PosterPDF Available

Applying Situation-Person-Driven Semantic Similarity On Location-Specific Cognitive Frames For Improving Location Prediction

Poster

Applying Situation-Person-Driven Semantic Similarity On Location-Specific Cognitive Frames For Improving Location Prediction

Abstract

In this paper, within the scope of optimizing location prediction systems and in line with the Ontology Design Pat- tern technique, we propose a way to encapsulate the user’s transient experience when visiting locations into n-ary relational objects, which we call Location-Specific Cognitive Frames.
Applying Situation-Person-Driven Semantic
Similarity On Location-Specific Cognitive
Frames For Improving Location Prediction
Antonios Karatzoglou1,2and Michael Beigl2
1Robert Bosch GmbH,
Corporate Sector Research and Advance Development,
antonios.karatzoglou@de.bosch.com,
2Karlsruhe Institute of Technology,
{michael.beigl, antonios.karatzoglou}@kit.edu,
Germany
In recent years, there has been an increase in demand for personalization,
user-tailored solutions and context awareness. When one thinks of personaliza-
tion, it becomes apparent how important it is to understand the users’ individual
point of view, how users perceive and understand the world. Minsky introduced
in the 1970s’ so called (cognitive) Frames to capture and represent the personal,
experience- and situation-gained knowledge of individuals [1]. Today, however
most applications rely solely on a user’s predefined or learned set of preferences.
This makes them incapable of reflecting the aforementioned personal dynamics
and thus they are in need of improvement. In this paper, within the scope of opti-
mizing location prediction systems and in line with the Ontology Design Pattern
technique, we propose a way to encapsulate the user’s transient perception of
locations and define our own Location-Specific Cognitive Frames. Furthermore,
we apply semantic similarity analysis methods in order to build dynamic and
user-centered location-specific constructs instead of using a rigid location tax-
onomy as a basis for all the users. We use the resulting constructs to improve
the location prediction accuracy.
Location prediction has come to represent a broad research field. Most of the
work done today applies probabilistic or diverse machine learning based mod-
elling methods on plain GPS data, like Ashbrook et al. did in [2]. However, in
the last years, many researchers have been using semantic knowledge to describe
context information at a higher level aiming at improving the prediction with
promising results. The described context comprises locations, activities or even
the users themselves. By doing so, they dissociate themselves from the usual
GPS trajectories and they build their models upon so called semantic trajecto-
ries. Semantic trajectories describe movement as a sequence of meaningful and
human understandable annotated locations (e.g. home, office, gym, mall, ..).
Ying et al. introduces as one of the first semantic trajectories as a basis for a lo-
cation prediction algorithm [3]. Karatzoglou et al. investigate the use of artificial
neural networks to model semantic trajectories in different semantic layers [4].
Some use ontologies in their attempt to create more solid and reusable location
and trajectory representations. Wannous et al.’s research [5] for example focuses
on modeling movement by combining ontologies with rules. However, the afore-
mentioned work rarely goes beyond clusters and hierarchical structures when it
comes to modeling semantic locations and trajectories.
Our goal is to cover the full semantic spectrum of locations and how these
are actually being perceived by the user. For this purpose, we need to bring loca-
tions together with the temporary situation and user’s experience. The location
entities are linked to the corresponding entities, which capture and mirror the
whole experience of a person visiting a place, such as the purpose of visit, the
time, the activity, her companion, but also even more personal concepts like her
personality, mood and her overall mental state at that time. By taking a closer
look at our goal and the requirements mentioned above, we can easily identify so
called Poly-hierarchies and complex N-ary relations in them. We consider this
as our conditional framework. In this paper, we adopt the workaround method
for solving the N-ary issue, and we propose the creation and use of an extra class
in order to be able to embrace locations and their special meaning to the user
in its whole. In tangible terms, we propose to model the set of relationships of a
location entity to each of the aforementioned corresponding entities, like time,
purpose of visit, mood, etc., as a single resource, which we call Location-Specific
Cognitive Frames (LSCFs). Each instance of this class captures a personal and
temporary view on a location. An office for instance becomes a place of plea-
sure during the annual Christmas party, where the employee links it with a
different time(evening), different mood(unstressed), and different activities(eat,
drink, dance). We hypothesize that the use of such LSCFs in trajectories can
significantly support location prediction by enabling higher accuracy. Due to the
expectation of a great number of instances that differ only slightly to each other,
we propose furthermore the appliance of semantic similarity analysis metrics (for
instance Tversky’s Similarity equation) in order to cluster them appropriately
and overcome human fuzziness or noise in our conclusions.
Some first preliminary results based on a 5-week long real trajectory data
set of 4 users substantiate our hypothesis with our approach achieving an up to
32% higher f-score in comparison to Ying’s first approach.
References
1. Minsky, Marvin: A Framework for Representing Knowledge. Technical Report, Mas-
sachusetts Institute of Technology (1974),
2. Ashbrook, Daniel and Starner, Thad: Using GPS to learn significant locations and
predict movement across multiple users. PUC Vol.7, Nr.5, 275–286, (2003)
3. Ying, Josh Jia-Ching and Lee, Wang-Chien and Weng, Tz-Chiao and Tseng, Vincent
S.: Semantic Trajectory Mining for Location Prediction. Proc. 19th ACM SIGSPA-
TIAL, 34–43 (2011)
4. Karatzoglou, A., Sent¨urk, H., Jablonski, A., Beigl, M.: Applying Articial Neural
Networks on Two-Layer Semantic Trajectories for Predicting the Next Semantic
Location. Proc. 26th ICANN, (2017)
5. Wannous, R., Malki, J., Boujou A., Vincent, C.: Modelling mobile object activities
based on trajectory ontology rules considering spatial relationship rules. Modeling
approaches and algorithms for advanced computer applications, 249-258 (2013)
Article
Full-text available
