LINGUINI - acquiring individual interest profiles by means of adaptive natural language dialog [Elektronische Ressource] / Rosmary Stegmann

Rosmary Stegmann

Sujets

• Informatik

Informations

• Publié par: technische_universitat_munchen
• Publié le: 01 janvier 2006
• Nombre de lectures: 16
• Langue: English
• Poids de l'ouvrage: 2 Mo

Extrait

Fakultät für Informatik
der Technischen Universität München



LINGUINI
–
Acquiring Individual Interest Profiles
by Means of Adaptive Natural Language Dialog


Rosmary Stegmann


































Fakultät für Informatik
der Technischen Universität München

Lehrstuhl Informatik XI




LINGUINI
–
Acquiring Individual Interest Profiles
by Means of Adaptive Natural Language Dialog



Rosmary Stegmann





Vollständiger Abdruck der von der Fakultät für Informatik der Technischen Universität
München zur Erlangungen des akademischen Grades eines
Doktors der Naturwissenschaften (Dr. rer. nat.)
genehmigten Dissertation.

Vorsitzender: Univ.-Prof. Dr. Uwe Baumgarten
Prüfer der Dissertation: 1. Univ.-Prof. Dr. Johann Schlichter
2. Univ.-Prof. Dr. Manfred Pinkal
Universität des Saarlandes


Die Dissertation wurde am 30.03.2006 bei der Technischen Universität München
eingereicht und durch die Fakultät für Informatik am 19.07.2006 angenommen.





























Abstract



User information is needed by adaptive systems in order to tailor information and product
offers to the needs and preferences of individual users. Personalized Recommender Systems
are adaptive systems that automatically generate recommendations on the basis of individual
user profiles. Most existing Recommender Systems, however, are based on rather simple and
mainly standardized profile information, which often delimits the adequacy of the recommen-
dations they generate for an individual user. More adequate recommendations could be gener-
ated on the basis of more individual and representative user profiles that also integrate com-
plex information, for example about personal interests or lifestyle. Furthermore, most adap-
tive systems acquire profile information only for their own purposes and do not allow for an
exchange of this information with other applications the user wants to use. Above all, existing
explicit profiling methods suffer from severe drawbacks which limit their utilizability in prac-
tice. Especially for mobile scenarios, in which a spoken language interaction with the user is
required, no suitable explicit profiling methods exist as yet that integrate a solution for all of
the above mentioned problems.
This thesis presents a solution for acquiring detailed information about personal interests
of users by means of an adaptive natural language dialog. We have developed a comprehen-
sive explicit profiling framework, LINGUINI, which integrates a dialog management and
profile management approach. Because of the natural language processing methods applied,
this profiling approach is especially suitable for situations in which spoken language is re-
quired (e.g. in a vehicle), but it is also applicable with a user interface for typed input and
output (e.g. for Internet and E-Commerce platforms). The acquired information can be used
by various types of adaptive systems for which user interests are relevant.
During our profiling dialog, users are able to formulate their interests in their own words.
The dialog adapts to each user individually and is able to find and talk about new interests
related to the interests already mentioned by the user. The dialog management approach inte-
grates a sociological target group model that clusters users into groups according to their in-
terests. The groups do not serve as user profiles, however, but are used for providing clues
about suitable next questions or related topics. With this adaptive approach, we are able to
create truly personalized profiles that are different for each user in contents and structure. By
employing the lexical-semantic network GermaNet, our profiling approach allows for repre-
senting interests in a semantically structured way and for interpreting and storing new user
information dynamically that has not been predefined in the user model before.
We implemented our adaptive profiling approach as a comprehensive prototype system
and evaluated it by means of a user study which investigates user acceptance, dialog adapta-
bility, and profile quality. The study shows that users, in fact, appreciate the adaptive capabili-
ties of the profiling system. The users’ willingness to apply the system is high and they con-
sider this approach very suitable for a variety of mobile and non-mobile situations and adap-
tive applications.


iii
















































iv

Acknowledgements



This work emerged from my activities as a research assistant at the chair for Informatics XI:
Applied Informatics/Cooperative Systems (Prof. Johann Schlichter) at the Technische Univer-
sität München.
First, I would like to thank my supervisor Johann Schlichter for providing the opportunity
to work on this interesting research topic, and my second supervisor Manfred Pinkal, espe-
cially for his support with respect to the Natural Language Processing aspects of this work.
Their helpful suggestions and comments greatly contributed to the success of this thesis.
I also thank my colleagues for the friendly and cooperative working atmosphere. In par-
ticular I would like to thank Wolfgang Wörndl for constantly encouraging me and for discuss-
ing and commenting on large parts of this work. I also thank Michael Koch, from whom I
learned a lot, for helping me with the initiation of this project. I cannot thank enough Georg
Groh and Michael Galla for their support during the preparation of my mathematics exam,
which was a prerequisite for the dissertation in Informatics. Special thanks also to Thomas
Leckner for the excellent cooperation during several years of common work on personaliza-
tion and recommender systems.
I am very grateful to the students who contributed to the prototype implementation of
LINGUINI, especially to the incomparable commitment of Kristof Unterweger and Manuel
Giuliani. Many thanks also to the participants of the user study for patiently answering LIN-
GUINI’s nosy questions and for the multitude of valuable comments and suggestions they
provided in their questionnaires.
I would also like to thank the researchers from other institutions for their support and the
many inspiring and visionary discussions, which encouraged me to dare the impossible. Rep-
resentative for all of them, these are Mathias Bauer, Aljoscha Burchardt, Katrin Erk, Alexan-
der Felfernig, Gerd Fliedner, Iryna Gurevych, Anne Hackett, Dominik Heckmann, Elco
Herder, Vera Hollink, Anthony Jameson, Ivana Kruijff-Korbayova, Jonas Kuhn, Claudia
Kunze, Martin Lacher, Lothar Lemnitzer, Alexandros Paramythis, Bodo Polzer, Helmut
Schmid, Sebastian Stegmann, Stefan Weibelzahl, and Ingrid Zukerman.
I am very grateful to Colleen Gruban who helped me proofread this work and improve my
English. Last but not least, my deep gratitude belongs to Patrick, my family and friends for
their understanding and great support in all situations.











v




























vi

Contents


LIST OF FIGURES ............................................................................................................................................XI
LIST OF TABLES ...........................................................................................................................................XIII
LIST OF ACRONYMS .....................................................................................................................................XV

1 INTRODUCTION............................. 1
1.1 BACKGROUND .............................................................................................................................................. 1
1.1.1 Adaptive Systems and Personalized Recommender Systems ............................................................... 1
1.1.2 User Models......................................................................................................................................... 1
1.1.3 Acquiring Information about Users..................................................................................................... 2
1.1.4 Acquiring Information and Generating Recommendations in a Mobi