Mashup Based Content Search Engine for Mobile Devices

Mashup based content search engine for mobile devices is proposed. Example of the proposed search engine is implemented with Yahoo!JAPAN Web SearchAPI, Yahoo!JAPAN Image searchAPI, YouTube Data API, and Amazon Product Advertising API. The retrieved results are also merged and linked each other. Therefore, the different types of contents can be referred once an e-learning content is retrieved. The implemented search engine is evaluated with 20 students. The results show usefulness and effectiveness on e-learning content searches with a variety of content types, image, document, pdf files, moving picture.


INTRODUCTION
Mashup technology is defined as search engine with plural different APIs.Mashup has not only the plural APIs, but also the following specific features.

1) it enables classifications of the contents in concern by using web 2.0,
2) it may use API from the different sites, 3) it allows information retrievals from both sides of client and server, 4) it may search contents as an arbitrary structured hybrid contents which is mixed contents formed with the individual contents from the different sites, 5) it enabling to utilize REST, RSS, Atom, etc. which are formed from XML conversions.
There are some Mashup tools, Yahoo!Pipes 1 , Microsoft Popfly 2 , etc. while there are some services using Mashup technology, ChaMap -Enjoy Geo Communication!- 3 , Newsgraphy 4 , Flowser on Amazon 5 Although Mashup allows content search which is same as portal, Mashup has the aforementioned different features from portal.Therefore, Mashup is possible to create more flexible search engine for any purposes of content retrievals.The search system which is proposed here is that make it possible to control the graph in the 3D space display with these 1 http://pipes.yahoo.com/pipes/ 2 http://www.microsoft.com/ja-jp/dev/default.aspx 3 http://chamap.net/ 4http://newsgraphy.com/ 5 http://www.flowser.com/peculiarity on Android devices.Mashup technology utilized search engine for e-learning content retrievals is proposed.Example of the proposed search engine is implemented with Yahoo!JAPAN Web SearchAPI, Yahoo!JAPAN Image searchAPI, YouTube Data API, and Amazon Product Advertising API.The retrieved results are also merged and linked each other.Therefore, the different types of contents can be referred once an e-learning content is retrieved.The implemented search engine is evaluated with 20 students.The results show usefulness and effectiveness on e-learning content searches with a variety of content types, image, document, PDF files, moving picture.
The following section describes the proposed search engine followed by implementation results and experimental results with 20 students which show usefulness and effectiveness of the proposed content search engine for mobile devices.Then conclusion is described together with some discussions on future investigations.

A. Background of the Proposed Search Engine
Figure 1 shows research background of the proposed search engine.

Research background of the proposed mashup based content search engine
There are seven key components, user involvement, user contribution, arranging by users, reliability, rich experiences, www.ijarai.thesai.orglong tail, and distributed system.Facebook, Mixi, etc. of social network system allows user involvement.Users may contribute to create huge database through amazon.com,for instance.Long tail becomes true with amazon.com,Rakuten, etc, Not only contents but also rearrangement of the contents can be done by users through Hatena Bookmark, Nikoniko, etc. Reliable Q&A is available through Wikipedia, Yahoo answer, etc.Also rich experiences are obtained through e-mail communications with Gmail, for instance.These activities can be done based on Web 2.0 technology utilized distributed system.In particular, Mashup technology allows efficient and effective information retrievals for the distributed information providing systems.

B. Content Types and Representation Models for the Content Search
There are some contents types for search.Namely, document, moving picture, a variety of types of contents, images, etc.Also there are some representation models for the different search content types, helical, star, star-helical, and star slide models.Figure 2 shows an example of star slide model of representation of contents types (Blue star) and the search results (Smile marks).Namely, there are the different content types on the top and there are search results under the content types.With the touch panel function, these content types can be selected (swipe in horizontal direction) and also search results can be selected through swipe in vertical direction.Under the icon of these five content types, candidate contents are aligned below the icons as search results.In order to show the candidate contents in a 3D representation, the sizes of icons and the candidate contents are changed depending on their locations.Namely, the icons which are located near forward are displayed with relatively large size while those which are located near backward are displayed with comparatively small size as shown in Figure 4. Icon sizes are changed depending on their locations Also swipe for candidate URLs in vertical direction and flick for rotation of candidate URLs in horizontal direction are available as shown in Figure 5.

