Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Are You for Real? Detecting Identity Fraud via Dialogue Interactions

T0 review · 4 major / 6 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 A trainable dialogue system that asks knowledge-graph-derived questions can detect identity fraud more accurately than rule-based questioning.

desk verdict A genuinely new task formulation with a sensible architecture, but the central accuracy claim is measured on a simulator that shares its key covariate with the learned policy, so the 88.4% result is not evidence about real fraudsters. read the letter →

arxiv 1908.06820 v1 pith:ZERSJIJ6 submitted 2019-08-19 cs.CL cs.AI

Weikang Wang , Jiajun Zhang , Qian Li , Chengqing Zong , Zhifei Li This is my paper · ORCID
classification cs.CLcs.AI
keywords identityfrauddetectiondialoguesystemknowledgegraphhierarchicalreinforcementlearningpolicygradientusersimulationderivedquestions
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 tries to establish that identity fraud in loan applications can be detected by a dialogue system that asks questions derived from a knowledge graph, rather than by checking documents or asking applicants to repeat their claimed details. The key move is to ask about facts a fraudster would not think to memorize, such as the nearest subway station to the school they claim to have attended. The paper builds a personal knowledge graph for each applicant, adds map-derived facts, and trains a hierarchical dialogue policy with reinforcement learning. On a simulator whose applicants answer according to fitted human behavior curves, the full system reaches 88.4% recognition accuracy, compared with 78.7% for a hierarchical rule-based questioner and 74.8% for flat random questioning. If the simulator's assumptions hold, this points toward an automated, interpretable anti-fraud dialogue that needs no labeled fraud examples.

What carries the argument

The load-bearing mechanism is the derived-question generator over a completed personal knowledge graph. For each claimed fact such as 'graduated from Nanjing University,' the system looks up related triplets, such as (Nanjing University, FoundedDate, 1902) or map-generated (Nanjing University, SubwayStation, Gulou Subway Station), and turns the tail entity into a multiple-choice question. The dialogue manager treats that graph as a directed network of user, personal-information, and answer nodes; a knowledge-graph-based state tracker computes node embeddings by max-pooled message passing, and a hierarchical policy has a manager choose which personal-information node to verify and a worker choose which answer node to ask about. A static 'spread degree' feature, equal to the search-engine hit count of each fact, is included in the state representation because the paper finds it correlates with whether applicants know the fact. This machinery lets the system learn to prioritize discriminating questions and to skip facts that are entailed by already-known neighboring facts.

What would settle it

Run the same derived questions on real loan applicants whose true identity status is known from verified documents, and compare their answer rates per search-engine-hit-count bin against Figure 4; if fraudsters often answer very obscure derived questions correctly, or legitimate applicants often miss very common ones, the simulator's fitted curves, and the 88.4% accuracy built on them, will not transfer to real detection.

Watch

Extended reading notes

Core claim

The paper's central claim is that a trainable dialogue agent can detect identity fraud more accurately than rule-based agents by asking derived questions, which are questions built from knowledge-graph triplets whose head is a piece of the applicant's claimed personal information. The agent constructs a directed graph with the applicant's user node, personal-information nodes (school, company, residence, birthplace), and answer nodes; it then selects questions through a manager-worker hierarchical policy, with a message-passing state tracker that propagates what the applicant does or does not know across related facts. In the simulated test set, the full model (Full-S) reaches 88.4% accuracy and 9.8 average turns, versus 78.7% for the Hierarchical Rule baseline, and the learned policy shows interpretable behavior, such as prioritizing school and company verification and overriding individual worker mistakes.

Load-bearing premise

The load-bearing premise is that real applicants' and fraudsters' answer behavior follows the curves fitted in Figure 4, where knowing a fact rises with its search-engine hit count, plus the rule that someone who knows two facts that together imply a third must know the third.

Editorial extensions

If this is right

  • Data-driven dialogue policies can outperform hand-crafted rule systems on fraud detection, by about ten accuracy points in the simulated setting.
  • The policy learns to verify the most frequently forged information types first, namely school and company, and to gather more evidence before deciding when a worker's signals conflict.
  • Message passing across the knowledge graph reduces redundant questions, because knowing one nearby fact can make another fact inferable.
  • The manager's terminal decisions mostly follow worker verdicts, with conditional probabilities around 0.95 to 0.96, yet the manager can overrule a single worker, making the overall system robust to individual errors.
  • With no labeled fraud data, the whole system can be trained from rule-based demonstrations and reinforcement learning, which matters for deployment where fraud labels are scarce.

