Hierarchical Text Classification with Reinforced Label Assignment

Jiawei Han; Jingjing Tian; Xiang Ren; Yuning Mao

REVIEW 3 major objections 5 minor 1 cited by

Hierarchical Text Classification with Reinforced Label Assignment

T0 review · 3 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 Hierarchical text classification is treated as a Markov decision process in which a learned policy places each object in the label hierarchy and decides when to stop, producing consistent multi-path label sets.

desk verdict Genuinely new RL-for-HTC idea with a learned stop action, but the headline 33.4% Macro-F1 gain over flat classifiers is inflated by untuned flat baselines. read the letter →

arxiv 1908.10419 v1 pith:BBPX25MQ submitted 2019-08-27 cs.IR cs.CLcs.LG

Yuning Mao , Jingjing Tian , Jiawei Han , Xiang Ren This is my paper · ORCID

classification cs.IRcs.CLcs.LG

keywords hierarchicaltextclassificationreinforcementlearninglabelassignmentpolicyMarkovdecisionprocessmulti-labelgradientstopactionMacro-F1

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

The paper tries to show that hierarchical text classification is better solved as a decision process than as a set of independent label predictions. It trains a reinforcement-learned label assignment policy that places each document into the label hierarchy one step at a time, with a dedicated stop action, so that training and inference follow the same traversal and predictions are always consistent sub-trees. Because Macro-F1 weights every label equally, the reported 33.4% average gain over flat classifiers indicates the policy is especially recovering rare, deep labels that flat models ignore. On five datasets and four different neural encoders, the paper reports this gain plus better results than existing hierarchical methods, without requiring post-processing.

What carries the argument

The load-bearing object is the label-assignment Markov decision process and its policy network. At time $t$, the state is formed by concatenating the object embedding $e_d$ with the embedding $l_t$ of the most recently assigned label and projecting the result through two ReLU layers to $s_t$; the action space holds the embeddings of all unvisited children of every label already selected, plus a trainable stop embedding $e_{\mathrm{stop}}$, and the policy is a softmax over the matrix product $A_t s_t$. The reward at each step is the difference between the current and previous example-based F1 scores, so the cumulative reward measures whether the whole assignment improved. A supervised pre-training phase walks only ground-truth paths using local per-parent sigmoid losses and a flat loss over all labels, and the final objective mixes the reinforcement loss with that supervised loss.

What would settle it

Run HiLAP with a state that also aggregates all assigned labels (for example, the mean or attention-weighted sum of their embeddings) on a business-review or gene-function dataset; a clear improvement in Macro-F1 would show the original Markov state is insufficient, while a tie would support it. Separately, retrain the flat baselines with per-label threshold tuning and check whether the 33.4% average gain persists.

Watch

Extended reading notes

Core claim

The central claim is that label dependencies in hierarchical text classification should be modeled by one policy that interacts with the hierarchy directly. Starting at the root, the policy repeatedly selects either an unvisited child of any label already assigned to the object or a learned stop action, and the labels visited when it stops are the prediction. The reward at each step is the change in example-based F1, and the policy is trained with a policy-gradient objective plus a greedy self-critical baseline, so the agent learns to look ahead rather than optimize each placement in isolation. This formulation guarantees label consistency by construction, supports multi-path and non-leaf predictions, and removes the mismatch between hierarchical training and flat inference. The paper further claims that the framework is encoder-agnostic: swapping the base encoder only changes the object embedding, and the same reinforcement setup improves all four tested encoders.

Load-bearing premise

The policy's state contains only the embedding of the most recently assigned label, so the learned decisions implicitly assume that labels assigned earlier do not matter for where to go next or when to stop.

Editorial extensions

If this is right

Predictions are consistent by construction: because the policy only adds children of labels already chosen, contradictory label sets cannot occur and no post-processing is needed; the paper reports 0% inconsistent predictions on RCV1.
Sparse, deep labels benefit most: reported gains concentrate on lower-level labels and labels with few training examples, which flat classifiers tend to ignore.
The framework inherits improvements from any differentiable encoder: better base models should transfer directly to the policy without redesign.
Training and inference use the same traversal, reducing the exposure bias of local top-down classifiers, and the learned stop action removes the need for thresholding or mandatory leaf-node prediction.

