Searching the Internet for information is one of the most if not the most popular activity. People use Internet for various reasons but searching is one of the most common and popular reason. Searching is a widely and extensively researched issue and still remains the one. In recent times it has become an industry and Google has emerged as the clear leader. But others like Yahoo, Bing, etc. are also giving a stiff competition and trying hard to win people by introducing different tweaks and optimizations in the interfaces and implementation. Also the information available over the Internet is increasing exponentially and it is becoming harder every day to find desired information.
Most search engines do well in finding known and factual information e.g. Capital of a country, information about a movie, data of birth, and other information like this. Searching problems catering such information have a very good success rate and modern search engines have seemed to solve the problem of general information searching.
But things start going wrong when someone doesn't know what he is searching or what should he search? When someone tries to learn about a new domain or a subject but he has no idea about the field then what he does, he writes a small query related to his domain and tries to find the information using his preferred search engine. The problem with this approach is that a small query will return many results but since the user has no general idea about the domain, he starts exploring different results without knowing what he should search. This is where the contemporary search engines lack in their functionalities. It is not that developers haven't yet realized this problem; in fact a lot of work is being done in this area. But it still remains an issue to search and explore.
2.2 History of Exploratory Search
Exploratory search has been around for quite some time in some shape or another, but it emerged as an individual and significant field of study in 2005 when a workshop on exploratory search interfaces was held at University of Maryland (White et al. 2005). This workshop brought together researchers from different domains like Information Retrieval (IR), Information Seeking (IS), Human Computer Interaction (HCI), etc and various issues related to exploratory search were discussed including but not limited to interfaces, evaluation, cognitive processes, etc. This workshop proved to be a cornerstone for the field of exploratory search and brought to attention many new ideas and dimensions. It motivated researchers to set out on the path of exploratory search and explore these issues to find a reasonable solution to a very challenging problem.
This Workshop inspired researchers to write a special issue of Communications of the ACM in 2006 (White et al 2006). This issue discussed the progress made thus far in the field of exploratory search. It discussed the different systems supporting exploratory search and prevailing issues and directions for supporting it at bigger scale and levels.
In 2006 an ACM SIGIR workshop was held with the objective to focus on the techniques and tools necessary to evaluate existing exploratory search systems since it is imperative to analyze the impact they make on overall performance of the users while searching.
In 2008 a workshop was held by the name of "Information Seeking Support systems". Researchers realized that the term "exploratory search" can be confusing sometimes so they tried to put it under the broader category of Information Seeking Support Systems (ISSS). This workshop urged the researchers to come up with new ideas related to three main objective; better HCI models, new tools and techniques for supporting ISSS, and new techniques and way to evaluate ISSS. Researcher delivered many new ideas and various issues were also pointed out regarding models, tools, and evaluation.
In March 2009, IEEE Computer Magazine featured ISSS as the cover feature and published recent development and new directions for the future.
2.3 Explanation of Exploratory Search
Exploratory search, as the name suggests deals with the situations when the basic purpose of searching of a user is exploration. The field of exploratory search was started with the idea to improve the techniques, tools and technologies to help searchers in the process of exploration. Its aim is to make process of searching more engaging for user, where the user has more command and liberty to specify the information needs and play and manipulate results. Different researchers have tried to define in exploratory search in various ways. There are different terms used in literature like exploratory search, information seeking support system, interactive search, human computer information retrieval, all describing the same concept. Every type of search involves bit of exploration but there are certain properties which can be associated with exploratory search.
White et al. (White et al 2006) stated that exploratory search is one which deals with the situations when the information need demands exploring the field while the users have no or very little knowledge about the domain they want to search and they don't know how to form the query or choose proper keywords. It also caters to the situation when the required information in not indexed properly.
Marchionini (Marchionini 2006) divided all types of information needs into three main categories lookup, learn, and investigate and stated that all search activities fall under these three categories. He further described that lookup activities are the most basic type of search activities and mostly user succeed in finding their required information rather easily. Search activities in learn and investigate categories constitute exploratory search. Learning about some domain or investigating a query are the tasks which are harder to accomplish using the traditional search engine, are the main target of exploratory search.