Reading between the lines

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

  • Because the same 'spread degree' signal both generates the simulated applicants' answer probabilities and feeds the policy's state features, the reported accuracy may partly reflect the simulator's own assumptions; a live trial with verified identities would be needed to measure real-world transfer.
  • If the fitted answer-rate curves generalize, the same derived-question strategy could extend beyond the four personal-information types to employment history, family relationships, or property records by building knowledge graphs from public data.
  • The transitive-knowledge hypothesis suggests that an explicit entailment model of facts could further improve question selection, beyond the message-passing approximation used here.
  • A testable extension would field the trained policy against a human-populated set of real and fake applicant profiles, comparing its decisions with document-based ground truth.
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.

Referee Report

4 major / 6 minor

Summary. This paper proposes an interactive dialogue system for detecting identity fraud in loan applications. The system constructs a knowledge graph for each applicant, and uses hierarchical reinforcement learning with a graph-neural-network state tracker to ask 'derived questions' about personal information. A heuristic user simulator is built using three hypotheses: that answer probability increases with web 'spread degree', approximated by search-engine hit counts, and that knowledge is closed-loop transitive. The simulator's answer probabilities are fitted to 31 volunteers. Experiments on the simulator report 88.4% accuracy for the full model (Full-S), outperforming rule-based baselines (74.8% and 78.7%). The paper also presents ablation studies and analyses of the learned manager and worker policies.

Significance. If validated, this work would introduce a novel application of task-oriented dialogue to anti-fraud, with an interpretable questioning policy and a practical KG completion method. The authors deserve credit for releasing code, carrying out a small human experiment to ground the simulator, and being explicit about the premise that applicants answer without tools (footnote 2). However, all effectiveness claims rest on a simulator whose answer-generation mechanism shares a key feature with the policy's state representation. The reported accuracy primarily measures the policy's ability to exploit that shared feature, not real-world detection. The significance is therefore conditional on additional validation or on re-scoping the claims to a simulation study.

major comments (4)
  1. [Section 5.1 vs. Section 3.1] The simulator defines p(k_i=1) using binned curves of log(Freq(eh_i, et_i)+1) fitted from 31 volunteers (Fig. 4). The agent's state representation includes the same 'spread degree on the internet' as a static feature of answer nodes (Section 3.1). Consequently, Full-S can learn to select questions whose spread degree separates the fraud and non-fraud curves, and the 88.4% accuracy in Fig. 5 measures exploitation of exactly the relationship that generates the simulated applicants. No held-out human evaluation is provided, so the abstract's claim that the system 'can effectively detect fraudsters' is not supported.
  2. [Section 5.1, Hypothesis 3] The simulator imposes the closed-loop transitivity assumption by calibrating sampled knowledge variables until they agree with the hypothesis. The message-passing state tracker (Section 3.1, Eq. 1) propagates known information along graph paths. The Full-S vs. HP-S ablation gain therefore may reflect that the model and simulator share the same structural assumption, rather than evidence that message passing captures a real property of human knowledge. The authors do not independently validate Hypothesis 3; the 31 volunteers' responses are only used to fit the spread-degree curves, not to test transitivity.
  3. [Section 5.3 and Fig. 5] The test set is generated from the same simulator used for training, with no distribution shift; the train/dev/test split is only over the 906 applicants' personal information, while the answer-generation process is identical across all splits. Thus the test accuracy does not measure generalization to real applicants. In addition, Fig. 5 reports 10-epoch averages without error bars or confidence intervals, and the small human sample (31 volunteers) yields no uncertainty quantification for the fitted curves in Fig. 4. The sensitivity of all results to these fitted probabilities is unknown.
  4. [Section 5.1, simulation validity] The simulator assumes that fraudsters' knowledge of derived questions depends only on identity state and spread degree. Real fraudsters may prepare or memorize answers to high-spread facts, or may differ from the 31 volunteers in other systematic ways. Because this behavior is not represented in the simulator, the reported advantage over rule-based systems could shrink or reverse under a more realistic user model. This is a validity threat to the central effectiveness claim.
minor comments (6)
  1. [Section 5.4] Typo: 'possbile' should be 'possible'.
  2. [Title] The title in the PDF displays 'Are Y ou for Real?' with an unwanted space in 'You'.
  3. [Fig. 5] The figure would benefit from error bars or variance reporting; the text states that test results are averaged over 10 epochs but does not report the spread.
  4. [Footnote 1] The author affiliation footnote appears garbled ('Nanjing University Baidu No.3 Gulou Street'); it should likely be a standard affiliation list.
  5. [Section 5.6, Rule 2] The phrase 'all workers' decisions are both "Non-Fraud"' is awkward; it should read 'all workers' decisions are "Non-Fraud"'.
  6. [Section 5.3] The collection of 906 applicants' personal information is described in one sentence; if these are real applicants, more detail on data source, anonymization, and privacy considerations is needed.

