Recurrent neural networks for dynamic user intent prediction in human-database interaction

Vamsi Meduri, Kanchan Chowdhury, Mohamed Elsayed

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Prediction of human intent during a user interaction session with the database received a significant amount of attention in the recent past [1, 8]. State-of-the-art intent detection approaches such as [5] insist that the human intent is dynamic and is constantly changing throughout the user session. While the usage of classifiers like SVMs and decision trees have been proposed to capture static user intent [2], such models become ineffective in predicting dynamic or ever-changing human intent. Recurrent Neural Networks (RNNs) are powerful temporal predictors and have recently been prominent in the database research community for tasks such as entity matching [3, 6]. In this work, we discuss the application of RNNs to the problem of dynamic user intent prediction during a human-database interaction. We propose two variants of SQL-specific embedding vectors for RNNs. We also propose active learning strategies for RNNs which consume a fraction of the held-out training data to produce competitive prediction quality as full training or supervised learning. Our experiments on real user sessions upon the NYCTaxiTrip dataset [9] evaluate the effectiveness of vanilla, LSTM and GRU based RNNs.

Original languageEnglish (US)
Title of host publicationAdvances in Database Technology - EDBT 2019
Subtitle of host publication22nd International Conference on Extending Database Technology, Proceedings
EditorsZoi Kaoudi, Helena Galhardas, Irini Fundulaki, Berthold Reinwald, Melanie Herschel, Carsten Binnig
PublisherOpenProceedings.org
Pages654-657
Number of pages4
ISBN (Electronic)9783893180813
DOIs
StatePublished - Jan 1 2019
Event22nd International Conference on Extending Database Technology, EDBT 2019 - Lisbon, Portugal
Duration: Mar 26 2019Mar 29 2019

Publication series

NameAdvances in Database Technology - EDBT
Volume2019-March
ISSN (Electronic)2367-2005

Conference

Conference22nd International Conference on Extending Database Technology, EDBT 2019
CountryPortugal
CityLisbon
Period3/26/193/29/19

Fingerprint

Recurrent neural networks
Supervised learning
Decision trees
Classifiers
Experiments

ASJC Scopus subject areas

  • Information Systems
  • Software
  • Computer Science Applications

Cite this

Meduri, V., Chowdhury, K., & Elsayed, M. (2019). Recurrent neural networks for dynamic user intent prediction in human-database interaction. In Z. Kaoudi, H. Galhardas, I. Fundulaki, B. Reinwald, M. Herschel, & C. Binnig (Eds.), Advances in Database Technology - EDBT 2019: 22nd International Conference on Extending Database Technology, Proceedings (pp. 654-657). (Advances in Database Technology - EDBT; Vol. 2019-March). OpenProceedings.org. https://doi.org/10.5441/002/edbt.2019.79

Recurrent neural networks for dynamic user intent prediction in human-database interaction. / Meduri, Vamsi; Chowdhury, Kanchan; Elsayed, Mohamed.

