pattern entailment)
§3.2 motivates swapping entities of the same category, and §1 frames it as 'context pattern entailment'. This is the load-bearing premise for why augmentation helps; the paper acknowledges it fails sometimes ('national tennis centre' example).
standard math CRF/Viterbi and LSTM-CNN standard machinery
Equation (1) uses the standard CRF; backpropagation through LSTM and CNN stacks is standard. No new mathematical claims are made.
domain assumption Augmentation noise does not systematically bias the fused representation
The bilateral network relies on the augmented sub-network being 'more generalized' (§3.3), a claim asserted without direct evidence; on SCAI the bilateral result (69.79) is below augmentation alone (69.98), suggesting the fusion is not universally beneficial.

how reviews work

0 comments

Cite this review

Pith. "Pith review of FlexNER: A Flexible LSTM-CNN Stack Framework for Named Entity Recognition." pith.science (2026). https://pith.science/paper/M5UII47B

@misc{pith2026190805009,
  author       = {Pith},
  title        = {Pith review of: FlexNER: A Flexible LSTM-CNN Stack Framework for Named Entity Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M5UII47B}},
  note         = {Machine review of arXiv:1908.05009}
}

read the original abstract

Named entity recognition (NER) is a foundational technology for information extraction. This paper presents a flexible NER framework compatible with different languages and domains. Inspired by the idea of distant supervision (DS), this paper enhances the representation by increasing the entity-context diversity without relying on external resources. We choose different layer stacks and sub-network combinations to construct the bilateral networks. This strategy can generally improve model performance on different datasets. We conduct experiments on five languages, such as English, German, Spanish, Dutch and Chinese, and biomedical fields, such as identifying the chemicals and gene/protein terms from scientific works. Experimental results demonstrate the good performance of this framework.

Figures

Figures reproduced from arXiv: 1908.05009 by the authors.

**Figure 2.** An overview of bilateral architecture 2. Build the categorical sentence set S = {S1, S2, ..., Sc}. The main idea is to classify the training samples according to the type mention of entities. Let e(i,1) ∈ Ei and ei,1 ∈ sj , then sj ∈ Si . In [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 20 canonical work pages

[1]

arXiv preprint arXiv:1511.08308 (2015)

Chiu, J.P., Nichols, E.: Named entity recognition with bidirectional lstm-cnns. arXiv preprint arXiv:1511.08308 (2015)

arXiv 2015
[2]

arXiv preprint arXiv:1809.08370 (2018)

Clark, K., Luong, M.T., Manning, C.D., Le, Q.V.: Semi-supervised sequence mod- eling with cross-view training. arXiv preprint arXiv:1809.08370 (2018)

arXiv 2018
[3]

Journal of Machine Learning Research 12(Aug), 2493–2537 (2011)

Collobert, R., Weston, J., Bottou, L., Karlen, M., Kavukcuoglu, K., Kuksa, P.: Natural language processing (almost) from scratch. Journal of Machine Learning Research 12(Aug), 2493–2537 (2011)

work page 2011
[4]

arXiv preprint arXiv:1810.04805 (2018)

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidirec- tional transformers for language understanding. arXiv preprint arXiv:1810.04805 (2018)

arXiv 2018
[5]

arXiv preprint arXiv:1702.05538 (2017)

DeVries, T., Taylor, G.W.: Dataset augmentation in feature space. arXiv preprint arXiv:1702.05538 (2017)

arXiv 2017
[6]

Bioinformatics 33(14), i37–i48 (2017)

Habibi, M., Weber, L., Neves, M., Wiegandt, D.L., Leser, U.: Deep learning with word embeddings improves biomedical named entity recognition. Bioinformatics 33(14), i37–i48 (2017)

work page 2017
[7]

In: Proceedings of IJCAI 2019 (2019) 10 Zhu et al

Hu, W., Chan, Z., Liu, B., Zhao, D., Ma, J., Yan, R.: Gsn: A graph-structured network for multi-party dialogues. In: Proceedings of IJCAI 2019 (2019) 10 Zhu et al

work page 2019
[8]