Circularity Check

1 steps flagged · score 6.0 of 10

Fraud-detection accuracy is measured only on a simulator whose answers are generated from the same search-frequency 'spread degree' feature the learned policy observes, so the 88.4% result is largely a read-out of the fitted p(k|Freq) curves rather than independent evidence.

  1. fitted input called prediction [Section 3.1 (Static Features), Section 5.1 (User Simulator and Human Experiments), evaluated in Fig. 5]
    "we use the “spread degree on the internet” to distinguish different answer nodes because we find there is an obvious correlation between this “spread degree” feature and applicants’ behavior in our human experiments ... The number of the retrieved results will be the “spread degree” feature of va. ... For both fraudsters and normal applicants, p(ki = 1) is proportional to the “spread degree” of ti. ... we first estimate the function relations between p(ki = 1) and Freq(ehi, eti) via human experiments."

    The user simulator sets the probability that an applicant knows each fact as a fitted function of search-engine hit counts, Freq(ehi, eti), while Section 3.1 gives the learned dialogue policy the exact same search-derived hit count as the static 'spread degree' node feature. The RL agents are trained and evaluated against rewards sampled from this fitted p(k|Freq) while having Freq as an input feature. Full-S can therefore learn to select questions whose hit-count separates the two fitted curves, and the reported 88.4% accuracy (Fig. 5) mostly measures the separation of those curves, not detection of real applicants.

full rationale

The load-bearing empirical claim is the abstract's assertion that the trainable system can detect fraudsters and beat rule-based systems. That claim is supported only by Fig. 5, measured on the authors' own heuristic simulator. Section 5.1 fits p(ki=1) to binned log Freq(ehi,eti) from 31 volunteers and generates simulated applicant answers from those fitted curves; Section 3.1 gives the learned policy the same search-derived 'spread degree' as a static node feature in KG-DST. The learned policy is therefore optimized to exploit exactly the Freq-to-knowledge relationship that defines the simulated answers, so the reported accuracy is substantially a read-out of the fitted separation between fraud and non-fraud curves. This matches the fitted-input-called-prediction pattern. It is not a formal equation-level identity, so the paper is not fully circular in its RL machinery, but the central empirical claim is not independently supported: there is no held-out human evaluation, and Footnote 2 itself restricts the premise to applicants who answer without external help. The paper's self-citations (Wang et al. 2018, 2019) are related-work references and are not load-bearing, so they do not raise the score. Hypothesis 3 is additionally imposed on the simulator and mirrored by message passing, so the HP-S versus Full-S ablation partly rewards self-consistency; this is a secondary concern rather than a separate fully formal circular step. Overall, one core evaluation step reduces to the fitted relationship, giving a circularity score of 6.