In addition to context awareness and proactive behaviour, personalization constitutes one of the most important factors to consider when building digital assistance systems. In the field of location prediction, personal information can lead to a significant improvement of the predictive performance of the respective models. However, most of the existing approaches handle this type of information separate from the actual location and movement data, making them incapable of taking the entire existing underlying dynamics into account. In the presented work, inspired by the Minsky’s frame system theory in the 1970s, we evaluate an adapted ontological construct, which we call context-specific cognitive frame (CSCF), in order to capture the entire experience of a user at a given moment. Moreover, we use CSCFs to associate situations that arise when visiting a certain location with the respective context information as well as the emotions and the personality of the user. We show that our method can be used to provide a flexible, more accurate and therefore more personalized user experience using the example of location prediction.
Chapter
Current location-aware systems rely increasingly on location prediction techniques in order to provide their services in a timely fashion. At the same time, it has been shown that the use of additional context information, that is, elevating the degree of semantic enrichment of movement data, can lead to significant better results both in analyzing as well as in modeling human trajectories and predicting upon them. In this work, we propose a Multi-Channel Convolutional Neural Network (CNN) based approach for capturing all the available context dimensions in our semantic trajectory dataset aiming at achieving a higher prediction accuracy compared to a vanilla locations-only Single-Channel CNN. Moreover, we investigate whether and to what degree time, activity, companionship and the user’s emotional state have an impact on the predictive performance of our multi-dimensional CNN. We evaluate our model using a real-world dataset and compare it among others to a probabilistic Markov Chain model and a vanilla CNN at two semantic representation levels. It can be shown that especially for higher level representations, the present approach is able to outperform the baseline models achieving an overall higher accuracy and F1-Score.
Article
Full-text available
Wearable computers have the potential to act as in- telligent agents in everyday life and assist the user in a variety of tasks, using context to determine how to act. Location is the most common form of con- text used by these agents to determine the user's task. However, another potential use of location context is the creation of a predictive model of the user's fu- ture movements. We present a system that automati- cally clusters GPS data taken over an extended period of time into meaningful locations at multiple scales. These locations are then incorporated into a Markov model that can be consulted for use with a variety of applications in both single{user and collaborative sce- narios.
Conference Paper
Location-awareness and prediction play a steadily increasing role as systems and services become more intelligent. At the same time semantics gain in importance in geolocation application. In this work, we investigate the use of artificial neural networks (ANNs) in the field of semantic location prediction. We evaluate three different ANN types: FFNN, RNN and LSTM on two different data sets on two different semantic levels each. In addition we compare each of them to a Markov model predictor. We show that neural networks perform overall well, with LSTM achieving the highest average score of 76,1%.
Conference Paper
Research on predicting movements of mobile users has attracted a lot of attentions in recent years. Many of those prediction techniques are developed based only on geographic features of mobile users' trajectories. In this paper, we propose a novel approach for predicting the next location of a user's movement based on both the geographic and semantic features of users' trajectories. The core idea of our prediction model is based on a novel cluster-based prediction strategy which evaluates the next location of a mobile user based on the frequent behaviors of similar users in the same cluster determined by analyzing users' common behavior in semantic trajectories. Through a comprehensive evaluation by experiments, our proposal is shown to deliver excellent performance.
Article
This is a partial theory of thinking, combining a number of classical and modern concepts from psychology, linguistics, and AI. Whenever one encounters a new situation (or makes a substantial change in one's viewpoint) he selects from memory a structure called a frame, a remembered framework to be adopted to fit reality by changing details as necessary. A frame is a data-structure for representing a stereotyped situation, like being in a certain kind of living room, or going to a child's birthday party. Attached to each frame are several kinds of information. Some of this information is about how to use the frame. Some is about what one can expect to happen next. Some is about what to do if these expectations are not confirmed. The "top levels" of a frame are fixed, and represent things that are always true about the supposed situation. The lower levels have many "alota" that must be filled by specific instances or data. Collections of related frames are linked together into frame-systems. The effects of important actions are mirrored by transformations between the frames of a system. These are used to make certain kinds of calculations economical, to represent changes of emphasis and attention and to account for effectiveness of "imagery". In Vision, the different frames of a system describe the scene from different viewpoints, and the transformations between one frame and another represent the effects of moving from place to place. Other kinds of frame-systems can represent actions, cause-effect relations, or changes in conceptual viewpoint. The paper applies the frame-system idea also to problems of linguistic understanding: memory, acquisition and retrieval of knowledge, and a variety of ways to reason by analogy and jump to conclusions based on partial similarity matching.
Modelling mobile object activities based on trajectory ontology rules considering spatial relationship rules. Modeling approaches and algorithms for advanced computer applications
  • R Wannous
  • J Malki
  • A Boujou
  • C Vincent
Wannous, R., Malki, J., Boujou A., Vincent, C.: Modelling mobile object activities based on trajectory ontology rules considering spatial relationship rules. Modeling approaches and algorithms for advanced computer applications, 249-258 (2013)