In: Proceedings of ICLR 2019 (2019)

Hu, W., Lin, Z., Liu, B., Tao, C., Tao, Z., Ma, J., Zhao, D., Yan, R.: Overcoming catastrophic forgetting for continual learning via model adaptation. In: Proceedings of ICLR 2019 (2019)

work page 2019

Show all 29 references

[9]

In: Proceedings of COLING 2016 (2016)

Hu, W., Zhang, J., Zheng, N.: Diﬀerent contexts lead to diﬀerent word embeddings. In: Proceedings of COLING 2016 (2016)

2016
[10]

Journal of cheminformatics 3(1), 41 (2011)

Jessop, D.M., Adams, S.E., Willighagen, E.L., Hawizy, L., Murray-Rust, P.: Os- car4: a ﬂexible architecture for chemical text-mining. Journal of cheminformatics 3(1), 41 (2011)

2011
[11]

In: Proceedings of NeuIPS 2014 (2014)

Jiang, L., Meng, D., Yu, S.I., Lan, Z., Shan, S., Hauptmann, A.: Self-paced learning with diversity. In: Proceedings of NeuIPS 2014 (2014)

2014
[12]

Bioinformatics 32(2), 276–282 (2015)

Kaewphan, S., Van Landeghem, S., Ohta, T., Van de Peer, Y., Ginter, F., Pyysalo, S.: Cell line name recognition in support of the identiﬁcation of synthetic lethality in cancer from text. Bioinformatics 32(2), 276–282 (2015)

2015
[13]

Laﬀerty, J., McCallum, A., Pereira, F.C.: Conditional random ﬁelds: Probabilistic models for segmenting and labeling sequence data (2001)

2001
[14]

arXiv preprint arXiv:1603.01360 (2016)

Lample, G., Ballesteros, M., Subramanian, S., Kawakami, K., Dyer, C.: Neural architectures for named entity recognition. arXiv preprint arXiv:1603.01360 (2016)

2016 arXiv
[15]

arXiv preprint arXiv:1603.01354 (2016)

Ma, X., Hovy, E.: End-to-end sequence labeling via bi-directional lstm-cnns-crf. arXiv preprint arXiv:1603.01354 (2016)

2016 arXiv
[16]

In: Proceedings of ACL 2009 (2009)

Mintz, M., Bills, S., Snow, R., Jurafsky, D.: Distant supervision for relation ex- traction without labeled data. In: Proceedings of ACL 2009 (2009)

2009
[17]

arXiv preprint arXiv:1707.02459 (2017)

Ni, J., Florian, R.: Improving multilingual named entity recognition with wikipedia entity type mapping. arXiv preprint arXiv:1707.02459 (2017)

2017 arXiv
[18]

arXiv preprint arXiv:1802.05365 (2018)

Peters, M.E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., Zettlemoyer, L.: Deep contextualized word representations. arXiv preprint arXiv:1802.05365 (2018)

2018 arXiv
[19]

In: Proceedings of CoNLL 2009

Ratinov, L., Roth, D.: Design challenges and misconceptions in named entity recog- nition. In: Proceedings of CoNLL 2009. pp. 147–155 (2009)

2009
[20]

Bioinformatics 28(12), 1633–1640 (2012)

Rockt¨ aschel, T., Weidlich, M., Leser, U.: Chemspot: a hybrid system for chemical named entity recognition. Bioinformatics 28(12), 1633–1640 (2012)

2012
[21]

In: Proceedings of IJCNLP 2017 (2017)

Saito, I., Suzuki, J., Nishida, K., Sadamitsu, K., Kobashikawa, S., Masumura, R., Matsumoto, Y., Tomita, J.: Improving neural text normalization with data augmentation at character-and morphological levels. In: Proceedings of IJCNLP 2017 (2017)

2017
[22]

In: Proceedings of CoNLL 2002 (2002)

Sang, E.F.T.K.: Introduction to the conll-2002 shared task: Language-independent named entity recognition. In: Proceedings of CoNLL 2002 (2002)

2002
[23]