Reading between the lines

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

Not tested in the paper: comparing HiLAP against flat classifiers that tune a threshold per label would likely narrow the 33.4% headline gain, since the paper notes such tuned flat baselines are stronger.
If the Markov state is the bottleneck, augmenting it to pool embeddings of all assigned labels rather than only the most recent one should help on datasets with many labels per object, such as gene-function prediction; this is a direct experiment the paper does not run.
The same MDP formulation should transfer beyond text to any domain with a label taxonomy and a differentiable object encoder, such as product categorization or image classification, because the policy consumes only the object embedding and label embeddings.
The stop action effectively learns a data-dependent threshold on when further placement is not worth the F1 reward; reading its probability along a traversal could provide a confidence score and a controllable precision-recall trade-off, which the paper does not explore.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Colleague,

You should know about this one. Mao et al. formulate hierarchical text classification as an MDP where an agent walks the label tree/DAG and decides when to stop, trained with policy gradient and self-critical learning. That is genuinely new as far as the cited literature goes—prior work either made flat predictions, used local classifiers, or applied global constraints without a learned placement policy. The framework is general over base encoders, and the code and data are public. The experiments are extensive: five datasets, four encoders, plus functional genomics. The ablation shows the gain is concentrated on rare, lower-level labels, which is a believable mechanism.

The soft spots, in order of importance. First, the headline number: the average 33.4% Macro-F1 improvement over flat classifiers compares against flat baselines with default thresholds. The paper admits in Appendix B that it does not do per-label threshold tuning because it is "infeasible in our case." That is exactly the regime where Macro-F1 is threshold-dominated. HiLAP's learned stop action is an adaptive threshold, so part of the reported gain is an adaptive-threshold effect, not hierarchy modeling. That does not invalidate the comparison against HMCN and other structured baselines, but the abstract's central quantitative claim is overstated as written. Second, the MDP state is only the object embedding plus the embedding of the most recently assigned label, while the action space contains unvisited children of every assigned label. The paper gives no Markov or sufficiency argument; if the history of assigned labels matters (and it should), the policy is not fully controlling the stated MDP. Empirically it works, but this needs at least an analysis or a controlled ablation. Third, no error bars and some key hyperparameters (lambda, alpha, gamma, max episode steps) are undisclosed, which hurts reproducibility. These are all fixable in a revision.

The central empirical claim—that the method beats HMCN and other HTC baselines by a large margin—looks solid, and the code release gives it credibility. It deserves a serious referee, but I would ask the authors to rerun the flat baselines with threshold calibration, report variance, disclose hyperparameters, and address the state-space issue. I would not take the 33.4% number at face value.

Referee Report

3 major / 5 minor

Summary. The paper proposes HiLAP, a reinforcement-learning framework for hierarchical text classification. Label assignment is cast as an MDP: starting from the root, the agent repeatedly selects either a child of one of the already assigned labels or a special stop action; the reward is the change in example-based F1. The state is a projection of the object embedding and the last assigned label embedding, and the policy is trained with self-critical policy gradient after a supervised pre-training stage that combines flat and local losses. The framework is evaluated on RCV1, NYT, Yelp, FunCat, and GO with TextCNN, HAN, bow-CNN, and a feed-forward base model. The paper reports consistent Macro-F1 gains, including an average 33.4% improvement over flat classifiers, and ablations showing benefits for sparse bottom-level labels and 0% label inconsistency.

Significance. If the results hold, HiLAP is a useful and general recipe for HTC: it is encoder-agnostic, guarantees hierarchy-consistent predictions without post-processing, supports multi-path and non-mandatory-leaf prediction, and the released code and data support reproducibility. The reward-shaping and self-critical training details are standard, and the main value is in demonstrating that a learned stop action and hierarchical exploration can improve sparse-label Macro-F1 across text and non-text domains. However, the headline comparison to flat classifiers is weakened by threshold calibration, and the MDP state definition needs clarification before the central claims can be accepted as stated.

major comments (3)

