Nowadays, people are accustomed to using general search engines such as Google and Bing to meet the information needs of daily life. Researchers and students can also use Google Scholar or Microsoft Academic Search to search for information. But they still need the services of digital libraries to access some databases. Academic digital libraries play an important role in academic context and provide necessary services for college students and teachers. With the development of information library systems, especially the application of discovery platforms, users can not only use digital library to search for paper print resources (e.g., books, journals, and magazines) but also search for electronic resources in different databases, such as the articles in ACM database. A large number of users access the academic digital library every day, which results in the large scale of user data that the researchers can analyze.
In addition, due to the rapid development of smart devices and the increasing ownership of different devices, users can use different devices to access academic digital library services. Switching devices are becoming more popular for users, especially in the face of complex search tasks. Montanez et al. defined the “cross-device transition” as the behavior that users change the device when searching on the web (Wang et al., 2013). They also found that the searches for information about adults and shopping triggered more device transitions (Wang et al., 2013). In addition, the probability of device transitions changes over time, and predictions of user’s behavior across device settings have recently become a hot topic (Montanez et al., 2014). The users’ network history, such as initial queries and time intervals, helps predict and support cross-device searches for users (Han et al., 2015a).
The prediction of user’s behavior requires a large amount of data support, and the massive log data from digital libraries can help to achieve it. Predictions of search tasks (Wang et al., 2013) and device types in cross-device settings (Montanez et al., 2014) help support search engines to provide users with appropriate content, for example, they can determine the display of pictures in the search result page according to the size of the device’s screen. Analyzing users’ cross-device search behavior can help users interact more effectively with online information resources in a multi-device environment. In addition, machine learning tools and methods have been widely used in a variety of industries. Digital libraries should also actively explore the application of machine learning in work and research. This helps improve digital library services and reuse user log data.
However, the existing research on cross-device transition mainly focuses on general search engines and their users, and there is a lack of studies on academic digital library users. Since many academic digital libraries can be accessed via phones, tablets, and other devices, it is increasingly popular to observe digital libraries’ cross-device searches. For example, a user can access a library service using a mobile library application (APP) on a mobile device, which can promote the device transition between the desktop and other devices. The user’s behavior of the library also shows the characteristics of cross-device transition.
Consequently, our work focuses on academic digital library users’ cross-device behaviors. This study can extend research on cross-device interaction and cross-device search to the domain of digital library. The data used in this paper were from a traditional transaction log of academic digital library.
The users’ activities on academic digital library online public access catalog (OPAC) can reflect the interactions they did, so we proposed the first research question:
RQ1- What kind of activities do academic digital library OPAC users conduct before and after device transition?
Predicting the next device and next activity after device transition can help library to provide the device-appropriate services and resume the user’s continuous behavior. For example, when users use mobile phones to search a book, the digital libraries can provide the basic information. When users change to another device (such as computers, laptops, and tablets), they can provide more kinds of information resources (e.g., picture and video) about the book. So, we aim to understand the second question:
RQ2- What are the important features to predict the next device and the activity after device transition?
2 Related Works
2.1 User’s information behavior in OPAC
Since the advent of OPACs, more and more libraries have provided the OPAC service for their services, and the OPAC has also become an important symbol of digital libraries. OPAC offers people with an additional option to search for the online information, especially for searching academic information, such as e-books and academic papers (Esposito et al., 1998). Users also have different information activities on the OPAC, such as searching for information, browsing the information, and gaining the knowledge. Therefore, user’s behavior on the OPAC is an important aspect of users’ information behavior research.
Much research has focused on the users’ experience when using OPAC. Solomon (1993) studied the children’s information retrieval behavior on OPAC and proposed the children’s behavior patterns that affect the success and failure of information retrieval as well as the children’s search intentions, actions, search strategies, and search terms. Shishido (1997) studied the difficulties that college students faced when using the OPAC through a survey in two universities. They found that users are not very satisfied with OPACs, and the main difficulties include subject search, keyboard operation, and interpretation. Sakong (2003) analyzed the difficulties encountered by web-based OPAC users in searching for information and proposed some suggestions to help users solve these difficulties, such as improving the interface design of the search results page.
