PULP: A system for exploratory search of scientific literature

Alan Medlar, Kalle Ilves, Ping Wang, Wray Buntine, Dorota Głowacka

    Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

    Abstract

    Despite the growing importance of exploratory search, information retrieval (IR) systems tend to focus on lookup search. Lookup searches are well served by optimising the precision and recall of search results, however, for exploratory search this may be counterproductive if users are unable to formulate an appropriate search query. We present a system called PULP that supports exploratory search for scientific literature, though the system can be easily adapted to other types of literature. PULP uses reinforcement learning (RL) to avert the user from context traps resulting from poorly chosen search queries, trading off between exploration (presenting the user with diverse topics) and exploitation (moving towards more specific topics). Where other RL-based systems suffer from the "cold start" problem, requiring sufficient time to adjust to a user's information needs, PULP initially presents the user with an overview of the dataset using temporal topic models. Topic models are displayed in an interactive alluvial diagram, where topics are shown as ribbons that change thickness with a given topics relative prevalence over time. Interactive, exploratory search sessions can be initiated by selecting topics as a starting point.

    Original languageEnglish
    Title of host publicationProceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016)
    EditorsJaved Aslam, Ian Ruthven, Justin Zobel
    Place of PublicationNew York, NY, USA
    PublisherAssociation for Computing Machinery (ACM)
    Pages1133-1136
    Number of pages4
    ISBN (Print)9781450340694
    DOIs
    Publication statusPublished - 7 Jul 2016
    EventACM International Conference on Research and Development in Information Retrieval 2016 - Pisa, Italy
    Duration: 17 Jul 201621 Jul 2016
    Conference number: 39th

    Conference

    ConferenceACM International Conference on Research and Development in Information Retrieval 2016
    Abbreviated titleSIGIR 2016
    CountryItaly
    CityPisa
    Period17/07/1621/07/16

    Keywords

    • Exploratory search
    • Topic models
    • Bandit algorithms
    • Scientific literature search
    • Query formulation

    Cite this

    Medlar, A., Ilves, K., Wang, P., Buntine, W., & Głowacka, D. (2016). PULP: A system for exploratory search of scientific literature. In J. Aslam, I. Ruthven, & J. Zobel (Eds.), Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016) (pp. 1133-1136). New York, NY, USA: Association for Computing Machinery (ACM). https://doi.org/10.1145/2911451.2911455
    Medlar, Alan ; Ilves, Kalle ; Wang, Ping ; Buntine, Wray ; Głowacka, Dorota. / PULP : A system for exploratory search of scientific literature. Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016). editor / Javed Aslam ; Ian Ruthven ; Justin Zobel. New York, NY, USA : Association for Computing Machinery (ACM), 2016. pp. 1133-1136
    @inproceedings{4e7512bac985480ab0897c1f3e999b97,
    title = "PULP: A system for exploratory search of scientific literature",
    abstract = "Despite the growing importance of exploratory search, information retrieval (IR) systems tend to focus on lookup search. Lookup searches are well served by optimising the precision and recall of search results, however, for exploratory search this may be counterproductive if users are unable to formulate an appropriate search query. We present a system called PULP that supports exploratory search for scientific literature, though the system can be easily adapted to other types of literature. PULP uses reinforcement learning (RL) to avert the user from context traps resulting from poorly chosen search queries, trading off between exploration (presenting the user with diverse topics) and exploitation (moving towards more specific topics). Where other RL-based systems suffer from the {"}cold start{"} problem, requiring sufficient time to adjust to a user's information needs, PULP initially presents the user with an overview of the dataset using temporal topic models. Topic models are displayed in an interactive alluvial diagram, where topics are shown as ribbons that change thickness with a given topics relative prevalence over time. Interactive, exploratory search sessions can be initiated by selecting topics as a starting point.",
    keywords = "Exploratory search, Topic models, Bandit algorithms, Scientific literature search, Query formulation",
    author = "Alan Medlar and Kalle Ilves and Ping Wang and Wray Buntine and Dorota Głowacka",
    year = "2016",
    month = "7",
    day = "7",
    doi = "10.1145/2911451.2911455",
    language = "English",
    isbn = "9781450340694",
    pages = "1133--1136",
    editor = "Javed Aslam and Ian Ruthven and Justin Zobel",
    booktitle = "Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016)",
    publisher = "Association for Computing Machinery (ACM)",
    address = "United States of America",