[Section 4.4, Appendix B] The abstract's headline claim of a 33.4% average Macro-F1 improvement over flat classifiers is not supported as stated because the flat baselines use an uncalibrated decision threshold whereas HiLAP replaces thresholding with a learned stop action. Macro-F1 in multi-label classification is highly sensitive to the decision threshold, and the paper itself notes in Appendix B that per-label scutFBR threshold tuning (as used by Johnson and Zhang 2014) is infeasible in its setting. The comparison to the same base models therefore conflates hierarchy modeling with an adaptive-threshold effect. Please either calibrate the flat baselines (for example, by tuning a global threshold or per-label thresholds on the validation set) or explicitly qualify the 33.4% claim as including the thresholding effect.
[Section 2.2, "States and Policy Network"] The MDP formulation is incomplete because the state st = ReLU(W1_l ReLU(W2_l [ed; lt])) contains only the last assigned label, while the action space At contains the unvisited children of every label assigned so far and the example-F1 reward depends on the entire set of assigned labels. For example, after a leaf label is assigned it no longer appears in At but must still be counted in the reward. No Markovianity or sufficiency argument is provided, so the process as defined is a POMDP rather than an MDP. Please include an aggregate of the assigned-label history in the state or provide an empirical ablation showing that the last-label-only state is sufficient.
[Tables 2-4 and Figure 4] The empirical comparison reports single-run scores with no error bars, multiple seeds, or statistical significance tests, although the text describes improvements as "significant" and the datasets are very imbalanced. Given that the headline gain is large and Macro-F1 values for rare labels are unstable, please report at least the mean and standard deviation over several random seeds and, where feasible, a paired significance test over examples or folds.

minor comments (5)

[Table 1 and Appendix A] The Yelp test-set size is reported as 37,265 in Table 1 but as 37,517 in Appendix A; please reconcile the discrepancy.
[Appendix D] Please report the exact values of the mixing ratio lambda, the scaling factor alpha, the discount factor gamma, and the maximum episode steps; Appendix D gives network sizes and batch size but not these hyperparameters, which are needed for reproducibility.
[Section 2.2, "Actions"] For DAG hierarchies such as Yelp and GO, the paper does not specify how duplicate children reachable through multiple parents are handled when constructing the action space; please clarify this in the action-space definition.
[Figure 5] The popularity bins (P1-P3) and level definitions are not described in the text; the caption only says that labels are divided into three equal-sized categories, so please add the exact construction used for the popularity-based and level-based gains.
[Author affiliations] The affiliation numbering in the header lists affiliation 3 twice rather than providing unique affiliation numbers for the two universities; this typo should be fixed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HiLAP's RL formulation and reported gains are self-contained, and the self-cited reward-shaping component is not load-bearing.

full rationale

We walked the paper's derivation chain. HiLAP defines an MDP with states as concatenations of object and current-label embeddings, actions as unvisited children plus a stop action, and rewards as the difference in example-based F1. This is a genuine learning objective, not a fitted parameter disguised as a prediction. The headline metric, Macro-F1, is not the training reward, and the reported numbers are measured on held-out test sets against external baselines, so the 33.4% improvement claim is an empirical result rather than a tautology. The self-citation to Mao et al. (2018) is used only to motivate intermediate reward shaping, a standard technique; the central claim does not depend on any uniqueness theorem or unverified self-cited result. The paper's guarantee of 0% label inconsistency follows directly from the construction of the action space (only children of already assigned labels are available), so it is a structural property, not a circularly derived empirical finding. The potential concern that flat baselines are not threshold-tuned (Appendix B) bears on fair comparison and correctness, not on circularity. No equation or claim in the paper reduces to its own inputs by construction. Therefore no significant circularity is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard reinforcement learning machinery plus several domain assumptions and undisclosed hyperparameters. The most fragile items are the unchecked Markov property of the truncated state and the absence of reported loss weights, discount factor, and episode length.

free parameters (6)

Loss mixing ratio lambda = not reported
OSL = lambda*Of + (1 - lambda)*Ol in Section 3.2; the reported results depend on a value that is not disclosed.
Scaling factor alpha = not reported
Total loss is O = ORL + alpha * OSL; alpha is not reported.
Discount factor gamma = not reported
Cumulative future reward v_t uses gamma in Section 2.2; no value or sensitivity analysis is given.
Maximum episode steps = not reported
Appendix D says it is set 'reasonably large' depending on the average label count; exact numbers for each dataset are not given.
Batch size = 32
Appendix D states batch size 32 performs best on the validation set; this is a hand-tuned optimization choice.
Policy network hidden sizes = W2_l size 500 on text, 1000 on genomics; W1_l and label embeddings 50 on text, 1000 on genomics
Set by hand in Appendix D without sensitivity analysis.

