Project

Back to overview

Relevance Criteria Combination for Mobile Information Retrieval (RelMobIR)

English title Relevance Criteria Combination for Mobile Information Retrieval (RelMobIR)
Applicant Crestani Fabio
Number 156864
Funding scheme Project funding (Div. I-III)
Research institution Istituto del Software (SI) Facoltà di scienze informatiche
Institution of higher education Università della Svizzera italiana - USI
Main discipline Information Technology
Start/End 01.03.2015 - 28.02.2019
Approved amount 245'252.00
Show all

Keywords (4)

information retrieval; user and context modelling; document relevance; mobile information access

Lay Summary (Italian)

Lead
This proposal addresses the theoretical foundations of context-awareness for mobile Information Retrieval (IR) by studying a fundamental aspect: relevance. In particular, it addresses the problem of combining different relevance criteria for mobile IR (e.g. topical, geographical, temporal, etc.), in order to determine the effective overall relevance of a document to a user information need.
Lay summary

I telefoni cellulari sembrano fornire grandi opportunità per l'accesso efficace alle informazioni, data la loro maggiore capacità di personalizzazione e localizzazione. Tuttavia il lavoro teorico dietro l'uso efficace di queste caratteristiche sta registrando un sostanziale ritardo. Questo progetto  mira ad affrontare i fondamenti teorici della contestualizazione della mobile Information Retrieval (IR) studiandone un aspetto fondamentale: il concetto di rilevanza. Infatti, i sistemi IR mobili sensibili al contesto (context-aware) sono in grado di stimare la rilevanza di una determinata pagina in relazione a diversi aspetti. Il primo aspetto è la rilevanza al topic (topicality) stessa per il quale esiste già una lunga storia di ricerca. Tuttavia, altri tipi di rilevanza svolgono un ruolo molto importante nell'IR mobile, ad esempio la rilevcanza geografica, la rilevanza temporale o la rilevanza all'opinione. Mentre alcuni tentativi sono stati effettuati per tentare di combinarli. Questa proposta affronta il problema di combinare diversi criteri di rilevanza per IR mobile, al fine di determinare la rilevanza complessiva di un documento al bisogno di informazioni dell'utente.

Direct link to Lay Summary Last update: 08.03.2015

Responsible applicant and co-applicants

Employees

Publications

Publication
Understanding Mobile Search Task Relevance and User Behaviour in Context
Aliannejadi Mohammad, Harvey Morgan, Costa Luca, Pointon Matthew, Crestani Fabio (2019), Understanding Mobile Search Task Relevance and User Behaviour in Context, in the 2019 Conference, Glasgow, Scotland UKACM, New York, USA.
Personalized Context-Aware Point of Interest Recommendation
Aliannejadi Mohammad, Crestani Fabio (2018), Personalized Context-Aware Point of Interest Recommendation, in ACM Transactions on Information Systems, 36(4), 1-28.
A Collaborative Ranking Model with Multiple Location-based Similarities for Venue Suggestion
Aliannejadi Mohammad, Rafailidis Dimitrios, Crestani Fabio (2018), A Collaborative Ranking Model with Multiple Location-based Similarities for Venue Suggestion, in the 2018 ACM SIGIR International Conference, Tianjin, ChinaACM, New York, USA.
Target Apps SelectionTowards a Unified Search Framework for Mobile Devices
Aliannejadi Mohammad, Zamani Hamed, Crestani Fabio, Croft W. Bruce (2018), Target Apps SelectionTowards a Unified Search Framework for Mobile Devices, in The 41st International ACM SIGIR Conference, Ann Arbor, MI, USAACM, New York, USA.
A Collaborative Ranking Model with Contextual Similarities for Venue Suggestion
Aliannejadi Mohammad, Crestani Fabio (2018), A Collaborative Ranking Model with Contextual Similarities for Venue Suggestion, in Italian Information Retrieal Workshop, CEUR Workshop Proceedings, Germany.
In Situ and Context-Aware Target Apps Selection for Unified Mobile Search
Aliannejadi Mohammad, Zamani Hamed, Crestani Fabio, Croft W. Bruce (2018), In Situ and Context-Aware Target Apps Selection for Unified Mobile Search, in the 27th ACM International Conference, Torino, ItalyACM, New York, USA.
Venue suggestion using social-centric scores
AliannejadiMohammad (2018), Venue suggestion using social-centric scores, in ECIR Workshop on Social Aspects in Personalization and Search, Grenoble, FrancearXiv, USA.
A Cross-Platform Collection for Contextual Suggestion
Aliannejadi Mohammad, Mele Ida, Crestani Fabio (2017), A Cross-Platform Collection for Contextual Suggestion, in the 40th International ACM SIGIR Conference, Shinjuku, Tokyo, JapanACM, New York, USA.
Venue Appropriateness Prediction for Personalized Context-Aware Venue Suggestion
Aliannejadi Mohammad, Crestani Fabio (2017), Venue Appropriateness Prediction for Personalized Context-Aware Venue Suggestion, in the 40th International ACM SIGIR Conference, Shinjuku, Tokyo, JapanACM, New York, USA.
Personalized Keyword Boosting for Venue Suggestion Based on Multiple LBSNs
Aliannejadi Mohammad, Rafailidis Dimitrios, Crestani Fabio (2017), Personalized Keyword Boosting for Venue Suggestion Based on Multiple LBSNs, in European Conference on Information Retrieval, Aberdeen, UKSpringer International Publishing, Cham.
Personalized ranking for context-aware venue suggestion
Aliannejadi Mohammad, Mele Ida, Crestani Fabio (2017), Personalized ranking for context-aware venue suggestion, in the Symposium, Marrakech, MoroccoACM, New York, USA.
Venue Appropriateness Prediction for Contextual Suggestion
AliannejadiMohammad, MeleIda, CrestaniFabio (2016), Venue Appropriateness Prediction for Contextual Suggestion, TREC, Gaithersburgh, USA.
University of Lugano at TREC 2015: ContextualSuggestion and Temporal Summarization Tracks
AliannejadiMohammad, BahrainianSeyed Ali, GiachanouAnastaisa, CrestaniFabio (2015), University of Lugano at TREC 2015: ContextualSuggestion and Temporal Summarization Tracks, TREC, Gaithersburgh, USA.
A Joint Two-Phase Time-Sensitive Regularized Collaborative Ranking Model for Point of Interest Recommendation
AliannejadiMohammad, RafailidisDimitrios, CrestaniFabio, A Joint Two-Phase Time-Sensitive Regularized Collaborative Ranking Model for Point of Interest Recommendation, in IEEE Transactions on Knowledge and Data Engineering (TKDE), 1.