C. Examples on Content Search Operations
The icon of the application software, ELDOXEA: elearning document search engine appears on the screen as shown in Figure 9. ELDOXEA is proposed for e-learning content search [1]- [5].In this section, example of the search operation with the keyword of "java" is demonstrated as shown in Figure 10.First, users have to key-in their keyword through the dialog box.After the research results appear on the screen, then users have to select the candidate the search results with swipe and flick operations.
When you click the icon, then Figure 11 (a) of dialog box with search start of radio button appears followed by keyboard screen as shown in Figure 11 (b).When you key-in the keyword, "java" for example, then search result of screen images appear as shown in Figure 12.Examples of swipe and flick operations are shown in Figure 13.Thus users may take a look at research results freely and easily.After that, when user click one icon of the candidate of the search results from the candidates of icons as the search results, the content of the candidate appears as shown in Figure 14.There are some keywords in the Meta tag of the source code of URLs.By using the number of common keywords in the Meta tag keywords for both content types of URLs, the distance between URLs is defined.Thus the closest URL can be retrieved using the distance between URLs.This is also specific feature of the Mashup technology of search engine.Figure 16 Small portion of the report from the users (20 of students) about ELDOXEA Although most of those reports are positive, effective search and easy to use, a small number of negative report say that merged contents are not so easy to refer.Link structure has to be clearly shown in the screen.Through experiments with 20 of students who use ELDOXEA for e-learning content search, it is found that the proposed search engine provides e-learning content retrievals comfortably and in a comprehensive manner.On the other hand, there is the report which says that merged contents which are closely related to the retrieved content are not so easy to refer.It should be solved in the future.

Fig. 1 .
Fig.1.Research background of the proposed mashup based content search engine

Fig. 2 .
Fig.2.Content type and search result selections by star slide model of representation methodFigure3shows five content types as example.Namely, there are Youtube for moving picture, Amazon.comfor product, Yahoo for document, Yahoo search for image, and Yahoo for Web searches as shown in Figure3.

Fig. 3 .
Fig.3.Five different content types are available for search

Fig. 4 .
Fig.4.Icon sizes are changed depending on their locations

Fig. 5 .
Fig.5.Swipe for candidate URLs in vertical direction and flick for rotation of candidate URLs in horizontal direction Away3D on Andoroide is used for 3D representation of icons while the APIs which are shown in Figure 6 are used for Mashup.These are aligned in the pentagonal shape as shown in Figure 7.Meanwhile, example of icon movement through swipe for the case of Yahoo search is shown in Figure 8. Candidate of URL icon is selected with swipe (up and down operations).

Fig
Fig.6.APIs used for mashup

Fig. 9 .
Fig.9.Example of display image of smartphone of which the proposed search engine is implemented github.com/legnoh/ledoxea

Fig. 15 .
Fig.15.Example of linked retrieved results (Merged contents of Yahoo search results with the other retrieved contents) III.EXPERIMENTS 20 students in the laboratory of department of information science, Saga University are participated to the experiments using the developed ELDOXEA.Then more than 30 of reports are sent as shown in Figure 16.

Fig
Fig.16.Figure16Small portion of the report from the users (20 of students) about ELDOXEA IV. CONCLUSION Mashup based content search engine for mobile devices is proposed.Example of the proposed search engine is implemented with Yahoo!JAPAN Web SearchAPI, Yahoo!JAPAN Image searchAPI, YouTube Data API, and Amazon Product Advertising API.The retrieved results are also merged and linked each other.Therefore, the different types of contents can be referred once an e-learning content is retrieved.The implemented search engine is evaluated with 20 students.The results show usefulness and effectiveness on elearning content searches with a variety of content types, image, document, PDF files, moving picture.