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

The central evaluation is built on the heuristic user simulator. The simulator's answer probabilities are fitted to 31 volunteers' responses binned by search-engine hit count, and the closed-loop transitivity rule is an unvalidated assumption. The dialogue system's policies are then trained and scored against this same simulator, using the same hit-count feature as input. The listed free parameters and axioms are the load-bearing extra assumptions the reader must accept on top of the standard RL and GNN machinery.

free parameters (5)
  • p_know_fraud_bins = bin-averaged ratios from 31 volunteers (10 bins in log(Freq+1)); values shown in Fig. 4
    The simulator's probability that a fraudster applicant knows a fact is computed from these fitted ratios; the dialogue system's questioning strategy is trained and evaluated against this simulator.
  • p_know_nonfraud_bins = bin-averaged ratios from 31 volunteers (10 bins); values shown in Fig. 4
    Same as the fraud bins, but for non-fraud applicants; the simulated normal applicants answer accordingly.
  • fake_item_sampling_weights = School:Company:Residence:BirthPlace = 2:2:1:1; number of fake items sampled uniformly from 1 to 4
    Used to generate simulated fraudsters in the user simulator; chosen from the intuition that forging School or Company yields larger loans.
  • reward_weights = rm_crt=3, rm_wrg=3, rw_crt=1, rw_wrg=1, rturn=0.1
    Hand-set rewards in the RL training; they shape the learned policy and the resulting turn counts and accuracy.
  • action_mask_thresholds = worker decision after at least 3 answer nodes explored or all explored; manager decision after all workers decided or…
    Domain rules encoded as action masks in Section 3.2; they constrain the policy and affect how quickly decisions are made.
assumptions (5)
  • domain assumption Loan applicants answer questions without external help (no QA systems or search tools) under time constraints.
    Stated in footnote 2; the entire derived-question strategy relies on fraudsters being unable to look up facts during the interview.
  • domain assumption The probability that an applicant knows a fact is proportional to the fact's spread degree on the web, approximated by the number of search-engine hits for the head and tail entities.
    Hypotheses 1 and 2 in Section 5.1; validated only on 31 volunteers, not on a real applicant population.
  • ad hoc to paper For any closed loop of triplets, if an applicant knows two of them, they know all of them.
    Hypothesis 3 in Section 5.1; used to calibrate the simulated applicants' answers and to justify skipping redundant questions.
  • domain assumption The external Chinese KG (OwnThink) and Amap POI data are complete enough that nearly all personal information entities can find triplets for derived question generation.
    Section 2; derived question generation depends on KG coverage, and the completion procedure keeps only the nearest POI for each type.
  • standard math Max aggregation in message passing (Eq. 1) is a suitable dialogue-state representation.
    Standard GNN design choice, not externally validated for this task; the authors compare with K=0 in HP-S as an ablation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Are You for Real? Detecting Identity Fraud via Dialogue Interactions." pith.science (2026). https://pith.science/paper/ZERSJIJ6

@misc{pith2026190806820,
  author       = {Pith},
  title        = {Pith review of: Are You for Real? Detecting Identity Fraud via Dialogue Interactions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZERSJIJ6}},
  note         = {Machine review of arXiv:1908.06820}
}
read the original abstract

Identity fraud detection is of great importance in many real-world scenarios such as the financial industry. However, few studies addressed this problem before. In this paper, we focus on identity fraud detection in loan applications and propose to solve this problem with a novel interactive dialogue system which consists of two modules. One is the knowledge graph (KG) constructor organizing the personal information for each loan applicant. The other is structured dialogue management that can dynamically generate a series of questions based on the personal KG to ask the applicants and determine their identity states. We also present a heuristic user simulator based on problem analysis to evaluate our method. Experiments have shown that the trainable dialogue system can effectively detect fraudsters, and achieve higher recognition accuracy compared with rule-based systems. Furthermore, our learned dialogue strategies are interpretable and flexible, which can help promote real-world applications.

Figures

Figures reproduced from arXiv: 1908.06820 by the authors.

Figure 1
Figure 1. Dialogue examples of two possible fraud de [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An example of the KG for Nanjing University. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of our approach. To build the directed graph for dialogue management, we reverse directions [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: We can find that the statistical distributions of real behavior patterns of normal applicants and fraudsters are distinguishable and the results agree with our first two intuitions. [0.00, 1.84) [1.84, 3.68) [3.68, 5.53) [5.53, 7.37) [7.37, 9.21) [9.21, 11.05) [11.05, …
Figure 5
Figure 5. Figure 5: Performance of different systems. Tested on 10 epochs using the best model during training. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Accuracy curves of different neural models in [PITH_FULL_IMAGE:figures/full_fig_p007_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

  1. Joint Detection of Fraud and Concept Drift inOnline Conversations with LLM-Assisted Judgment

    cs.CL 2025-05 reject novelty 4.0 of 10

    A hybrid chat-fraud pipeline with classifier, drift detector, and LLM judge is described, but only the classifier stage is measured and the dataset citation is incorrect.

