=Paper=
{{Paper
|id=Vol-1438/paper8
|storemode=property
|title=User Controlled News Recommendations
|pdfUrl=https://ceur-ws.org/Vol-1438/paper8.pdf
|volume=Vol-1438
|dblpUrl=https://dblp.org/rec/conf/recsys/IngvaldsenGO15
}}
==User Controlled News Recommendations==
https://ceur-ws.org/Vol-1438/paper8.pdf
User Controlled News Recommendations
Jon Espen Ingvaldsen Jon Atle Gulla Özlem Özgöbek
Norwegian University of Science and Norwegian University of Science and Department of Computer Engineering,
Technology, Department of Computer Technology, Department of Computer Ege University,
and Information Science, Trondheim and Information Science, Trondheim Izmir,
Norway Norway Turkey
jonespi@idi.ntnu.no jag@idi.ntnu.no ozlem.ozgobek@ege.edu.tr
ABSTRACT location, personal interest profile and time. When designing user-
The adoption of mobile devices is pushing the Internet into a more friendly systems for mobile devices, we need to be careful about
personal and context aware space. A common challenge for online the amount of buttons and menu items introduced. In this paper
news services is to deliver contents that are interesting to read. In we will describe the news recommender system prototype and its
this paper, we describe the user interface design of the mobile user interface where the users can control their news
SmartMedia news recommender prototype. Through deep analysis stream recommendations from three toggleable buttons.
of textual news contents it is able to deliver local, recent and
personalized news experiences, and the user interface is designed
to give the users control over the news stream compositions. We 2. IMPLEMENTATION
will present its innovative user interface and the approach taken to The backend of the news recommender prototype developed is
transform raw textual data into well defined and meaning bearing constructed as a pipeline of operations harvesting and
entities. transforming Rich Site Summary (RSS) entries and raw text data
into a semantic and searchable representation. The pipeline and its
Categories and Subject Descriptors operations are implemented with using Apache Storm2. This
H.4.7 [Information Systems Applications] Communications distributed computing framework enable scalability and ability to
Applications – Information browsers handle large amounts of news items from a magnitude of
publishers continuously.
General Terms As shown in Figure 1, the news processing pipeline consists of
Algorithms, Design, Experimentation, Human Factors. five steps. The first step creates an input stream by continuously
monitoring a large set of RSS feeds. Whenever a new news item
Keywords occurs, properties such as the title, lead text and HTML sources
Recommender system, news, mobile, user interfaces, user control are extracted. The HTML sources are parsed and cleaned to
extract a representative body text. In the second step, natural
.
language processing operations such as language identification,
sentence detection and part-of-speech tagging is applied to extract
entity mentions from the textual data. The third step uses
1. INTRODUCTION supervised models to map entity mentions to referent entities in
The Smartmedia project1 at NTNU targets construction of context the WikiData3 and Geonames4 knowledge bases. These models
aware news experiences based on deep understanding of text in combine textual similarities, graph relations and entity frequency
continuous news streams [4, 9]. The goal of the Smartmedia and co-occurrence statistics to classify the relevance of multiple
project is to deliver a mobile and context aware news experience referent candidates. First Story Detection (FSD) is applied in the
based on deep understanding of textual contents, combining both forth step to group news items describing the same news story. In
geo spatial exploration and context aware recommendations. The the fifth step this semantic representation is indexed and made
system is designed with scalability in mind and ability to support searchable. As this backend architecture is stream based, it is able
multiple languages. to index and promote recent news items.
Privacy is an important aspect when engineering recommender WikiData is the community-created knowledge base of Wikipedia
systems and exploitation of user interaction and context data. [12]. Since its public launch in 2012, the knowledge base has
When dealing with personal data and privacy, transparency tools gathered more than 15 millions entities, including more than 34
are tools that can provide to the concerned individual clear million statements and over 80 million labels and descriptions in
visibility of aspects relevant to these data and the individual’s more than 350 languages [3]. Most geographical entities in
privacy. The combination of transparency tools and user control WikiData provide a reference to Geonames containing more
yields viable functionality to empower users to protect their detailed geographical properties. In the implementation of the
privacy [5]. Smartmedia prototype, the news and entity information including
In the Smartmedia project, we want to build transparent news news text, titles, publication timestamps, entity labels and
recommender systems where the user can control gathered data
and how their news streams are composed based on geo spatial 2
http://storm.apache.org/
3
https://www.wikidata.org/
1 4
http://research.idi.ntnu.no/SmartMedia https://www.geonames.org/
geospatial properties are indexed in a Lucene based search index. personal interest factor is disabled, the user retrieve a news
This index makes the news items and their related entities composition which is general and without such bias.
searchable and creates a foundation for detailed querying.