In: Proceedings of COLING 2014

Shi, B., Zhang, Z., Sun, L., Han, X.: A probabilistic co-bootstrapping method for entity set expansion. In: Proceedings of COLING 2014. pp. 2280–2290 (2014)

2014
[24]

In: Proceedings of HLT-NAACL

Tjong Kim Sang, E.F., De Meulder, F.: Introduction to the conll-2003 shared task: Language-independent named entity recognition. In: Proceedings of HLT-NAACL

2003
[25]

arXiv preprint arXiv:1711.07010 (2017)

Xu, J., Wen, J., Sun, X., Su, Q.: A discourse-level named entity recogni- tion and relation extraction dataset for chinese literature text. arXiv preprint arXiv:1711.07010 (2017)

2017 arXiv
[26]

In: Proceedings COLING 2018 (2018)

Yang, J., Liang, S., Zhang, Y.: Design challenges and misconceptions in neural sequence labeling. In: Proceedings COLING 2018 (2018)

2018
[27]

arXiv preprint arXiv:1703.06345 (2017)

Yang, Z., Salakhutdinov, R., Cohen, W.W.: Transfer learning for sequence tagging with hierarchical recurrent networks. arXiv preprint arXiv:1703.06345 (2017)

2017 arXiv
[28]

arXiv preprint arXiv:1704.06393 (2017) FlexNER 11

Zhou, L., Hu, W., Zhang, J., Zong, C.: Neural system combination for machine translation. arXiv preprint arXiv:1704.06393 (2017) FlexNER 11

2017 arXiv
[29]

In: Proceedings of BI 2016

Zhu, H., Zeng, Y., Wang, D., Xu, B.: Brain knowledge graph analysis based on complex network theory. In: Proceedings of BI 2016. Springer (2016)

2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

arXiv preprint arXiv:1511.08308 (2015)

Chiu, J.P., Nichols, E.: Named entity recognition with bidirectional lstm-cnns. arXiv preprint arXiv:1511.08308 (2015)

arXiv 2015

[2] [2]

arXiv preprint arXiv:1809.08370 (2018)

Clark, K., Luong, M.T., Manning, C.D., Le, Q.V.: Semi-supervised sequence mod- eling with cross-view training. arXiv preprint arXiv:1809.08370 (2018)

arXiv 2018

[3] [3]

Journal of Machine Learning Research 12(Aug), 2493–2537 (2011)

Collobert, R., Weston, J., Bottou, L., Karlen, M., Kavukcuoglu, K., Kuksa, P.: Natural language processing (almost) from scratch. Journal of Machine Learning Research 12(Aug), 2493–2537 (2011)

work page 2011

[4] [4]

arXiv preprint arXiv:1810.04805 (2018)

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidirec- tional transformers for language understanding. arXiv preprint arXiv:1810.04805 (2018)

arXiv 2018

[5] [5]

arXiv preprint arXiv:1702.05538 (2017)

DeVries, T., Taylor, G.W.: Dataset augmentation in feature space. arXiv preprint arXiv:1702.05538 (2017)

arXiv 2017

[6] [6]

Bioinformatics 33(14), i37–i48 (2017)

Habibi, M., Weber, L., Neves, M., Wiegandt, D.L., Leser, U.: Deep learning with word embeddings improves biomedical named entity recognition. Bioinformatics 33(14), i37–i48 (2017)

work page 2017

[7] [7]

In: Proceedings of IJCAI 2019 (2019) 10 Zhu et al

Hu, W., Chan, Z., Liu, B., Zhao, D., Ma, J., Yan, R.: Gsn: A graph-structured network for multi-party dialogues. In: Proceedings of IJCAI 2019 (2019) 10 Zhu et al

work page 2019

[8] [8]

In: Proceedings of ICLR 2019 (2019)

Hu, W., Lin, Z., Liu, B., Tao, C., Tao, Z., Ma, J., Zhao, D., Yan, R.: Overcoming catastrophic forgetting for continual learning via model adaptation. In: Proceedings of ICLR 2019 (2019)

work page 2019