-
1. Abstract
tion and utilisation of the neural processes that underpin
the IS&R process, termed “NeuraSearch”, has recently
Information Retrieval (IR) process is complex because it drawn increasing interest in the Information Retrieval
involves a gap between the representation of an Informa- and Science communities [
11, 12, 13, 14, 15, 16
]. This
tion Need (IN) (i.e. the formulated query) and the actual interest has focused on gaining an understanding of how
IN. This gap can become widen when searchers are expe- the diferent components of IR emerge from measurable
riencing an ill-defined IN. As a result of this phenomenon, activity in the brain. These studies have employed a wide
searchers were left unsatisfied with the results obtained range of brain imaging techniques to probe brain activity
in response to their initial retrieval formulation [
1
], and related to brain states involved in processing relevance
must engage in further interaction with the system to and information need. Moshfeghi et al. [
11, 17, 18, 19, 20
]
resolve their needs. conducted a series of studies using fMRI to understand
In the past, to close the gap between an actual IN and brain regions activated during relevance judgement and
its representation, IR systems have employed feedback information need. Results from one study [
11
] revealed
techniques. An example of such an approach is the rel- that brain regions, including the inferior parietal lobe,
evance feedback technique where feedback is gathered inferior temporal gyrus, and superior frontal gyrus, are
through explicit [
2
], implicit [
3
], afective and/or physi- activated during the relevance judgement process. Also,
ological feedback [
4
]. Despite the robustness of explicit another study from this series [
17
] indicated that IN
refeedback in improving retrieval efectiveness [
2
], it is not lfected a process of switching between internal and
exalways applicable or reliable due to the cognitive burden ternal information sources. Another study [
19
] revealed
that it places on users [
5
]. To overcome this cognitive how transitions between diferent segments of an
inforburden, implicit feedback is proposed where relevance is mation search task were reflected in activity changes in
inferred from the interactional data indirectly and unob- large-scale brain networks. While having a high spatial
trusively [
6
]. For example, researchers try to understand resolution, the fMRI technique does not have a high
temhow task [
3
], dwell time [
7
] and click-through [
6
] relate poral resolution. Thus such techniques are instrumental
to relevance. However, a problem occurs when actions in localising brain regions associated with phenomena
are taken as an indication of relevance without suficient such as relevance judgement or information need
realievidence to support their efectiveness [
8
]. For exam- sation rather than monitoring their changes in real-time.
ple, Kelly and Belkin [
9
] show that the implicit feedback Apart from the fMRI technique, other studies have
measures based on user interaction with the full con- applied diferent brain measurement techniques in
ortent of documents can often be unreliable and dificult to der to investigate the concept of IR, especially relevance.
measure or interpret. For example, the technique of Magnetoencephalography
Our position is that if we can monitor brain regions (MEG) has been used to understand the concept of
releactivated during an Information Seeking and Retrieval vance of visual information [
14
]. The related technique
(IS&R) process, we can reduce (or possibly eliminate) the of Electroencephalography (EEG) has also been applied
gap between a formulated query and its IN. This would in several studies to investigate the concept of relevance
then help searchers better satisfy their IN. To do so, we to text information [
21, 13, 22, 23, 24, 25
]. EEG technique
need to be able to identify brain regions activated from has high temporal resolution but lacks high spatial
resthe early stages of an IN (i.e. at its visceral level [
10
]) un- olution. Thus such techniques are crucial for real-time
til the termination of the IS&R process, either as a result monitoring of the phenomena of interest.
of IN satisfaction or search abandonment. The identifica- With the encouraging results obtained from these
studies, the IR community now better understands the
possibilities and limitations of employing neuroscience for IR.
However, this is still an open research problem.
Therefore, it is important to further study the brain activity
underlying IR processes and how they can be harnessed
and used to improve and help searchers in their search
tasks and experience. This would be possible via a
collabDESIRES 2021 – 2nd International Conference on Design of
Experimental Search & Information REtrieval Systems, September
15–18, 2021, Padua, Italy
" yashar.moshfeghi@strath.ac.uk (Y. Moshfeghi)
~ http://academic.yashmosh.com (Y. Moshfeghi)
0000-0000-0000-0000 (Y. Moshfeghi)
© 2021 Copyright for this paper by its authors. Use permitted under Creative
CPWrEooUrckReshdoinpgs IhStpN:/c1e6u1r3-w-0s.o7r3g CCoEmmUoRns LWiceonsrekAstthribouptionP4r.0oIncteerenadtiionnagl s(CC(CBYE4U.0)R.-WS.org)
orative efort to understand what exactly happens inside
the human brain in real-time while they are performing a
search process, from the realisation of an IN, till stopping
a search process, including engaging, comprehending,
perceiving, processing and judging the information they
encounter during this process and how they ultimately
contribute to the satisfaction of the realised IN.