Reference graph

Works this paper leans on

32 extracted references · 20 canonical work pages · cited by 1 Pith paper

  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]

    Peter W Battaglia, Jessica B Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Vinicius Zambaldi, Mateusz Malinowski, Andrea Tacchetti, David Raposo, Adam Santoro, Ryan Faulkner, et al. 2018. Relational inductive biases, deep learning, and graph networks. arXiv preprint arXiv:1806.01261

    arXiv 2018

  4. [4]

    Nisha Bhaskaran, Ifeoma Nwogu, Mark G Frank, and Venu Govindaraju. 2011. Lie to me: Deceit detection via online behavioral learning. In Face and Gesture 2011, pages 24--29. IEEE

    work page 2011

  5. [5]

    Lu Chen, Bowen Tan, Sishan Long, and Kai Yu. 2018. Structured dialogue policy with graph neural networks. In Proceedings of the 27th International Conference on Computational Linguistics, pages 1257--1268

    2018

  6. [6]

    Carlos Florensa, Yan Duan, and Pieter Abbeel. 2017. Stochastic neural networks for hierarchical reinforcement learning. arXiv preprint arXiv:1704.03012

    arXiv 2017

  7. [7]

    Milica Ga s i \'c , Filip Jur c \' c ek, Blaise Thomson, Kai Yu, and Steve Young. 2011. On-line policy optimisation of spoken dialogue systems via live interaction with human subjects. In 2011 IEEE Workshop on Automatic Speech Recognition & Understanding, pages 312--317. IEEE

    work page 2011

  8. [8]

    Kallirroi Georgila, James Henderson, and Oliver Lemon. 2006. User simulation for spoken dialogue systems: Learning and evaluation. In Ninth International Conference on Spoken Language Processing

    work page 2006

Show all 32 references
  1. [9]

    Martin Graciarena, Elizabeth Shriberg, Andreas Stolcke, Frank Enos, Julia Hirschberg, and Sachin Kajarekar. 2006. Combining prosodic lexical and cepstral systems for deceptive speech detection. In 2006 IEEE International Conference on Acoustics Speech and Signal Processing Pro...

    2006

  2. [10]

    He He, Anusha Balakrishnan, Mihail Eric, and Percy Liang. 2017. Learning symmetric collaborative dialogue agents with dynamic knowledge graph embeddings. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1...

    2017

  3. [11]

    Guoliang Ji, Kang Liu, Shizhu He, and Jun Zhao. 2016. Knowledge graph completion with adaptive sparse transfer matrix. In Thirtieth AAAI Conference on Artificial Intelligence

    2016

  4. [12]

    Gangeshwar Krishnamurthy, Navonil Majumder, Soujanya Poria, and Erik Cambria. 2018. A deep learning approach for multimodal deception detection. arXiv preprint arXiv:1803.00344

    2018 arXiv

  5. [13]

    Tejas D Kulkarni, Karthik Narasimhan, Ardavan Saeedi, and Josh Tenenbaum. 2016. Hierarchical deep reinforcement learning: Integrating temporal abstraction and intrinsic motivation. In Advances in neural information processing systems, pages 3675--3683

    2016

  6. [14]

    Sarah Ita Levitan, Angel Maredia, and Julia Hirschberg. 2018. Acoustic-prosodic indicators of deception and trust in interview dialogues. Proc. Interspeech 2018, pages 416--420

    2018

  7. [15]

    Xiujun Li, Yun-Nung Chen, Lihong Li, Jianfeng Gao, and Asli Celikyilmaz. 2017. End-to-end task-completion neural dialogue systems. arXiv preprint arXiv:1703.01008

    2017 arXiv

  8. [16]

    Xiujun Li, Zachary C Lipton, Bhuwan Dhingra, Lihong Li, Jianfeng Gao, and Yun-Nung Chen. 2016. A user simulator for task-completion dialogues. arXiv preprint arXiv:1612.05688

    2016 arXiv

  9. [17]

