Students’ Research Works – Autumn 2018: User Experience and Implicit Feedback (PeWe.UX)

Visualization of Eye-Tracking Records

Martin Civáň
Abstract: One of the most efficient ways how to do eye-tracking experiment with many participants is to record more people at one session that one by one. However usual eye-tracking recording programs are not optimal for this kind of task. So the UXR infrastructure has been developed to help us with it. Unfortunately, the client nor the server part does not include any visualization tool to take a look at the recorded data.
In our work, we are trying to develop a visualization tool compatible with the data from the UXR infrastructure. One of the considered approaches are to create import module to Ogama. While exploring Ogama, some UX problems were detected and possibly experiment set-up export plugin was considered to be implemented.

to the topto the main

Discovery of gaze patterns in navigational tasks on the Web

Gabriel Csikmák

Abstract: Nowadays, the use of the Web has become an everyday activity of almost every person in the world. Their most popular actions include searching for information, which they complete on so-called web search engines. However, their behavior during interactions with these resources can be affected in several ways. There are various efforts that try to understand, also to describe the behavior of users on the Web at the same time.
One possible way of observing is to use a gaze tracking device to get valuable information about the interaction of users on the Web. For this purpose, eye-tracking methodology can be used, which is increasingly used for such types of observation. The information obtained by this tool allows us to determine different patterns of behavior based on user properties or also based on the situation on the Web.
The aim of this work is to understand the behavior of users when performing navigational tasks on the Web, also to determine how their behavior can be influenced while interacting with the browser. It also aims to suggest a solution for information gaining from user interaction and then perform analysis on a diverse sample of users for this type of solution.

to the topto the main

Discovery of view patterns in navigational tasks on the Web

Patrik Kovács

Abstract: Nowadays a considerable effort is invested into the development of methods aiming to influence the viewing patterns of Web users. It is a challenging task to understand and describe the behavior of Web users. To overcome this problem, we need to assign different patterns of behavior that is specific to user profiles and also to the information presented on web pages. With eye tracking devices we can obtain additional feedback from users about their interaction with the web.
In our work we shall focus on navigational elements of websites. We are going to test the effectiveness of different types of menus on the sample of users. Data from the eye tracker will be used for measuring and evaluating the effectiveness of different types of menus.
We hope that our findings will help the web designers to create more effective menus.

 to the topto the main

Analysis of gaze patterns based on user interaction in navigation tasks

Jakub Kubanyi

Abstract: User behavior differs in various aspects. User skill plays an important role in user behavior. We can view on user skill from different perspectives. A significant part of the research takes place on the skill of searching for information and navigate the Web.
We can use an eyetracking technology to track user activity. With the eyetracker, various viewing metrics, such as the number of fixations or the duration of fixations, can be obtained. The user view can also be observed at the level of saccades, allowing us to explore the gaze patterns. User interaction itself can be explored on prepared stimuli. Examples are websites where the user performs web navigation tasks.
The work analyzes the possibilities of exploring the user skill based on gaze patterns of user. We are preparing a suitable experiment to get the data on a diverse sample of users. We are going to design a method of processing and analyzing data to extract patterns and then classify the degree of user skill.

to the topto the main

Automatic Text Comprehension Detection

Vladimír Kuchár

Abstract: Reading is one of the main ways in which information is acquired and understanding of the text being read is essential to gaining new knowledge. The reader’s ability to understand the text is influenced by their knowledge, their current psychological state and the difficulty of the text itself. One of the ways to verify the level of comprehension is to test the reader based on the text. Another possibility is to determine the level of comprehension automatically based on the physiological characteristics of a person that were recorded during reading. It is known that the movement of the eyes while reading with comprehension reflects the thought processes behind the reading process.
This work focuses on the automatic text comprehension detection using eyetracking and natural language processing of texts. The current development of natural language processing allows us to automatically calculate many features based on the text being read that affect its complexity. These text features are directly associated with specific eye movements. The knowledge gained from the analysis was used in designing of our method and selection of interesting eyetracking features, features based on the text itself and their combination. In this moment we are working on preprocessing the data gained in conducted user study and verifying our model.

to the topto the main

Support of website usability testing

Andrej Nemeček

Abstract: The usability of websites can be tested by using several tools. In our project we mainly focus on information architecture tests, namely the method of card sorting and tree testing. Both methods can be used to understand how users think of website structure and content. Without user understanding, it is impossible to design good information architecture.
For remote and unmoderated testing, we often loose some valuable data. Existing solutions usually only show the participant’s overall results, but do not show enough data of participant behavior while performing test tasks. Our goal is to properly represent these data.
We focus on the usability of the websites and the methods for testing usability. We analyze existing tools for testing the information architecture. Consequently, our goal is to create a tool that tracks the participant’s complete behavior during card sorting and tree testing and represent the results of these two tests.

to the topto the main

Comparison of scanpaths

Michal Oláh

Abstract: Eye tracking reseach has been applied in many areas, including cognitive psychology, medicine and now it’s often used in exploring user interactions with websites in order to test usability or modeling user’s behavior.
The approaches used in analysis often have different limitations and drawbacks. In many scenarios, it’s not enough to initially pre-process and remove invalid data to get relevant results when comparing scanpaths. This problem could be caused by anomalous scanpaths identified with respect to the analyzed data. Often, significant deviations in data may distort the results of the analysis and therefore it‘s necessary to identify this type of data and analyze its origin.
In our work we focus on the comparison of scanpaths and the identification of anomalous scanpaths. Our main objective is to design and implement a method of comparison of scanpaths in order to identify local anomalous scanpaths. We believe that by combining the previously analyzed approaches, we will achieve relevant results in the proposed approach and successfully verify the proposed method in finding a common scanpath.

to the topto the main

Automatic Categorization of Users Based on their Web Navigation

Ľubomír Samotný

Abstract: The aim of this work is to create an automated method of categorizing users according to their cognitive styles. That is why we need to get enough data on how users of different cognitive styles are behaving on the web. Based on these, we will design and verify a classifier that can properly assign a cognitive style to the user. It is proved that users with different cognitive style perceive and process the information provided to them differently. By automatically determining a user’s cognitive style only during his web browsing, it would be possible to start personalizing a site that would be better suited to his website perception, thereby improving his user experience and browsing speed. Two dimensions of cognitive styles were described and verified in psychological research in the past, specifically Verbal-Imager and Wholist-Analyst dimensions. At the beginning of the experiment, we will perform a short test to determine user’s inclusion to these two cognitive style dimensions. The experiment will then continue with search-based tasks on the Web. These are designed to detect differences in users navigating behavior of different cognitive styles. For the classifier to work properly, it is therefore necessary to collect enough data from eye-tracking, but also multiple metrics measured by Javascript on the front-end.

to the topto the main

Probabilistic Models for Analysis of Performed Task based on Eye

Andrej Vítek

Abstract: Gaze and visual patterns of observer while looking on visual stimuli is affected by 3 factors. First factor (bottom-up) is related to visual stimuli and its properties. Second factor (top-down) is related to personal characteristics of observer like task at hand, experience, mood or gender. Third factor is related to characteristics of oculomotor system. As a result of these factors, the observer’s gaze contains a great amount of information about the user as well as about what they are looking at.
Models, which are using information about gaze, are often based only on evaluation of aggregated metrics of gaze, thereby losing temporal information about visual exploration of observer. In this work, we focus on the creation of features by using probabilistic models, which could abstract this information from data. We investigate, which probabilistic model and with what input features has best ability to model data from selected domain and whether the subsequent clustering of created probabilistic models can improve its ability.

to the topto the main