    }

    Medlar, A, Ilves, K, Wang, P, Buntine, W & Głowacka, D 2016, PULP: A system for exploratory search of scientific literature. in J Aslam, I Ruthven & J Zobel (eds), Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016). Association for Computing Machinery (ACM), New York, NY, USA, pp. 1133-1136, ACM International Conference on Research and Development in Information Retrieval 2016, Pisa, Italy, 17/07/16. https://doi.org/10.1145/2911451.2911455

    PULP : A system for exploratory search of scientific literature. / Medlar, Alan; Ilves, Kalle; Wang, Ping; Buntine, Wray; Głowacka, Dorota.

    Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016). ed. / Javed Aslam; Ian Ruthven; Justin Zobel. New York, NY, USA : Association for Computing Machinery (ACM), 2016. p. 1133-1136.

    Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

    TY - GEN

    T1 - PULP

    T2 - A system for exploratory search of scientific literature

    AU - Medlar, Alan

    AU - Ilves, Kalle

    AU - Wang, Ping

    AU - Buntine, Wray

    AU - Głowacka, Dorota

    PY - 2016/7/7

    Y1 - 2016/7/7

    N2 - Despite the growing importance of exploratory search, information retrieval (IR) systems tend to focus on lookup search. Lookup searches are well served by optimising the precision and recall of search results, however, for exploratory search this may be counterproductive if users are unable to formulate an appropriate search query. We present a system called PULP that supports exploratory search for scientific literature, though the system can be easily adapted to other types of literature. PULP uses reinforcement learning (RL) to avert the user from context traps resulting from poorly chosen search queries, trading off between exploration (presenting the user with diverse topics) and exploitation (moving towards more specific topics). Where other RL-based systems suffer from the "cold start" problem, requiring sufficient time to adjust to a user's information needs, PULP initially presents the user with an overview of the dataset using temporal topic models. Topic models are displayed in an interactive alluvial diagram, where topics are shown as ribbons that change thickness with a given topics relative prevalence over time. Interactive, exploratory search sessions can be initiated by selecting topics as a starting point.

    AB - Despite the growing importance of exploratory search, information retrieval (IR) systems tend to focus on lookup search. Lookup searches are well served by optimising the precision and recall of search results, however, for exploratory search this may be counterproductive if users are unable to formulate an appropriate search query. We present a system called PULP that supports exploratory search for scientific literature, though the system can be easily adapted to other types of literature. PULP uses reinforcement learning (RL) to avert the user from context traps resulting from poorly chosen search queries, trading off between exploration (presenting the user with diverse topics) and exploitation (moving towards more specific topics). Where other RL-based systems suffer from the "cold start" problem, requiring sufficient time to adjust to a user's information needs, PULP initially presents the user with an overview of the dataset using temporal topic models. Topic models are displayed in an interactive alluvial diagram, where topics are shown as ribbons that change thickness with a given topics relative prevalence over time. Interactive, exploratory search sessions can be initiated by selecting topics as a starting point.

    KW - Exploratory search

    KW - Topic models

    KW - Bandit algorithms

    KW - Scientific literature search

    KW - Query formulation

    UR - http://www.scopus.com/inward/record.url?scp=84980398298&partnerID=8YFLogxK

    U2 - 10.1145/2911451.2911455

    DO - 10.1145/2911451.2911455

    M3 - Conference Paper

    SN - 9781450340694

    SP - 1133

    EP - 1136

    BT - Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016)

    A2 - Aslam, Javed

    A2 - Ruthven, Ian

    A2 - Zobel, Justin

    PB - Association for Computing Machinery (ACM)

    CY - New York, NY, USA

    ER -

    Medlar A, Ilves K, Wang P, Buntine W, Głowacka D. PULP: A system for exploratory search of scientific literature. In Aslam J, Ruthven I, Zobel J, editors, Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2016). New York, NY, USA: Association for Computing Machinery (ACM). 2016. p. 1133-1136 https://doi.org/10.1145/2911451.2911455