To customize the time-factor, the user is presented with a calendar
When a user is opening the news app on the mobile a request where it is possible to move in time and retrieve either recent or
containing user id, location and preferences are sent to the historic news items. When, the time-factor is disabled the user
backend. Here, a multi factor search query is formed to retrieve will retrieve news solely based on the other relevance factors
relevant news entries from the index. (location and personal interests).
Figure 2b shows an example of how news stories are presented.
Here we see one news article “Theresa May urges media restraint
3. USER INTERFACE in coverage of terror suspects” from the Guardian about politics
A web-based user interface is developed to make the news stream and terror, followed by another news story from BBC. The three
contents explorable on mobile devices. In this interface, the user is circular buttons on the bottom of the screen allow users to toggle
allowed to extract news items that are relevant to the geo special whether their locality, personal interest profile and time setting
locality context, personal interests and given point of time. These such influence news story retrieval.
three relevance factors are customizable and the user can select
whether or not they should influence the retrieval and ranking of By clicking on a news story, the user gets the ingress of the news
available news items. story and a list of the most salient entities for the selected news
story. Figure 1c shows the ingress and relevant WikiData entities
To customize the geographical locality, the user specifies a from the news article about Theresa May. As we can see, our
circular relevance region on a map. Figure 2a shows an example news story about politics and terror related to Syria, Theresa May,
of such a relevance region. By default, the relevance region is set ISIL and Sky News. By hovering these items, the user is
to users current GPS location with a 50 km radius. By moving the presented with their textual WikiData description. On figure 2c,
region or modifying the radius, users can generate a local we can see that the WikiData entity for Theresa May contains the
newspaper for any region of the world. If the location factor is description “British politician”.
disabled, it means that the system is recommending news from
any location in the world and news that are not containing In general, the three buttons at the bottom of the screen for
location information. location, interest profile and time can at any time be activated and
de-activated to provide very different recommendation strategies.
In the current Smartmedia prototype, we have predefined a For example, keeping all buttons active with default parameters
handful of user interest profiles. Examples of such profiles are means that the system will recommend news articles that have
stock trader, soccer fan, technology geek, etc. Each profile recently takes place in the vicinity of the reader and are consistent
consists of a weighted concept vector, where each entry is a with her profile. Figure 3 describes different combinations of
WikiData entry associated with an interest score between 0 and 1. recommendation factors and summarizes how the user can control
By selecting any of these interest profiles, the retrieved news will the retrieval and composition of news items.
be influenced and biased towards the interest topics. When the
Figure 1. Steps of the news stream processing pipeline.
a) b) c)
Figure 2. Screenshots from the Smartmedia prototype. a) The map query interface. b) Presentation of news stories. c) Presentation
of news details.
Both these NewsStand and News@Hand have user interfaces
4. RELATED WORK targeting desktops and larger device screens. They both provide
People nowadays have access to more worldwide news
user control over the retrieved set of news, either through a map
information than ever before. As Internet services get more
or category based navigation or preferences settings.
information about their users and their context, they can deliver
personal and customized contents and user experiences. Tran and Herder [11] have looked at the studied news event
timelines and shown that manually constructed timelines are
The prototype system, described in this paper, share similarities to
subjective and often missing important dates or other information.
other academic news applications such as NewsStand [8, 10] and
By complementing the timelines with elements extracted
News@Hand [1, 2]. Both these systems map textual news
algorithmically from multiple sources, it is possible to create more
contents to entities defined in a knowledge base.
objective and argumentative timelines. However, the manual
NewsStand targets geo spatial exploration of news. It is an processing and editing efforts are still needed to enhance the
example application of a general framework developed to enable communicative qualities of the timelines, and to adapt it to the
people to search for information using a map query interface. It needs of the readers
utilize maps both to explore and find news stories and to visualize
Parra et al. [6, 7] presents SetFusion, a visual user-controllable
and present single news events.