assumptions (5)

standard math Gradient estimates from the policy gradient theorem (Williams 1992) and the self-critical baseline (Rennie et al. 2017) are unbiased for the stated objective Og.
Used without proof in Section 2.2; this is standard reinforcement learning background.
domain assumption Each object's ground-truth labels form a consistent sub-tree of the hierarchy, and labels may be non-leaf nodes.
Stated in the problem formulation and Figures 1 and 2; for the GO and Yelp DAGs this connectivity assumption is applied to directed acyclic graphs without validation.
ad hoc to paper The state st, built only from the object embedding and the last assigned label embedding, contains enough information for optimal action selection.
Section 2.2 defines st this way but does not prove Markovianity or sufficiency, even though the action space includes children of all previously assigned labels.
ad hoc to paper Greedy inference using the highest-probability action at each step gives a good approximation of the learned policy.
Section 2.2 says at inference 'we greedily select labels with the highest probability as Lhat'; no analysis of the approximation gap is given.
domain assumption Example-based F1 is an appropriate reward and its per-step differences provide useful shaped feedback.
Section 2.2 sets r_t = F1_t - F1_{t-1}; this choice equates the learning objective with the EBF evaluation metric and is not independently validated.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Hierarchical Text Classification with Reinforced Label Assignment." pith.science (2026). https://pith.science/paper/BBPX25MQ

@misc{pith2026190810419,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Text Classification with Reinforced Label Assignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BBPX25MQ}},
  note         = {Machine review of arXiv:1908.10419}
}

read the original abstract

While existing hierarchical text classification (HTC) methods attempt to capture label hierarchies for model training, they either make local decisions regarding each label or completely ignore the hierarchy information during inference. To solve the mismatch between training and inference as well as modeling label dependencies in a more principled way, we formulate HTC as a Markov decision process and propose to learn a Label Assignment Policy via deep reinforcement learning to determine where to place an object and when to stop the assignment process. The proposed method, HiLAP, explores the hierarchy during both training and inference time in a consistent manner and makes inter-dependent decisions. As a general framework, HiLAP can incorporate different neural encoders as base models for end-to-end training. Experiments on five public datasets and four base models show that HiLAP yields an average improvement of 33.4% in Macro-F1 over flat classifiers and outperforms state-of-the-art HTC methods by a large margin. Data and code can be found at https://github.com/morningmoni/HiLAP.

Figures

Figures reproduced from arXiv: 1908.10419 by the authors.

**Figure 1.** We aim at consistent, multi-path, and nonmandatory leaf node prediction. For a Caribbean restaurant with a beer bar, inconsistent prediction may place it to node “Beer Bars” but not “Bars”, which contradicts with each other; Single-path prediction may only recognize that it is a beer bar; Mandatory leaf node prediction would have to assign a leaf node “Dominican” even if the nation of the cuisine is uncertain. li… view at source ↗

**Figure 2.** An illustrative example of the label assignment policy. At t = 0, xi is placed at the root label and the policy would decide if xi should be placed to its two children (red). At t = 1, xi is placed at label “Restaurants”, which adds its three children as the candidates. At t = 6, the stop action is taken and the label assignment is thus terminated. We then take all the labels where xi has been placed (blue) as xi’s … view at source ↗

**Figure 3.** The architecture of the proposed framework HiLAP. One CNN model (Kim, 2014) is used as the base model for illustration. The object embedding ed generated by the base model is combined with the embedding of currently assigned label lt and used as the state representation st, based on which actions are taken by the policy network. The time corresponds to t = 1 in [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Performance comparison of different classification frameworks using the same base models. We compare HiLAP with its flat, supervised variants, and HMCN. Results show that HiLAP exhibits consistent improvement over flat classifiers and larger gains than HMCN. SVM uses …

**Figure 5.** Figure 5: Performance Study on Label Granularity and Popularity. We compute level-based and popularity-based Macro-F1 gains on NYT with bowCNN as base model. We denote the levels of the hierarchy with L1, L2, and L3 (left) and divide the labels into three equal sized categori…

**Figure 6.** Figure 6: One time step in HiLAP-SL. At t = 1, two (K = 2) local per-parent probabilities p Local 1 are measured independently and aggregated in the loss function O1 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Enhancing Text-Based Hierarchical Multilabel Classification for Mobile Applications via Contrastive Learning
cs.LG 2025-07 conditional novelty 5.0 of 10

HMCL, a level-wise negative-sampling contrastive pretraining scheme, improves HMCN's hierarchical multilabel classification for mobile apps and achieved a reported 10.70% KS improvement in a downstream credit-risk task.