Other researchers are concerned with the information recommendation and evaluation of information in the process of OPAC usage. Franke et al. (2006) proposed the information recommendation algorithm based on the OPAC usage history to improve the recommender services of digital libraries. Wang and Tang (2008) analyzed the information acquirement behavior of Chinese college students and teachers through the perspectives of search intention, search habits, and information evaluation. Queries can help understand the user’s interest; Jiang et al. (2015) analyzed the high-frequency queries in the OPAC log, and they found that the students in their university are mainly interested in information about mathematics, economics, management, and sociology. Willson and Given (2010) investigated the college students’ search behavior, especially the differences between OPAC and WWW. They found that the users’ OPAC search behavior tends to be similar with web search behavior, and they suggested the design of OPAC should be user focused.
In addition, clickstream analysis is also an important aspect of the user’s information retrieval behavior (Liu and Yuan, 2010), which helps to understand the user’s search intentions, search habits, and search efficiency. The OPAC clickstream can express the user’s academic information behavior; Jiang et al. (2017) used the clickstream data from transaction log of the university library OPAC and found that users are less willing to participate in exploratory search. They are more inclined to consider OAPC as a tool for searching academic information.
2.2 OPAC Usage in Multi-Device Environment
Although the general search engine (such as Google, Bing, and Baidu) is very popular (Kumar, 2011), professional information resources still play an important role in meeting the needs of users’ information (Missingham et al., 2009) and have their advantages (Yang et al., 2011).
Especially for academic information, users will choose a digital library to search for papers, books, course videos, etc. The OPAC log needs to be analyzed in the user search method study (Kern, 1983). As more and more digital libraries provide mobile Internet services, users can access library resources using mobile devices.
Previous works have focused on the needs of mobile services and how to evaluate mobile library services. Most users have positive comments on mobile library services and mobile phone applications (Pu et al., 2015). Research on the impact of devices on OPAC found that users prefer to use small screen devices to access OPAC (Cummings et al., 2010).
With the popularity of cross-device transition, the researchers also studied the query reformulation patterns of OPAC users on different devices (mobile phones, tablets, and desktops) and found that the transition pattern of search fields is affected by input functions and device interfaces (Wu & Bi, 2017a; 2017b). In summary, there is currently a lack of research that combines OPAC behavior with the impact of device, especially in cross-device search environments.
2.3 Prediction of Cross-Device Transition
It is now common that users’ Internet behaviors span different devices in daily life, especially facing complex tasks. Previous research on cross-device transitions has focused on search conditions (Wang et al., 2013; Montanez et al., 2014) and found that information needs for adults and shopping have led to more device transitions (Montanez et al., 2014).
The probability of device transitions is different at different times of the day. The prediction of cross-devices is also an important aspect in this field. The user’s network history, touch interaction on the mobile device, and time interval of device transition can support to predict user’s search behavior after the device transition (Han et al., 2015a; Han et al., 2015b). Wang et al. (2013) explored how to predict task resumption after switching devices; others learned a model based on the cross-device search characteristics to predict various aspects of cross-device transition, including the next device that the user might use (Han, et al., 2017).
Machine learning has gained more and more attention in the field of digital libraries. The theory and technology of machine learning can provide valuable support for digital library to develop more intelligent digital services (Esposito et al., 1998). Li et al. (2009) used a semi-supervised machine learning framework, combining with traditional literature retrieval methods to construct a ranking model for document retrieval structures based on semi-supervised learning of library user preferences. Sun et al. (2008) analyzed the transaction log data of digital library users within 1 year to predict user’s preference changes and proposed a method for predicting users’ future behavior preferences. Mandl et al. (2004) also proposed a framework for long-term continuous learning user preferences in information retrieval process based on the method of machine learning. In addition, the digital library also can re-rank the search results of OPAC according to the correlation between search results and user preferences (Furtado et al., 2009). Wu et al. (2018) modeled the information preparation behavior in the cross-device search context and analyzed the feature importance to describe the information preparation behavior. Search performance is an important indicators of search engines, Wu and Cheng (2018) studied the mobile touch interactions and search performance during the cross-device search process, they also developed the models for predicting the user’s search performance.
