This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
As the technology provided by mobile devices has been expanded exponentially, new areas have come to light. One of these areas is personalization. Personalization involves with building user profiles and retrieving the related data. This study proposes a new system (CPMC) for searching movies in mobile devices by providing the mobile users with interesting video clips according to their personalized preferences. A context proxy will be incorporated in order to collect context profiles and, therefore, displaying the results that much suits the users' contexts.Keywords
Personalization, contextualization, customization, movie, Mobile Information Retrieval (MIR), Multimedia Retrieval (MR).
Mobile devices have affected our lives significantly. They play an important role in our lives as well as our daily activities. We use mobile devices to make calls (video and voice), send instant messages and multimedia files, and make memos ...etc. Other uses became available after the great growing in mobile domain, such as: Global Positioning System (GPS), Bluetooth, camera...etc. There is infinite number of uses of mobile devices.
Also, the connection used by mobile devices has been enhanced. That means that the user can easily use the mobile device to connect to internet and download some supported files. In the past, these files were limited types. But in these days and as the mobile devices have been developed in order to support more types, it become easily to download a movie or even watch it as a TV stream technology. But there are some constraints that should be considered, such as: limited screen size, limited bandwidth, and battery issues. Unfortunately, these issues have been resolved partially by providing a big memory size and maintaining a good connection bandwidth.
As the memory size get bigger size and a good connection bandwidth have been granted, the price becomes higher too. So, what is the solution? And how to balance all items together? This leads us to the personalization. Personalization will make users comfortable while using the devices to retrieve data. Actually, it is a part of what is called Mobile Information Retrieval (MIR).
My contribution will be providing a personalized movie customization system for the mobile devices in addition to provide a context proxy on these devices in order to collect user information, analyze context information, and compile context profiles. In the opposite of current researches which focus on Multimedia Retrieval (MR) for desktop PC only, I aim to use MR for mobile devices. So, this contribution will be based on a combination between 2 fields: MIR and MR.Problem Statement
Previous approaches to building contextualized, personalized and customized systems have several limitations. One of these limitations, there is no focus on the search context in order to make suggestion. In other words, providing the results related to a search query (e.g. Singapore vs. Malay football game). On the other hand, there is no current research toward personalizing movies for mobile devices because of old belief: mobile devices cannot download and hence view movies. Finally, many approaches couldn't be applied on mobile devices.
This study proposes to provide users with a personalized movie customization system for the mobile devices in addition to provide a context proxy on these devices in order to collect user information, analyze context information, and compile context profiles.
The objective of this research is to develop a personalized movie customization system (CPMC) for the users of mobile devices in addition to provide a context proxy on these devices in order to collect user information, analyze context information, and compile context profiles. The profiles will help in retrieving the most appropriate video clips within a movie.Hypothesis
The proposed movies customization system for mobile devices (CPMC) which is personalized and contextualized will significantly achieve better consistency, conciseness, coverage and convenience.Assumptions
There is a big demand on making a personalized and contextualized customization on movies to be displayed on mobile devices There are no any effective existing customization on movies that is displayed on mobile devices The intelligent manager will be able to process context profiles, identify ambiguous query, expand query, and reorder the result movies based on user current interests in the context.
The user can use the friendly UI to easily customize his favorite video clips.Delimitations
- This research aims to focus only on movies and will not cover any other topics, such as: news and sport videos.
- This research is limited to the contextualized, personalized and customized approaches and does not cover any other searching preferences or techniques.
It will provide a simple way to personalize movies based on the user's context. While personalization of movies is very important for normal PC, it will become more preferred for mobile users for many reasons such as: limited bandwidth.Terms & Definitions
- User Profile: a collection of personal data (movies and clips) associated to a specific user
- Personalization: building a differentiator among users based on his/her own selections
- Mobile user: anyone who uses mobile device
- Consistency: Whether the generated video clip is consistent with user request on content semantics, such as involved players and event types
- Conciseness: Whether the generated video clip capture the main body of the match without including irrelevant events
- Coverage: Whether the generated video clip covers all important events happened in the match under current viewing time limit
- Convenience: Whether the mobile client can facilitate the user conveniently customize his/her favorite sports video clip
One of the existing researches, as proposed by (Cheng, Fu, Jiang, Liang, Lu, Luo, Ma, Wang, Xu, 2010), focuses on providing a personalized sports video customization system for the mobile device. Their system can provide timely multimedia service with the help of automatic live video analysis and meet user personalized preference. The interested video clips will be customized by users to select their favorite video clips by using a friendly UI. On the other hand, a server will be used to detect live events from sports video, which can generate both accurate event location and rich content description.
Another research has focus on the context search as proposed by (Adjouadi, Gui, Rishe, 2009). The research focuses on personalization strategies which explicitly and implicitly infer user search context from user current environment. It is very suitable for ambiguous queries and localized searches. So, it collects user information, analyses context information, and compiles context profiles. These context profiles, which reflect user current context, assist query expansion to solve ambiguity.