interface for hybrid recommender system. Their approach enables
News@hand combines textual features and collaborative users explore and control the importance of recommender
information to make news suggestions. It uses Semantic Web strategies using an interactive Venn diagram visualization. Their
technologies to describe the news contents and user preferences. evaluations indicate that this interface had a positive effect on the
Both news items and user profiles are represented in terms of user experience and improved users engagement. Their idea of
concepts appearing in domain ontologies, and semantic relations using the Venn diagram to explain intersections among
among those concepts are exploited to enrich the above recommendation approaches is transferable and valuable to the
representations, and enhance recommendations. news domain.
[2] Cantador, I. et al. 2008. Ontology-based personalised and
context-aware recommendations of news items.
Proceedings of the 2008 IEEE/WIC/ACM International
Conference on Web Intelligence and Intelligent Agent
Local Technology. 1, (2008).
Retrieve news from a geospatial area
[3] Erxleben, F. et al. 2014. Introducing Wikidata to the
Personal Linked Data Web. The Semantic Web–ISWC 2014.
Retrieve news matching the interest profile of (2014).
the user
Temporal
Retrieve news published after a given date [4] Gulla, J.A. et al. 2013. Learning User Profiles in Mobile
News Recommendation. Journal of Print and Media
Local and personal Technology Research. II, 3 (2013), 183–194.
Retrieve news that both relate to a geospatial
area and match users interest profile.
[5] Hansen, M. 2008. Marrying transparency tools with user-
Local and temporal
controlled identity management. The Future of Identity in
Retrieve news related to a given geospatial
area and published after a given date. the Information Society. (2008).
Personal and temporal
Retrieve news matching the interest profiles [6] Parra, D. et al. 2014. See what you want to see.
of the user and published after a given date. Proceedings of the 19th international conference on
Local, personal and temporal Intelligent User Interfaces - IUI ’14 (New York, New
Retrieve news with relevance for the selected York, USA, Feb. 2014), 235–240.
geospatial area and interest profile, and
published after a given date.
[7] Parra, D. and Brusilovsky, P. 2015. User-controllable
personalization: A case study with SetFusion.
International Journal of Human-Computer Studies.
Figure 3. Combinations of selectable recommendation factors
(2015).
[8] Samet, H. et al. 2014. Reading news with maps by
exploiting spatial synonyms. Communications of the
5. CONCLUSIONS AND FUTURE WORK ACM. 57, 10 (Sep. 2014), 64–77.
The predefined user profiles can be replaced or used in
combination with more personal profiles trained on traced
[9] Tavakolifard, M. et al. 2013. Tailored news in the palm
interaction logs from the system. As users leave interaction data
of your hand: a multi-perspective transparent approach to
behind, we can gather knowledge about what the users interests
news recommendation. (May 2013), 305–308.
are. However, for new users where no past interaction records
exist, we have a cold-start problem where we still benefit on
predefined stereotypes. [10] Teitler, B. and Lieberman, M. 2008. NewsStand: A new
view on news. Proceedings of the 16th ACM
In future work we plan to use trained personal profiles with
SIGSPATIAL international conference on Advances in
predefined stereotypes in combination. We will also gather user
geographic information systems. (2008).
feedback and evaluate to which extent users want to control and
customize their news presentations and study how their
requirements can be met in a mobile user interface design. [11] Tran, G. and Herder, E. 2015. Detecting Filter Bubbles in
Ongoing News Stories. Extended Proc. UMAP 2015.
Deep understanding of textual contents together with knowledge
(2015).
base structures provides a fundament for innovative and
intelligent applications. This paper has described one such
innovation from the news domain, and how its mobile user [12] Vrandečić, D. and Krötzsch, M. 2014. Wikidata: a free
interface allow users to control the composition of news. A collaborative knowledgebase. Communications of the
screencast video demonstrating the prototype and its user interface ACM. (2014).
is available at: http://vimeo.com/121835936
6. REFERENCES
[1] Cantador, I. et al. 2008. News@ hand: A semantic web
approach to recommending news. Adaptive hypermedia
and adaptive web-based systems. (2008).