The above related work on this topic is mainly carried out by a general search engine, because the access of the large-scale user log data is convenient. There is currently a lack of research on cross-device transition and cross-device search. The definitions of cross-device search and cross-device search session are also different between the general information and the academic information in OPAC. The previous work mainly focused on the perspectives about the search methods, OPAC services, users’ experience, etc. The studies that consider both OPAC user’s behavior and the impact of different devices still need more attention. This article uses a large-scale traditional OPAC log data from a university, including different devices (such as PCs, mobile phones, and tablets) to analyze the cross-device transition behavior of users on OPAC. We studied the characteristics of cross-device transition and discussed how to predict the user’s device transition behavior through modeling. It is hoped that this study will enrich OPAC research in a multi-device environment and provide some suggestions or implications through this research to help digital libraries make better use of their log data.
3.1 Data Collection
Our study was conducted by analyzing the large-scale OPAC log from a university’s academic digital library, spanning 6 months from February to October in 2015. There are 16,140,509 records, and each record (“User-Agent” field) included the time of login and logout, user ID, device type, query, search method, and so on. A sample of data in our study is shown in Figure 1.
In addition, if a user logs into OPAC using his/her personal account, the record will contain his/her personal information, including the student number or e-mail address.
In the dataset, each record reflects the user’s online activity in the OPAC system. There are five main types of user activities:
- 1) Browsing: it refers to the user browsing the webpage after searching on the OPAC system, such as browsing book information on a search result page.
- 2) Clicking: it means that a user clicks on a specific link on OPAC. For example, he/she clicks a search result after searching or he/she clicks on a notice or news link.
- 3) Personal Account: it refers to the users’ browsing personal information in the user center, such as borrowing history.
- 4) Querying: when the user submits a query in the search bar on OPAC, the system records the query and search fields.
- 5) System Service: it means that user resets the personal password or communicates with the support personnel.
In our dataset, the “User-Agent” field in each record contains device information when accessing the library OPAC. We analyzed the information about the device’s operating system and browser and identified three types of devices: PC (desktop), phone (mobile phone), and tablet.
3.2 Unique User Identification
In cross-device settings, it is necessary to identify the unique user and combine their log data on different devices together, which reflects the unique user’s behaviors on different devices (Montanez et al., 2014). There are a large number of duplicate IP addresses on college campus that are not suitable for identifying unique users. Therefore, we only select data that contain the user’s full account information, ensuring that unique users can be distinguished based on their account information. We cleaned up a large amount of data that did not contain users’ account information by programming. Since the user does not have to log in when using the library OPAC, it is hard to collect the unique user on different devices with the account information. We only identified 248 cross-device transitions from 121 unique users, which lead to the small-scale cross-device transitions.
The “cross-device transition” in this study means that two consequent records of the same user contain different device codes within 24 hours, and this threshold helps to cover more device transitions (Montañez et al., 2014). In the OPAC record, the “user-agent” field includes the device code, and we analyzed these codes to identify the device type and device’s operating system, which helps us identify the unique device. There are many device transitions between devices of the same type, such as switching from a PC to a PC, which means that users can use their laptop in the apartment and then use the desktop in the library. “Cross-device transition” can be divided into “pre-transition” and “post-transition”, i.e., before and after the switching device.
3.3 Data Analysis
During the process of data analysis, we have cleaned a large number of data without users’ account information in order to identify unique users by programming. We used the linear regression analysis to build the prediction model and analyzed the importance of feature. Before the analysis of feature importance, we used K-fold cross-validation (K=8) to improve the accuracy of model, which could avoid the low prediction accuracy due to the small scale of sample. The method of root mean square error (RMSE) was used to evaluate the predicted performance of the different feature groups.
4 Analysis of Results
4.1 Transitions between Different Devices
By analyzing the transition probabilities between different devices, we can understand the user’s device transition habits. Through our research, we found that most cross-device transitions occur from PC to PC, then PC to phone, and phone to PC, similar to previous work (Montañez et al., 2014). This finding helps reveal that users tend to use the same type of device when they access the digital library next. However, in our study, the probability of device transition from PC to phone is higher than the probability of device transition from phone to PC, which is the opposite in the study by Montañez et al. (2014).