Advances in Database Technology - EDBT 2019: 22nd International Conference on Extending Database Technology, Proceedings. ed. / Zoi Kaoudi; Helena Galhardas; Irini Fundulaki; Berthold Reinwald; Melanie Herschel; Carsten Binnig. OpenProceedings.org, 2019. p. 654-657 (Advances in Database Technology - EDBT; Vol. 2019-March).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Meduri, V, Chowdhury, K & Elsayed, M 2019, Recurrent neural networks for dynamic user intent prediction in human-database interaction. in Z Kaoudi, H Galhardas, I Fundulaki, B Reinwald, M Herschel & C Binnig (eds), Advances in Database Technology - EDBT 2019: 22nd International Conference on Extending Database Technology, Proceedings. Advances in Database Technology - EDBT, vol. 2019-March, OpenProceedings.org, pp. 654-657, 22nd International Conference on Extending Database Technology, EDBT 2019, Lisbon, Portugal, 3/26/19. https://doi.org/10.5441/002/edbt.2019.79
Meduri V, Chowdhury K, Elsayed M. Recurrent neural networks for dynamic user intent prediction in human-database interaction. In Kaoudi Z, Galhardas H, Fundulaki I, Reinwald B, Herschel M, Binnig C, editors, Advances in Database Technology - EDBT 2019: 22nd International Conference on Extending Database Technology, Proceedings. OpenProceedings.org. 2019. p. 654-657. (Advances in Database Technology - EDBT). https://doi.org/10.5441/002/edbt.2019.79
Meduri, Vamsi ; Chowdhury, Kanchan ; Elsayed, Mohamed. / Recurrent neural networks for dynamic user intent prediction in human-database interaction. Advances in Database Technology - EDBT 2019: 22nd International Conference on Extending Database Technology, Proceedings. editor / Zoi Kaoudi ; Helena Galhardas ; Irini Fundulaki ; Berthold Reinwald ; Melanie Herschel ; Carsten Binnig. OpenProceedings.org, 2019. pp. 654-657 (Advances in Database Technology - EDBT).
@inproceedings{a122618b745842c19e6f5a028c484cc8,
title = "Recurrent neural networks for dynamic user intent prediction in human-database interaction",
abstract = "Prediction of human intent during a user interaction session with the database received a significant amount of attention in the recent past [1, 8]. State-of-the-art intent detection approaches such as [5] insist that the human intent is dynamic and is constantly changing throughout the user session. While the usage of classifiers like SVMs and decision trees have been proposed to capture static user intent [2], such models become ineffective in predicting dynamic or ever-changing human intent. Recurrent Neural Networks (RNNs) are powerful temporal predictors and have recently been prominent in the database research community for tasks such as entity matching [3, 6]. In this work, we discuss the application of RNNs to the problem of dynamic user intent prediction during a human-database interaction. We propose two variants of SQL-specific embedding vectors for RNNs. We also propose active learning strategies for RNNs which consume a fraction of the held-out training data to produce competitive prediction quality as full training or supervised learning. Our experiments on real user sessions upon the NYCTaxiTrip dataset [9] evaluate the effectiveness of vanilla, LSTM and GRU based RNNs.",
author = "Vamsi Meduri and Kanchan Chowdhury and Mohamed Elsayed",
year = "2019",
month = "1",
day = "1",
doi = "10.5441/002/edbt.2019.79",
language = "English (US)",
series = "Advances in Database Technology - EDBT",
publisher = "OpenProceedings.org",
pages = "654--657",
editor = "Zoi Kaoudi and Helena Galhardas and Irini Fundulaki and Berthold Reinwald and Melanie Herschel and Carsten Binnig",
booktitle = "Advances in Database Technology - EDBT 2019",

}

TY - GEN

T1 - Recurrent neural networks for dynamic user intent prediction in human-database interaction

AU - Meduri, Vamsi

AU - Chowdhury, Kanchan

AU - Elsayed, Mohamed

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Prediction of human intent during a user interaction session with the database received a significant amount of attention in the recent past [1, 8]. State-of-the-art intent detection approaches such as [5] insist that the human intent is dynamic and is constantly changing throughout the user session. While the usage of classifiers like SVMs and decision trees have been proposed to capture static user intent [2], such models become ineffective in predicting dynamic or ever-changing human intent. Recurrent Neural Networks (RNNs) are powerful temporal predictors and have recently been prominent in the database research community for tasks such as entity matching [3, 6]. In this work, we discuss the application of RNNs to the problem of dynamic user intent prediction during a human-database interaction. We propose two variants of SQL-specific embedding vectors for RNNs. We also propose active learning strategies for RNNs which consume a fraction of the held-out training data to produce competitive prediction quality as full training or supervised learning. Our experiments on real user sessions upon the NYCTaxiTrip dataset [9] evaluate the effectiveness of vanilla, LSTM and GRU based RNNs.

AB - Prediction of human intent during a user interaction session with the database received a significant amount of attention in the recent past [1, 8]. State-of-the-art intent detection approaches such as [5] insist that the human intent is dynamic and is constantly changing throughout the user session. While the usage of classifiers like SVMs and decision trees have been proposed to capture static user intent [2], such models become ineffective in predicting dynamic or ever-changing human intent. Recurrent Neural Networks (RNNs) are powerful temporal predictors and have recently been prominent in the database research community for tasks such as entity matching [3, 6]. In this work, we discuss the application of RNNs to the problem of dynamic user intent prediction during a human-database interaction. We propose two variants of SQL-specific embedding vectors for RNNs. We also propose active learning strategies for RNNs which consume a fraction of the held-out training data to produce competitive prediction quality as full training or supervised learning. Our experiments on real user sessions upon the NYCTaxiTrip dataset [9] evaluate the effectiveness of vanilla, LSTM and GRU based RNNs.

UR - http://www.scopus.com/inward/record.url?scp=85064918837&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064918837&partnerID=8YFLogxK

U2 - 10.5441/002/edbt.2019.79

DO - 10.5441/002/edbt.2019.79

M3 - Conference contribution

T3 - Advances in Database Technology - EDBT

SP - 654

EP - 657

BT - Advances in Database Technology - EDBT 2019

A2 - Kaoudi, Zoi

A2 - Galhardas, Helena

A2 - Fundulaki, Irini

A2 - Reinwald, Berthold

A2 - Herschel, Melanie

A2 - Binnig, Carsten

PB - OpenProceedings.org

ER -