Reference graph

Works this paper leans on

41 extracted references · 31 canonical work pages · cited by 1 Pith paper

[1]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Rahul Agrawal, Archit Gupta, Yashoteja Prabhu, and Manik Varma. 2013. Multi-label learning with millions of labels: Recommending advertiser bid phrases for web pages. In WWW, pages 13--24. ACM

work page 2013
[4]

Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473

arXiv 2014
[5]

Wei Bi and Jame T Kwok. 2015. Bayes-optimal hierarchical multilabel classification. TKDE, 27(11):2907--2918

work page 2015
[6]

Wei Bi and James T Kwok. 2011. Multi-label classification on tree-and dag-structured hierarchies. In ICML-11, pages 17--24

work page 2011
[7]

Lijuan Cai and Thomas Hofmann. 2004. Hierarchical document categorization with support vector machines. In CIKM, pages 78--87. ACM

work page 2004
[8]

Asli Celikyilmaz, Antoine Bosselut, Xiaodong He, and Yejin Choi. 2018. Deep communicating agents for abstractive summarization. In NAACL, pages 1662--1675

work page 2018

Show all 41 references

[9]

Ricardo Cerri, Rodrigo C Barros, Andr\' e CPLF de Carvalho, and Yaochu Jin. 2016. Reduction strategies for hierarchical multi-label classification in protein function prediction. BMC bioinformatics, 17(1):373

2016
[10]

Nicol \`o Cesa-Bianchi, Claudio Gentile, and Luca Zaniboni. 2006. Hierarchical classification: combining bayes with svm. In ICML, pages 177--184. ACM

2006
[11]

Amanda Clare and Ross D King. 2003. Predicting gene function in saccharomyces cerevisiae. Bioinformatics, 19(suppl\_2):ii42--ii49

2003
[12]

Siddharth Gopal and Yiming Yang. 2013. Recursive regularization for large-scale classification with hierarchical and graphical dependencies. In KDD, pages 257--265. ACM

2013
[13]

Boris Hayete and Jadwiga R Bienkowska. 2005. Gotrees: predicting go associations from protein domain composition using decision trees. In Biocomputing 2005, pages 127--138. World Scientific

2005
[14]

Rie Johnson and Tong Zhang. 2014. Effective use of word order for text categorization with convolutional neural networks. arXiv preprint arXiv:1412.1058

2014 arXiv
[15]

Yoon Kim. 2014. Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882

2014 arXiv
[16]

Daphne Koller and Mehran Sahami. 1997. Hierarchically classifying documents using very few words. In ICML, pages 170--178. Morgan Kaufmann Publishers Inc

1997
[17]

Siwei Lai, Liheng Xu, Kang Liu, and Jun Zhao. 2015. Recurrent convolutional neural networks for text classification. In AAAI, volume 333, pages 2267--2273

2015
[18]

David D Lewis, Yiming Yang, Tony G Rose, and Fan Li. 2004. Rcv1: A new benchmark collection for text categorization research. Journal of machine learning research, 5(Apr):361--397

2004
[19]

Jingzhou Liu, Wei-Cheng Chang, Yuexin Wu, and Yiming Yang. 2017. Deep learning for extreme multi-label text classification. In Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 115--124. ACM

2017
[20]

Tie-Yan Liu, Yiming Yang, Hao Wan, Hua-Jun Zeng, Zheng Chen, and Wei-Ying Ma. 2005. Support vector machines classification with a very large-scale taxonomy. Acm Sigkdd Explorations Newsletter, 7(1):36--43

2005
[21]

Yuning Mao, Xiang Ren, Jiaming Shen, Xiaotao Gu, and Jiawei Han. 2018. End-to-end reinforcement learning for automatic taxonomy induction. In ACL, pages 2462--2472. Association for Computational Linguistics

2018
[22]

Yu Meng, Jiaming Shen, Chao Zhang, and Jiawei Han. 2018. Weakly-supervised neural text classification. In CIKM

2018
[23]

Ioannis Partalas, Aris Kosmopoulos, Nicolas Baskiotis, Thierry Arti \` e res, George Paliouras, \' E ric Gaussier, Ion Androutsopoulos, Massih - Reza Amini, and Patrick Gallinari. 2015. LSHTC: A benchmark for large-scale text classification. CoRR, abs/1503.08581

2015 arXiv
[24]