As mentioned before, it is not mandatory for users to log in the OPAC with personal account, which leads to the less data which can be used for analysis. Therefore, the device transitions between the phones and tablets are much less in our work.
4.2 Temporal Characteristics of Cross-device Transition
Temporal characteristics are always an important perspective in the research about user’s information-seeking behavior. In this section, we introduced the basic temporal characteristics about the device transition in OPAC, such as the hour when cross-device transition occurred and the time interval of cross-device transitions.
Figure 2 shows the hourly distribution when a cross-device transition occurs. From the figure, we found that cross-device transition occurs mainly in two time periods (from 8 am to 11 am and from 4 pm to 7 pm). These two time periods are mainly for studying in university, and it obey the college students’ living habits. Besides, before 7 am every day, users will almost never switch the device when using OPAC.
Time interval has always been an important feature in predicting users’ cross-device search behavior (Han, et al., 2017). We analyzed the time interval between pre-transition and post-transition to help predict the users’ device transition next. In our research, we found that, in the morning, the time interval between device transitions will be longer. Conversely, in the afternoon, the time interval between device transitions is shorter.
4.3 Activity in Pre-transition and Post-transition
Based on the OPAC transaction log, we summarized the users’ online activities with the OPAC into five types: querying, browsing, clicking, personal accounts, and system services. Figure 3 shows the distribution of changes of activity after device transitions, which was to answer RQ1.
From the results, the most common activity change is querying–querying (11.29%), which indicates that users like to continue searching for information on OPAC after device transition. Followed by the clicking–querying (9.27%), such as viewing the full information of the book on the first device or viewing the location of the book in the library, and then issuing a query to search for information on the next device. When users engage in activities related to personal accounts, they prefer to use it again (8.47%). Then, there is a transition from clicking to querying (6.45%), and also the probability between the browsing to browsing, clicking to browsing, and clicking to clicking is same (5.65%).
This result helps to understand the cross-device transition habits of OPAC users. Before and after the device transition, users’ activities reflect their need for OPAC. By analyzing the user’s activities before and after the cross-device transition, we can understand the user’s activity preferences after the device transitions and can also understand the association between different activities before and after the device transition. Thus, the library OPAC system can provide tools to help users resume their pre-transitions based on their search or online activity history.
4.4 Feature Importance Detection
To answer RQ2, we extracted the users’ behavior features with counts from the dataset (Table 1). There were five types of online activities with OPAC mainly by three types of devices (PCs, mobile phones, and tablets), accordingly with eight different device’s operating systems. Some features of time were also used to analyze.
Feature Used in Predictive Model
|Type||Features and Counts||Explanation|
|Activity in pre- transition and post-transition||pre_transition_activity (x5)||The five types of activities on the first device||A1|
|post_transition_ activity (x5)||The five types of activities on the next device||A2|
|whether_the_activity_ is_same (x2)||Yes or no||WI|
|Device in pre- transition and post-transition||pre_transition_device (x3)||The first device used before transition||D1|
|post_transition_device (x3)||The next device used before transition||D2|
|whether_the_device_ is_same (x2)||Yes or No||WD|
|Device’s operating system in pre-||pre_transition_ operating_system (x8)||The operating system of first device||O1|
|transition and post-transition||post_transition_ operating_system (x8)||The operating system of next device||O2|
|whether_the_ operating_system_is_ same (x2)||Yes or No||WO|
|Time||time_interval_of_ device_transition (x1)||Time interval between the device transition||T|
|time_spent_in_pre- transition_session (x1)||Time used in the last session on the first device||T1|
|time_spent_in_post- ransition_session (x1)||Time used in the initial session on the next device||T2|
Due to the identification of unique user, the sample to build the model was small. In order to avoid the problem of low prediction accuracy resulted from the small sample, the K-fold cross-validation (K=8) method was used to improve the accuracy of model. And, we built the prediction model using linear regression analysis to predict the feature importance. The SPSS Modeler and its machine learning algorithm were used to analyze the feature importance.
First, we wanted to detect the important features and build a prediction model for next device after device transition.
