General Conclusion - Geographical Information Retrieval in Textual Corpora - page 91

4) Disambiguation of spatial features (SFs) [LEI 11].
5) Interpretation of the semantics carried by spatial, temporal and thematic
references [MAL 00, PER 07, SCH 07, BAT 11, MAT 11]. For example, Leidner and
Lieberman[LEI 11]aswellasPurvesandJones[PUR 11]showinterestinSFssuchas
“40kmtothenorthofBerlin”,whichdescribesavaguezonethathastobeinterpreted
by taking into account the meaning carried by “40 km”, “north” and the place of
reference “Berlin”.
6) Indexingofdocumentsbytakingintoaccountthethreegeographicdimensions:
for example, Vaid et al. [VAI 05] as well as Purves and Jones [PUR 11] show interest
in the spatial and thematic dimensions.
7) Matching and ranking of relevant results based on one or more geographic
dimensions [VAI 05, LIN 06, BUS 09b, URE 10, ALO 11, MAT 11, EGO 11,
FER 11].
8) Development of user interfaces [PUR 11].
9) Conception of frameworks and resources devoted to the evaluation of
geographic information retrieval system (GIRSs) [MAN 11, CAR 11, PUR 11].
In the T2I team of the laboratory of LIUPPA, we have chosen to work with a
priorityonspaceandtime,whicharegenerallyeasiertoprocessentrypoints,because
they are more stable in the language (e.g. recognition of named entities) and well
formalized,whenworkingonseveraldimensionsofgeographicinformationincluding
the theme.
As part of this reflection, which consists of proposing perspectives, we propose
extending these pieces of work following two main axes that we will qualify as being
intra- and interdimensional.
Concerning the intradimensional axis, we propose reworking on our models of
representation and reasoning. For the thematic dimension, we will review the theme,
initially dealt with as a set of terms, and we propose a thematic tiling based on the
definition of domain ontologies. For the spatial dimension, we are working on the
approximation of the spatial information by more accurate representations than
bounding boxes. This reflection targets challenges 2 and 5 mentioned earlier.
Concerning the interdimensional axis, we are working on a challenge identified
[ALO 11] that, in temporal and thematic IR, raises the following question “should a
relevant document in which the textual match and the temporal match are far away
from each other be penalized?”. It is mainly a question of identifying the
relationships between spatial, temporal or thematic information detected in a text.
This reflection targets challenge no. 7. Finally, we oppose an approach that we
qualify as qualitative against the specialized IRS federating approaches. From