Hao Peng, Jianxin Li, Yu He, Yaopeng Liu, Mengjiao Bao, Lihong Wang, Yangqiu Song, and Qiang Yang. 2018. Large-scale hierarchical text classification with recursively regularized deep graph-cnn. In WWW, pages 1063--1072

2018
[25]

Shengwen Peng, Ronghui You, Hongning Wang, Chengxiang Zhai, Hiroshi Mamitsuka, and Shanfeng Zhu. 2016. Deepmesh: deep semantic representation for improving large-scale mesh indexing. Bioinformatics, 32(12):i70--i79

2016
[26]

Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In EMNLP, pages 1532--1543

2014
[27]

Xipeng Qiu, Wenjun Gao, and Xuanjing Huang. 2009. Hierarchical multi-class text categorization with global margin maximization. In acl-ijcnlp 2009, pages 165--168

2009
[28]

Bo Qu, Gao Cong, Cuiping Li, Aixin Sun, and Hong Chen. 2012. An evaluation of classification models for question topic categorization. JASIST, 63:889--903

2012
[29]

Steven J Rennie, Etienne Marcheret, Youssef Mroueh, Jarret Ross, and Vaibhava Goel. 2017. Self-critical sequence training for image captioning. In CVPR, page 3

2017
[30]

Juho Rousu, Craig Saunders, Sandor Szedmak, and John Shawe-Taylor. 2005. Learning hierarchical multi-category text classification models. In ICML, pages 744--751. ACM

2005
[31]

Evan Sandhaus. 2008. The new york times annotated corpus. Linguistic Data Consortium, Philadelphia, 6(12):e26752

2008
[32]

Carlos N Silla and Alex A Freitas. 2011. A survey of hierarchical classification across different application domains. Data Mining and Knowledge Discovery, 22(1-2):31--72

2011
[33]

Carlos N Silla Jr and Alex A Freitas. 2009. A global-model naive bayes approach to the hierarchical prediction of protein functions. In ICDM'09, pages 992--997. IEEE

2009
[34]

Aixin Sun and Ee-Peng Lim. 2001. Hierarchical text classification and evaluation. In ICDM

2001
[35]

Ioannis Tsochantaridis, Thorsten Joachims, Thomas Hofmann, and Yasemin Altun. 2005. Large margin methods for structured and interdependent output variables. Journal of machine learning research, 6(Sep):1453--1484

2005
[36]

Celine Vens, Jan Struyf, Leander Schietgat, Sa s o D z eroski, and Hendrik Blockeel. 2008. Decision trees for hierarchical multi-label classification. Machine Learning, 73(2):185

2008
[37]

Ke Wang, Senqiang Zhou, and Yu He. 2001. Hierarchical classification of real life documents. In SDM, pages 1--16. SIAM

2001
[38]

Jonatas Wehrmann, Ricardo Cerri, and Rodrigo Barros. 2018. Hierarchical multi-label classification networks. In ICML, pages 5225--5234

2018
[39]

Ronald J Williams. 1992. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine learning, 8(3-4):229--256

1992
[40]

IPC WIPO. 2014. International patent classification (ipc). World Intellectual Property Organization, Geneve

2014
[41]

Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. 2016. Hierarchical attention networks for document classification. In NAACL, pages 1480--1489

2016

Pith tools

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

[1] [1]

URL: " 'urlintro :=

ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Rahul Agrawal, Archit Gupta, Yashoteja Prabhu, and Manik Varma. 2013. Multi-label learning with millions of labels: Recommending advertiser bid phrases for web pages. In WWW, pages 13--24. ACM

work page 2013

[4] [4]

Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473

arXiv 2014

[5] [5]

Wei Bi and Jame T Kwok. 2015. Bayes-optimal hierarchical multilabel classification. TKDE, 27(11):2907--2918

work page 2015

[6] [6]

Wei Bi and James T Kwok. 2011. Multi-label classification on tree-and dag-structured hierarchies. In ICML-11, pages 17--24

work page 2011

[7] [7]

Lijuan Cai and Thomas Hofmann. 2004. Hierarchical document categorization with support vector machines. In CIKM, pages 78--87. ACM

work page 2004

[8] [8]

Asli Celikyilmaz, Antoine Bosselut, Xiaodong He, and Yejin Choi. 2018. Deep communicating agents for abstractive summarization. In NAACL, pages 1662--1675

work page 2018