In order to answer the RQ2, we regarded the “post_ transition_device (D2)” as the target variable. Then, we regarded the “post_transition_activity (A2)” as the target variable to analyze the feature importance for predicting next activity after device transition. Because the probability of the model was <0.0005 (Sig.=.000), the models were statistically significant. Table 2 summarizes the top 5 important features according to the absolute value of Beta.
Feature Importance Analysis
|T1 (time_spent_in_pre-transition_ session)||-.006|
4.5 Prediction Performance of Feature Group
After the feature importance detection, we evaluated the prediction performance using the real results in the dataset. The input to the prediction model is the feature
combination. We compared the RMSE of the different feature combinations, as summarized in Table 3. The RMSE value reflects the deviation between the predicted value and the true value.
Feature Group Importance Analysis
|Predicting next device||Feature Group||Value of RMSE|
|Predicting next activity||Feature Group||Value of RMSE|
Table 3 summarizes that the feature combination to predict the next device performed better than to predict the users’ next activity. Adding the feature groups to the feature “pre-activity (A1)” can reduce RMSE of prediction
when predicting the next device. However, the feature combinations failed to reduce RMSE when predicting users’ activities, although the feature group “A1+T+D2” performed better than “A1+T”. This inferred that the feature “pre-activity (A1)” was more important in the prediction model for next interactions than that of next devices.
5 Discussion and Implication
5.1 Device Transition in Academic Digital Library Settings
Although research on cross-device transition and cross-device transition has received more attention recently, a perspective on the OPAC domain is lacking. We aim to understand the particular characteristics of cross-device transition in OPAC settings.
Through our research, we found that most cross-device transitions occur from PC to PC, then PC to phone and phone to PC, similar to the results of Montañez et al. (2014). The small amount of data with unique user’s information results in the small cross-device transition dataset. So, it may explain why the device transition in OPAC search mainly occurs between PC and PC. Besides, users always use the academic digital library to download papers, borrow books online, etc., and PCs can satisfy this need more than mobile devices.
Through our research, the time interval of cross-device transition on OPAC also differed from the time interval of cross-device transition on the general web. The mean time interval of device transition on OPAC was shorter in daytime, unlike the time interval on general web (Wang et al., 2013). In addition, the device transitions occurred on the general web were mainly in the afternoon, but in the academic digital library OPAC, there are two obvious peaks in the morning and in the afternoon. As well, the 65.32% of cross-device transitions occurred in less than 1 minute, which is much higher than the general web.
From our analysis, the searching activity accounts for the highest proportion after device transition, while clicking accounts for the most before device transition. This may reflect that after searching on the first device, OPAC users will continue to search on the next device, dissatisfied with the previous results or affected by external factors.
5.2 Prediction of Users’ Behavior after Device Transition
With machine learning methods and tools, intelligent information retrieval systems can implement different search results based on different unique users, even if they submit the same query. Search results vary from person to person. These are all achieved through machine learning methods, taking advantage of the user’s search habits, information needs, browsing habits, etc. According to the user’s search history, the search engine can select the best search results and recommend the query. This is also the result of machine learning, which analyzes the behavioral data of a large number of users. By studying the behavioral preferences of digital library users, it helps to provide better search result rank and provide better preference-based services.
In the cross-device search domain, the users’ behavior prediction is necessary and helps to recommend users much better queries, links, and services after device transitions. In this study, we mainly focused on the users’ cross-device OPAC transition behaviors, from the aspects of probabilities about the change of online activities after device transition, and how to predict the user’s next activity and next device in cross-device settings. We tried to explore how machine learning can help digital libraries better predict users’ behavior through this research.
Different with cross-device transitions when users access information on the general search engines (Han et al., 2015), the activity on the first device can predict the users’ next activity and next device after the device transition, which is better than the time interval of cross-device transitions. The research on the user’s behavior preference in the multi-device environment can make the full use of massive data of the digital library from many dimensions.
Machine learning tools and methods can help digital libraries make the most of their user log data. Digital libraries should also actively explore the application of machine learning in work and research. This helps improve digital library services and reuse user log data.
In this study, we studied the user’s cross-device OPAC behavior mainly from two aspects: the probability of cross-device transition and the activity in device transition. These features help predict cross-device behavior when users access OPAC. When transitions occur between different types of devices, we can use device information (such as IP or device code) to analyze the user’s context, such as their location.
