• Tidak ada hasil yang ditemukan

3.1.1 Apparatus

A host computer decorated with the eye-tracking experimental setup is shown in Figure 3.1.

The involved apparatus components are as follows.

1. Tobii X2-60 eye-tracker1— a binocular eye-tracker that samples the gaze at 60 Hz (60 times in a second) and called a “fixation picker” [86] was utilized for our fixation-

based research. The eye-tracker’s accuracy (the average difference between real stimuli position and measured gaze position) is 0.4and the precision (variation of the recorded data measured via root mean square of successive samples) is 0.34.

2. EPU— external processing unit to perform the gaze related calculations which reduces the computational load on the host computer.

1https://www.tobiipro.com/product-listing/tobii-pro-x2-60/

3.1. COMMON EXPERIMENTAL SETUP

3. Ethernet cable to transfer the processed gaze-data from EPU to host CPU;magnetic strip to attach the eye-tracker onto the display monitor.

4. A power connector to supply the power to EPU.

5. A magnetic strip to mount the eye-tracker.

6. A host computer with Windows 7, 64-bit operating system and 8 GB RAM was used for rendering the stimuli on 22" display monitor.

Computer before eye-tracker setup Computer after eye-tracker setup

Eye-tracker connections Eye-tracker

Eye-tracker EPU

Power supply connector Ethernet cable

Infrared sensor

Ethernet connector to CPU Ethernet connector to eye-tracker EPU

Ethernet port

Power port

Power pin USB connector to eye-

tracker EPU

USB port Power button Infrared

sensor

Figure 3.1: Eye-tracker setup used for collecting the gaze-data

The magnetic strip was attached at the bottom edge of display monitor onto which the eye-tracker was mounted for non-intrusive eye-movement tracking. The eye-tracker was connected to the EPU to send the sampled gaze-data for processing. The I-VT filter (Velocity-Threshold identification fixation filter [90, 128]) is used to detect the fixations.

The I-VT filter assigns an angular velocity (measured in degrees/second) to each gaze data point, and all the points with the velocity less than a threshold (default 30 degrees/second is utilized as per [90, 106,122]) are classified asfixation; The fixation location is an arithmetic

mean of all the points corresponding to a fixation. The EPU was connected to the host CPU using Ethernet cable for storing the processed data for further analysis. The necessary apparatus connections are shown in Figure 3.1.

3.1.2 Stimuli Preparation

Exp-I

Exp-II Exp-I

Exp-II

Figure 3.2: Sample stimuli webpages of Experiment-I (top two rows) and Experiment-II (bottom two rows).

In order to account for the user attentional behavior on webpages with varied visual features, we considered the webpages from multiple categories including SHOPPING, HEALTH, ARTS, HOME, NEWS, and SPORTS from Alexa2. The sample stimuli of Experiment-I3 and Experiment-II4 are shown inFigure 3.2. Each webpage was displayed in full-screen mode on a 22" display monitor and screenshot was captured (at a screen resolution shown in Table 3.1). The screen-shot helps in rendering the webpage as a whole avoiding the rendering latencies of individual web elements which may influence the attention allocation.

Additionally, the stimuli webpages (including HTML,CSS, and media) were stored locally

2https://www.alexa.com/topsites/category

3Stimuli screenshots & webpages are available athttp://bit.ly/2DaUoup&https://bit.ly/3l8avjv.

4Stimuli screenshots & webpages are available athttps://bit.ly/2C8JmKP&https://bit.ly/3l8avjv.

3.1. COMMON EXPERIMENTAL SETUP Table 3.1: Summary Experimental Details

Characteristic Experiment-I Experiment-II Experiment-III

Number of webpages 51 36 149

Number of participants 65 (33M; 32F) 42 (21M, 21F) 11 (4M, 7F)

Min. age of participants 21 years 20 years 21

Max. age of participants 34 years 32 years 25

Avg. age of participants 24.32 years 26.93 years NA

Standard deviation of age 3.24 years 2.69 years NA

Screen resolution 1680×1050 1280×720 1360×768

Fixation-threshold 100 ms 100 ms 100 ms

Each webpage display duration 5 seconds 5 seconds 5 seconds

Number of fixations on Text 11722 5726 NA

Number of fixations on Images 3013 2223 5762

Total number of fixations 14735 7949 5762

for the purpose of replication and reproducibility.

3.1.3 Experimental Setup

A slide-show presentation was prepared with a blank slide inserted after every webpage to reset the participants’ attention after every webpage. Counter-balance mode of stimuli presentation was considered to mitigate the ordering effects [161]. A 22" desktop monitor was used for displaying the stimuli. Each slide (including the blank webpage) was presented for five seconds duration which is typical for free-viewing analyses [134].

3.1.4 Procedure

Participants (see the details inTable 3.1) were individually invited to the experimental setup.

Each participant was briefed about the experimental setup. Instructions were provided that all the webpages are independent (to avoid the expectational bias about presentation sequence). In pre-experimentation step, eye-tracker calibration was performed using a 5-point method (participants gazed at the center of a red-color ball when it was moving along the four corners and the center of the screen as shown in Figure 3.3, second from left). Quality of the calibration was assessed and recalibration was performed, in case of errors. After successful calibration, an automated slideshow was presented usingcounter- balanced mode (to mitigate the stimuli ordering effects on attention) which wasfree-viewed by the participant while their eye-movements were captured by the eye-tracker. In the post-experimental survey, participants were askedhow many of the webpages shown in the experimentation was previously seen by you? for which each participant reported a number from 0 to the number of stimuli webpages. The reported median of zero known webpages mitigated the possiblememory-bias effect.

Stimuli Calibration

Calibration-Check

Gaze-tracking

Figure 3.3: Experimental procedure in four steps: (i) stimuli loading (ii) eye-tracker cali- bration for a participant (iii) checking for successful calibration (iv) gaze-tracking during stimuli presentation

3.1.5 Data Extraction

From users’ gaze data, fixations were extracted with the fixation duration of at least 100 milliseconds which is a typical threshold [55]. Accordingly, Fixation-Indices (FIs) were assigned. Now, the locally stored webpages were launched in a headless browser to replicate the screenshots of stimuli webpages, as seen by the participants. Based on the fixation- position, the DOM tree of the corresponding webpage is traversed (using a JavaScript program) and the leaf node element is identified, analogous to [19]. The corresponding fixated elements were identified from the fixation location. Among the extracted fixations, the repetitive fixations (multiple fixations from the same participant on a web element) are pruned. The repetitions might be due to the lack of further salient elements on a webpage, where, as perinhibition-of-return, participants are less likely to attend the already attended element [89]. Consequently, we considered the FI corresponding to the first attention