Datasets

TREC-CS Appropriateness

Author Aliannejadi, Mohammad; Mele, Ida; Crestani, Fabio
Publication date 07.08.2017
Persistent Identifier (PID) 10.1145/3077136.3080752
Repository contextual-appropriateness
Abstract
The contextual appropriateness collection consists of 1,969 pairs of trip descriptors and venue categories as features. In order to enable researchers to train their models using the contextual appropriateness of venues, we created another collection providing ground truth assessments for the contextual appropriateness of the venue categories. It completes the contextual information (i.e., trip type, group type, trip duration) for 10% of the whole TREC collection. This collection contains 760 rows including the features we already created using crowdsourcing and the context-appropriateness labels for venues. The 10% of labeled data allows to model the venues’ contextual appropriateness given the users’ context and to make prediction for the remaining 90% of the data. Below, you can see a histogram of venue-context appropriateness score ranges. We partition the histogram into 3 parts based on the scores range. Scores below −0.4 represent inappropriateness and score higher than +0.4 represent appropriateness. Scores between −0.4 and +0.4 do not provide much information and show no agreement among assessors (subjective task).

ISTAS: an in situ collection of cross-app mobile search queries

Author Aliannejadi, Mohammad; Zamani, Hamed; Crestani, Fabio; Croft, W. Bruce
Publication date 01.10.2018
Persistent Identifier (PID) 10.13140/RG.2.2.18890.82884
Repository istas
Abstract
In this data collection, we are particularly interested in providing the first dataset focusing on a unified search framework for mobile devices by collecting cross-app mobile queries as well as their target apps. To this end, we recruited 255 participants through an open online call asking them to install uSearch on their smartphones and let it run for at least 24 hours. During the study, we asked the participants to report their mobile searches using uSearch as soon as they did the search. A search report consisted of the search query as well as the app in which the search was done.

UniMobile: a collection of cross-app mobile search queries

Author Aliannejadi, Mohammad; Zamani, Hamed; Crestani, Fabio; Croft, W. Bruce
Publication date 01.10.2018
Persistent Identifier (PID) 10.13140/RG.2.2.15535.38565
Repository unimobile
Abstract
In this data collection, we are particularly interested in providing the first dataset focusing on a unified search framework for mobile devices by collecting cross-app mobile queries as well as their target apps. To this end, we initially asked crowdworkers to explain their latest search experience on their smartphones and used them to define various realistic mobile search tasks. Then, we asked another set of workers to select the apps they would choose to complete the tasks as well as the query they would submit. One of the key takeaway findings is that for the majority of the search tasks, most of the users prefer not to use Google Search. Here, we release the dataset for research purposes.

Collaboration

Group / person Country
Types of collaboration
Center for Intelligent Information Retrieval, Universirty of Massachusetts Amherst United States of America (North America)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Exchange of personnel
Faculty of Informatics/Universita' di Udine, Italy Italy (Europe)
- in-depth/constructive exchanges on approaches, methods or results

Awards

Title Year
PhD to Mohammad Aliannejadi, the student that worked in the project 2019

Abstract

With the tremendous increase in the use of mobile phones worldwide recent years have seen a great increase in their use to access information on the web. While all major search engines have released versions of their search platforms for mobile phones, their performance has not seen as great an increase as we whould have expected. In fact, while mobile phones seem to provide great opportunities for more effective IR, given their greater personalisation and localisation capabilities, the theoretical work behind the effective use of these characteristics has been lagging behind. So, while a few researchers worldwide have recognised the potential and have started working on context-awareness for mobile IR, their progress on a very important aspect of this area of research has been slow. This proposal aims at addressing this fundamental aspect: relevance. This proposal addresses the problem of combining relevance criteria for mobile IR, in order to determine the effective overall relevance of a document to a user information need. The work proposed will start by addressing how to best estimate these different relevance criteria for a document, given what is know of the user’s opinions, preferences and his spatio-temporal context. It will then address the problem of the best combination of these relevance estimates, under the assumption of their independence from each other. The system will be progressively improved removing this simplifying assumptions.
-