Analyzing the activities in a cross-device transition helps to understand the habits of OPAC users. For example, library OPAC systems can provide tools, resources, or services to help users resume their search tasks on pre-devices based on their activity history.
Predicting users’ behavior should consider various sources of data, such as data from different devices. Predicting the next device and the next activity can help the library provide smarter and device-related services. For example, it can recommend device-appropriate resources, such as pictures or videos, to readers on different devices. In addition, the academic digital library can re-rank search results based on users’ habits and search history. They can also provide recommendations about related queries, books, research articles, and so on based on the users’ interactions on the first device.
In our study, the characteristics could help to predict the users’ cross-device transition. Where device transition occurred between different types of device, we can use the device information, such as the device code, to analyze users’ contexts, e.g., their location, to recommend context-based search services.
In the context of more users using different devices, such as personal computers, iPads, and smartphones, they can access academic digital library OPAC services and other digital library services to access information anytime and anywhere. This leads to the device transition behavior and results in the richness and diversity of digital library users’ behavior data. Based on the OPAC transaction log of a university, this paper focused on the users’ cross-device transitions in the domain of library OPAC.
We investigated the changes in users’ online activity between device transitions and used linear regression models to detect important features to predict the user’s next device and next activity. The prediction performance of the feature group was evaluated using real results in the dataset. We found that the activity and time interval on the first device are more important for predicting the user’s next activity and the next device. In addition, the features of the operating system help to better predict the next device. The next device user used is more likely to predict the next activity after device transition. Our research also explores how machine learning can help digital libraries better serve customers.
Our research has some limitations in fact. Due to university requirements, users’ account information data obtained from mobile phones and tablets are much less than those from personal computers, which results in small data samples that can be analyzed. The small amount of data from mobile devices may influence the results of the probability of device transition and results in the transitions between PCs, and PCs account for the most. In addition, we did not connect to any specific user for qualitative research. Through the necessary qualitative research, we can gain an in-depth understanding of the user’s intention or external cause of the device transition. In the future, we will conduct these studies through questionnaires or group interviews to explore the reasons for cross-device transitions.
This work is an outcome of project “User Seeking Behavior Modeling and Search Technology Development Within Multi-Device Integrated Search Environment” (No. 71673204) supported by National Natural Science Foundation of China and an outcome of Wuhan University’s independent research project (Humanities and Social Sciences) “Human-Computer Interaction and Collaboration Team” (Whu2016020) supported by “the Fundamental Research Funds for the Central Universities”.
Cummings J. Merrill A. & Borrelli S. (2010). The use of handheld mobile devices: Their impact and implications for library services. Library Hi Tech 28(1) 22–40.
Esposito F. Malerba D. Semeraro G. Fanizzi N. & Ferilli S. (1998). Adding machine learning and knowledge intensive techniques to a digital library service. International Journal on Digital Libraries 2(1) 3–19.
Franke M. Geyer-Schulz A. & Neumann A. (2006). Building recommendations from random walks on library opac usage data. In H. Kiers & P. J. F. Groenen (Eds.) Data Analysis Classification and the Forward Search (pp. 235–246). Berlin Heidelberg: Springer.
Furtado C. A. Willrich R. Fileto R. de L Siqueira F. & Tazi S. (2009 October). Custom ordering on digital library information retrieval. In Proceedings of the XV Brazilian Symposium on Multimedia and the Web (p. 28). ACM Fortaleza Ceará Brazil.
Han S. He D. Yue Z. & Brusilovsky P. (2015a June). Supporting cross-device web search with social navigation-based mobile touch interactions. In International Conference on User Modeling Adaptation and Personalization (pp. 143-155). Cham Dublin Ireland: Springer.
Han S. He D. & Chi Y. (2017). Understanding and modeling behavior patterns in cross-device web search. Proceedings of the Association for Information Science & Technology 54(1) 150–158.
Han S. Yue Z. & He D. (2015b). Understanding and supporting cross-device web search for exploratory tasks with mobile touch interactions. ACM Transactions on Information Systems 33(4) 1–34.