Another approach has been proposed by (Boughanem, Daoud, Lechani, 2009). It is a personalized search approach involving a semantic graph-based user profile issued from ontology. User profile refers to the user interest in a specific search session defined as a sequence of related queries. It is built using a score propagation that activates a set of semantically related concepts and maintained in the same search session using a graph-based merging scheme.
In (Bouidghaghen, Bouidghaghen, Lechani, 2009), a novel situation-aware approach to personalize search results for mobile users has been discussed. It is based on a combination of geographical and temporal concepts inferred from concrete time and location information by some ontological knowledge. User's interests are inferred from past search activities related to the identified situations. So, the appropriate information that dynamically satisfies the user's interests according to his/her situation will be retrieved without any overload.
In (Mundur, Rao, Yesha, 2004), a technique for automatic summarizing videos has been proposed. The automatically generated video summaries capture the visual content of original clips and are suitable for wireless and mobile environments. It can be used in order to cluster multi-dimensional point data corresponding to the frame contents of the video. The data doesn't have any redundancy. It has been emphasized that this technique can be applied on speaker recognition, automatic indexing, and text recognition in video.
Another research has been conducted by (Avrithis, Castells, Corella, Fuentes, Mylonas, Vallet, 2005). Their work was a framework that can represent, capture, and manages any information about multimedia management system users that can be exploited. That means: collecting users' interactions to build profile whereby content semantics are linked to a rich representation of user preferences. It can be helpful in somehow by applying this framework to support a wide range of personalization facilities in a multimedia content management.
(Boughanem, Bouidghaghen, Lechani, 2009) has focused on dynamicity personalization. They have proposed an approach for a situation-aware personalized search. It has 3 steps: inferring semantic situations from low level location and time data, learning and maintaining user interests based on user's search history related to the identified situations, and selecting a profile to use for personalization given a new situation by exploiting a Case Based Reasoning (CBR) technique. A situation is represented as a combination of geographical and temporal concepts inferred from concrete time and location information by some ontological knowledge. User's interests are inferred from past search activities related to the identified situations. They are represented using concepts issued from a thematic ontology.
In (Doulamis, Litke, Papadakis, Skoutas, Varvarigou, 2005), an agent-based system has been proposed for mining textual and visual information from the web. The data collected by visiting 50 different web data sources, acquiring more than 63.000 documents and more than 6.5 GB of media content. As this system provides visual content mining and as it supports cross-language, it will be helpful for automatic language translation.
Another related work has been done in (Doulamis, Kollias, 2000). They suggested a new representation of visual content by constructing a fuzzy histogram for each video frame based on a collection of appropriate features, and therefore make an extraction by using video sequence analysis techniques. This is useful for the new emerging multimedia applications, such as content-based image indexing and retrieval, video browsing and summarization. The main contribution made by them and can be applied to my research is making segments based on an analysis of video sequences. For these video sequences, a color/motion segmentation algorithm will be applied.
Another personalization has been proposed by (Bhandarkar, Li, Wei, 2007). This system provides video personalization for mobile devices or any constrained client which is most relevant to the client's request while simultaneously satisfying multiple client-side system-level resource constraints. They have computed the relevance value of video summaries by giving their relative time durations and the relevance values of the corresponding original video segments. The used multi-stage client request aggregation strategy reduces the effective number of client requests processed by the server and thus reducing both, the server and network load and the client experienced latency.
While the framework proposed by (Echigo, Etoh, Masumitsu, Sekiguchi, Teraguch, 2001) is using metadata based on MPEG-7 for constructing video digests reflecting individual preference. They use metadata in order to generate short video clips from a long image sequence. They assign value changed in time interval of indexes that present the content of the scene. Next, they estimate content profiles as to generate typical digest video clips to meet similar demands from many users. The main problem with this framework is that it does not generate quite personalized digest video for all users. Instead it provides some typical digest video clips for users without well-learned user profiles to enable them to choose favorite ones, which can be combined individual requests from users.Research Methodology
The carrier's network is the server which provides data and voice services to subscribers. User submits mobile queries to network. The mobile devices play an extensive role of collecting, analyzing, and extracting context entities. Context profiles compiled at client side adapt to the mobile user's current situation.Client:
- In this module, the user will make inputs (voice or data).
- Hardware logic will collect surrounding environment inputs (temperature, GPS reading, altitude, e.g.).
- While the applications will collect other inputs such as the user calendar.
- The context aware proxy will further inspect inputs and extract context entities from the inputs. Finally, the proxy will compile the context profile and send it to the carrier server.
- In this module, the user context and the data collected by network context (i.e. user's location and query history) will help the carrier server to learn the user's situation.
- Intelligent manager identifies the ambiguous queries and further expands these queries with derived user context
- The expanded queries contains extra information which could facilitate search engines and improve the topic relevance of returned documents will be sent to search engines.
