Exploratory Search:

deep dives and rabbit holes, or inadequate technology?

What we’re doing today:

What is Exporatory Search?

Past work

Current investigations

What is exploratory search (ES)?

ES is:

  • as a task

  • Its ill-formed

  • repetitive, iterative

  • long-term

  • simultaneously broad and narrow

Overall:

  • The state-of-the-art in ES is woefully behind. Over the last 10 years, ESS has been left behind with bag-of-word approaches

  • As a task, there is a need for natural language query, visual query with task completion integrating with IR.

  • Ingwesen & Järvelin’s cognitive model of IS&R is most appropriate and integrates concepts of ES.

  • Lack of connection or resolution with scientometric theories

  • Dependency on self-reported data from questionnaires, and a lack of longitudinal studies, field studies.

  • There are few proxies for relevance and this impacts our conceptualization of relevance.

  • ES as a separate focus of application development is a dead end.

1

Intro

The problem space:

ES systems (ESS) have not been developed that address bias in citation behaviours

There are few available interfaces that support the ES task

Current information retrieval (IR) applications reinforce‘top of the list’behaviours

Master’s work

Co-occurrence in keywords

How might seeing this help us to understand how terms overlap?

What might I be missing when searching using keywords?

PhD work

VR studies

Creating digital twin using real-time ocean wave height data

And another on VR authorization

LIS in Canada paper

OA in the IR of the U15

Working at Crossref!

Comprehensive exam 1

Research questions:

RQ1: What are the ES theories/frameworks within LIS and HCI respectively, and are there any that specifically seek to bridge these two interdisciplinary fields?

RQ2: For exploratory search, what is the current state of the art in visualized exploratory search results in 2D, 2.5D, and 3D?

RQ3: What methods were employed, and what are the limitations and opportunities for advancement of knowledge?

Spoilers

  • The state-of-the-art in ES is woefully behind. Over the last 10 years, ESS has been left behind with bag-of-word approaches

  • As a task, there is a need for natural language query, visual query with task completion integrating with IR.

  • Ingwesen & Järvelin’s cognitive model of IS&R is most appropriate and integrates concepts of ES.

  • Lack of connection or resolution with scientometric theories

  • Dependency on self-reported data from questionnaires, and a lack of longitudinal studies, field studies.

  • There are few proxies for relevance and this impacts our conceptualization of relevance.

  • ES as a separate focus of application development is a dead end.

Comprehensive exam 2

study 1: IS&R methods and metadata quality

  • purpose: how the metadata affects discoverability
  • contributes to: CS and IS for metadata
  • problem statement: its unknown how metadata quality has been impacting IR in discovery systems and how it may affect systems built on LLMs

source:https://weaviate.io/assets/images/fig1-cef010de0310628aa3e0a1373b33c8e9.png

study 2: visual query and RAG IS&R method

  • purpose: explores use of RAG IR method for connecting users to body of literature grounded in pre-existing corpus using visual query
  • contributes to: HCI for understand visual representation and visual query, IR for information behaviours with RAG IR and visual query; IS for information representation
  • problem statement:There are limited ways to connect disconnected corpuses

Thank you!

Interested in this research area?

Poppy Riddle pnriddle@dal.ca