Jiang T. Chi Y. & Gao H. (2017). A clickstream data analysis of Chinese academic library OPAC users’ information behavior. Library & Information Science Research 39(3) 213–223.
Jiang T. Wang M. & Gao H. (2015). A search log analysis of OPAC users’ searching behavior - A case study of Wuhan University library. Documentation Information & Knowledge. 167 46–56.
Kumar S. (2011). Effect of web searching on the OPAC: A comparison of selected university libraries. Library Hi Tech News 28(6) 14–21.
Li M. Li H. & Zhou Z. H. (2009). Semi-supervised document retrieval. Information Processing & Management 45(3) 341– 355.
Liu Y. & Yuan P. (2010 April). A study of user downloading behavior in mobile internet using clickstream data. In 2010 Third International Symposium on Intelligent Information Technology and Security Informatics (pp. 255-257). IEEE Jinggangshan China.
- Export Citation
Liu, Y., & Yuan, P. (2010, April). A study of user downloading behavior in mobile internet using clickstream data. In)| false 2010 Third International Symposium on Intelligent Information Technology and Security Informatics(pp. 255-257). IEEE, Jinggangshan, China. 10.1109/IITSI.2010.79
Mandl T. & Womser‐Hacker C. (2004). A framework for long‐term learning of topical user preferences in information retrieval. New Library World105(5/6) 184–195.
Missingham R. Brettell R. White S. & Miskin S. (2009). Accessing information in a parliamentary environment: Is the OPAC dead?. Library Hi Tech 27(1) 42-56.
Montañez G. D. White R. W. & Huang X. (2014 November). Cross-device search. In Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management (pp. 1669-1678). ACM. Shanghai China.
Pu Y. H. Chiu P. S. Chen T. S. & Huang Y. M. (2015). The design and implementation of a mobile library app system. Library Hi Tech 33(1) 15–31.
Sakong B. H. (2003). A study on the searching behavior of web-based OPAC users. Jeongbo Gwan’ri Hag’hoeji 20(3) 81–110.
Shishido N. (1997). OPAC searching behavior of students at university libraries. Library & Information Science 37(37) 35–53.
Solomon P. (1993). Children’’s information retrieval behavior: A case chatman. Journal of the Association for Information Science and Technology44(5) 245–264.
Sun Y. Li H. Councill I. G. Lee W. C. & Giles C. L. (2008 October). Measuring user preference changes in digital libraries. In Proceedings of the 17th ACM conference on Information and knowledge management (pp. 1497-1498). ACM Napa Valley California USA.
Wang Y. Huang X. & White R. W. (2013 February). Characterizing and supporting cross-device search tasks. In Proceedings of the sixth ACM international conference on Web search and data mining (pp. 707-716). ACM Rome Italy.
Wang S. P. & Tang L. H. (2008). OPAC vs search engine – Information acquirement behavior of Chinese college students and teachers by taking Shanghai Jiaotong University as an example. Journal of Academic Library & Information Science 26(1) 63–68.
Willson R. & Given L. M. (2010). The effect of spelling and retrieval system familiarity on search behavior in online public access catalogs: a mixed methods study. Journal of the American Society for Information Science and Technology 61(12) 2461– 2476.
Wu D. & Cheng L. (2018 March). Predicting Search Performance from Mobile Touch Interactions on Cross-device Search Engine Result Pages. In International Conference on Information (pp. 560-570). Cham: Springer.
Wu D. Dong J. & Tang Y. (2018 July). Modeling and Analyzing Information Preparation Behaviors in Cross-Device Search. In International Conference on Cross-Cultural Design (pp. 232-249). Cham: Springer.
Wu D. & Bi R. (2017a June). Query reformulation patterns in cross-device OPAC search. In 2017 ACM/IEEE Joint Conference on Digital Libraries (JCDL) (pp. 1-2). IEEE Toronto ON Canada.
Wu D. & Bi R. (2017b). Impact of device on search pattern transitions: A comparative study based on large-scale library OPAC log data. The Electronic Library 35(4) 650–666.
Yang S. Q. & Hofmann M. A. (2011). Next generation or current generation? A study of the OPACs of 260 academic libraries in the USA and Canada. Library Hi Tech 29(2) 266–300.