- The search engines return all the documents related to the query based on the augmented context profile
- The returned documents are grouped into categories and ranked in terms of descending weights in the context profile
- It utilizes user's selection, as feedback, to adjust context weight calculation at client and server sides in order to improve the accuracy of server's responses
- As user changes activities/context, context entities are added to or dropped from the profiles
- If the user's situation remains the same or changes a little, then only updated weights in the context profiles are sent to network
- It manages 4 context profiles:
- User profile describes the user's characteristics, preferences, activities, emotions, and so on. For example, user's age, gender, height, weight, health conditions, past incidents, and planned events can be sorted into this category.
- Device profile describes the configuration and functions of the hardware. In addition, this profile includes the software applications and user interfaces. Examples of context entities in this profile are processor, memory, display type, device size, and OS.
- Environment profile describes surrounding area of the mobile user. This profile stores user location, weather, noise level, or temperature, etc.
- Data profile caches user's data in the local memory.
Design 1: One-Shot Experimental Design:-
In the experimental design, design 1 is applicable for our experiment since we have one group that is presented by the students and a posttest only for the reason that we don't have any previous work in this field that can be used as a benchmark.Treatments:
The treatment that is used in the experiment is the contextualized personalized movies customized online mobile directory.Observations:
The posttest is the observation for the contextualized and personalized movies customization for mobile devices in terms of consistency, conciseness, coverage and convenience.Participators:
We will do our experiment in the Faculty of Information Technology and Computer Science, UPM. Thirty postgraduate students from FSTKM who can understand and work on mobile devices applications were selected randomly as the sample of this experiment.
In the experiment, we introduce the contextualized and personalized movies customization for mobile devices (CPMC) and ask the students to build queries about movies.Data collection:
After using the contextualized and personalized movies customization for mobile devices, consistency, conciseness, coverage and convenience are recorded using the comments taken from the questionnaire that is filled by each participant upon finishing the searching task using this framework. The questionnaire has a scale for each variable from 1 to 5 where 1 stands for strongly disagree while 5 denotes strongly agree. The experimenter recorded all verbal comments or behavioral observations that were later analyzed using SPSS software. Table 2 shows the data that was collected.Data analysis:
- For these data, we performed two-way ANOVA using SPSS software to detect the averages in consistency, coverage, convenience and conciseness during the experiment time and to extract more information from our participators result.
- As it is shown below in table 2, the averages of consistency, coverage, convenience and conciseness are calculated to determine the effectiveness of the proposed system (CPMC).
- We found that the CPMC achieved reasonable consistency, conciseness, coverage and convenience.
- We found that the CPMC achieved a significantly high consistency. Movies have been checked by analyzing the generated video clips. Are they consistent with user request on content semantics, such as involved actress and event types? On average, the coverage has achieved 4.56 out of 5.
- We found that the CPMC achieved a significantly high conciseness. Movies have been checked by analyzing the generated video clips. Do they capture the main body of the movie without including irrelevant events? On average, the coverage has achieved 4.6 out of 5.
- We found that the CPMC achieved a significantly high coverage. Movies have been checked by analyzing the generated video clips. Do they cover all important events according to the user's contexts under current viewing time limit? On average, the coverage has achieved 4.5 out of 5.
- We found that the CPMC achieved a significantly high convenience. Movies have been checked by analyzing the generated video clips. Can the mobile client (user) facilitate the user conveniently customize his/her favorite movies video clip? On average, the coverage has achieved 4.6 out of 5.
- Cheng, Fu, Jiang, Liang, Lu, Luo, Ma, Wang, Xu (2010), Personalized Sports Video Customization for Mobile Devices, Springer-Verlag Berlin Heidelberg
- Adjouadi, Gui, Rishe (2009), A Contextualized and Personalized Approach for Mobile Search, International Conference on Advanced Information Networking and Applications Workshops
- Boughanem, Daoud, Lechani (2009), A Session Based Personalized Search Using An Ontological User Profile,SAC'09 March 8-12, 2009, Honolulu, Hawaii, U.S.A
- Mundur, Rao, Yesha (2004), Automatic Video Summarization for Wireless and Mobile Environments, IEEE Communications Society
- Avrithis, Castells, Corella, Fuentes, Mylonas, Vallet (2005), A Semantically-Enhanced Personalization Framework For Knowledge-Driven Media Services, ICWI 2005
- Boughanem, Bouidghaghen, Lechani, (2009), Dynamically Personalizing Search Results for Mobile Users, LNAI 5822, pages99-110, Springer-Verlag Berlin Heidelberg
- Doulamis, Litke, Papadakis, Skoutas, Varvarigou (2005), Retrieving, Adapting And Delivering Multimedia Content Using A Mobile Agent Architecture, IEEE 2005
- Doulamis, Kollias (2000), A fuzzy video content representation for video summarization and content-based retrieval
- Bhandarkar, Li, Wei (2007), Video personalization in resource-constrained multimedia environments, MM'07, September 23ââ‚¬"28, 2007, Augsburg, Bavaria, Germany
- Echigo, Etoh, Masumitsu, Sekiguchi, Teraguch (2001), Personalized Delivery of Digest Video Managed on MPEG-7