    Yujia Li, Daniel Tarlow, Marc Brockschmidt, and Richard Zemel. 2015. Gated graph sequence neural networks. arXiv preprint arXiv:1511.05493

    2015 arXiv

  10. [18]

    Shan Lu, Gabriel Tsechpenakis, Dimitris N Metaxas, Matthew L Jensen, and John Kruse. 2005. Blob analysis of the head and hands: A method for deception detection. In Proceedings of the 38th Annual Hawaii International Conference on System Sciences, pages 20c--20c. IEEE

    2005

  11. [19]

    Thomas O Meservy, Matthew L Jensen, John Kruse, Judee K Burgoon, Jay F Nunamaker, Douglas P Twitchell, Gabriel Tsechpenakis, and Dimitris N Metaxas. 2005. Deception detection through automatic, unobtrusive analysis of nonverbal behavior. IEEE Intelligent Systems, 20(5):36--43

    2005

  12. [20]

    Mathias Niepert, Mohamed Ahmed, and Konstantin Kutzkov. 2016. Learning convolutional neural networks for graphs. In International conference on machine learning, pages 2014--2023

    2016

  13. [21]

    Baolin Peng, Xiujun Li, Lihong Li, Jianfeng Gao, Asli Celikyilmaz, Sungjin Lee, and Kam-Fai Wong. 2017. Composite task-completion dialogue policy learning via hierarchical deep reinforcement learning. In Proceedings of the 2017 Conference on Empirical Methods in Natural Langua...

    2017

  14. [22]

    Ver \'o nica P \'e rez-Rosas, Mohamed Abouelenien, Rada Mihalcea, Yao Xiao, CJ Linton, and Mihai Burzo. 2015. Verbal and nonverbal clues for real-life deception detection. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2336--2346

    2015

  15. [23]

    Olivier Pietquin and Thierry Dutoit. 2006. A probabilistic framework for dialog simulation and optimal strategy learning. IEEE Transactions on Audio, Speech, and Language Processing, 14(2):589--599

    2006

  16. [24]

    Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. 2009. The graph neural network model. IEEE Transactions on Neural Networks, 20(1):61--80

    2009

  17. [25]

    Th \'e o Trouillon, Christopher R Dance, \'E ric Gaussier, Johannes Welbl, Sebastian Riedel, and Guillaume Bouchard. 2017. Knowledge graph completion via complex tensor factorization. The Journal of Machine Learning Research, 18(1):4735--4772

    2017

  18. [26]

    Alexander Sasha Vezhnevets, Simon Osindero, Tom Schaul, Nicolas Heess, Max Jaderberg, David Silver, and Koray Kavukcuoglu. 2017. Feudal networks for hierarchical reinforcement learning. In Proceedings of the 34th International Conference on Machine Learning-Volume 70, pages 35...

    2017

  19. [27]

    Weikang Wang, Jiajun Zhang, Qian Li, Mei-Yuh Hwang, Chengqing Zong, and Zhifei Li. 2019. Incremental learning from scratch for task-oriented dialogue systems. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3710--3720, Florence...

    2019

  20. [28]

    Weikang Wang, Jiajun Zhang, Han Zhang, Mei-Yuh Hwang, Chengqing Zong, and Zhifei Li. 2018. A teacher-student framework for maintainable dialog manager. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3803--3812, Brussels, Belgiu...

    2018

  21. [29]

    Tsung-Hsien Wen, David Vandyke, Nikola Mrk s i \'c , Milica Gasic, Lina M Rojas Barahona, Pei-Hao Su, Stefan Ultes, and Steve Young. 2017. A network-based end-to-end trainable task-oriented dialogue system. In Proceedings of the 15th Conference of the European Chapter of the A...

    2017

  22. [30]

    Jason D Williams, Kavosh Asadi, and Geoffrey Zweig. 2017. Hybrid code networks: practical and efficient end-to-end dialog control with supervised and reinforcement learning. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: L...

    2017

  23. [31]

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

    1992

  24. [32]

    Steve Young, Milica Ga s i \'c , Blaise Thomson, and Jason D Williams. 2013. Pomdp-based statistical spoken dialog systems: A review. Proceedings of the IEEE, 101(5):1160--1179

    2013

Pith tools

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