Fig 2.1 Search Activities (Marchionini 2006)
2.4 Non-Exploratory Search
A question comes to the mind that what is not exploratory search. Daniel Tunkelang who is an active researcher in the field of exploratory search has answered this question in his one blog post (Tunkelang 2008). The definition of exploratory search tells us that ambiguous requirements, inability to form exact queries, and limited knowledge of domain are some of the characteristics of exploratory search. So, according to Daniel Tunkelang if we are sure about our requirements and know exactly what we want to complete the task at hand and have enough knowledge to build queries and ask it precisely then this type of search is considered non-exploratory search. Queries formed for exploratory search tasks don't result in either success or failure and if a task is completed by giving a query and getting an exact result in return then this task in non-exploratory. The reason is that exploratory search is a consistent process of learning and knowledge building. The tasks which can be completed shortly and exactly don't fall into the category of exploratory search.
2.5 Exploratory Search Systems
Many Exploratory search systems have been developed over the years. Examples include mSpace, Flamenco, Apple's iTunes, Relation Browser, etc. Below is a brief overview of some of these systems.
2.5.1 Relation Browser
Relation Browser (Marchionini & Brunk 2003) aimsto provide an overview of a web space by exploring relationships among attributes in a small set and simple interaction mechanisms. It basically provides enhanced and more interactive overview of a web space as compared to simple sitemap and visualizations. It is important to provide an overview since then user is in a better place to understand the web space and it also helps him in exploration process since he has a general idea of what he can find here. So Relation Browser seeks to improve the situation further by providing additional aid and overviews.
2.5.2 Faceted Search
Faceted search is perhaps the most representative form of exploratory search. Whenever we talk about exploratory search the first thing which comes to mind is faceted browsing or faceted search. The reason is that the most systems which target exploratory search include some sort of faceted search. Examples are mSpace, Flamenco, Apple's iTunes, etc.
Faceted search allows users to search a domain based on its attributes (Schraefel 2009). It promotes learning and enables users to draw new ideas about the domain they are searching by just looking at interface. The reason is since the domain is divided into its attributes called facets so a user can have a look at the interface and can extract certain information about the domain easily. It helps him in better understanding the domain and guides him in exploration of his required information especially when he has no substantial knowledge of the domain.
A brief description of the few popular faceted search systems is described below.
2.5.3 mSpace
mSpace (Wilson & Schraefel 2008) is a faceted browser that lets users perform not only keyword search but also faceted search. The domain is divided into its attributes which are shown in columns and search criteria can be specified by making selections in different columns. Columns are not fixed and can be modified i.e. new columns and attributes can be added or deleted. A useful feature present in mSpace is of backward highlighting. If a user any columns from middle or right then not only this filter is applied to right columns but also all the related item to the left are also highlighted which makes it easier for users to make connections about the domain. Another feature is the ability to perform OR function in a column i.e. multiple items within a column can be selected. Sharing of information is also possible through the use of tagging, comments, and discussions. Preview cues can also be provided on demand which allows previewing the kind of information present in selected item and in turn help users in better exploring and querying the domain.
2.5.4 Apple's iTunes
iTunes (Schraefel 2009) is a multimedia application which lets users play different songs and tracks. It provides the option to build a library from local and online collection and enables to search the collection on the basis of facets. Faceted search is performed on the basis of three facets from music domain i.e. Genre, Artist, and Album. Selection made in any of the columns filters the results. It helps in drawing new conclusion about domains as well as provides unknown information. For example if a user selects the genre Jazz and then see an artist which he thought in the past that he only sings Pop music then this will be new discovery for him. Its interface is pretty similar to mSpace but backward highlighting feature is not provided in it.
2.5.5 Flamenco
Flamenco (Yee et al. 2003) is a faceted browser which makes use of rich metadata and hierarchical categorization to help user to guide to his required information. FLAMENCO stands for FLexible information Access using MEtadata in Novel COmbinations. The information is organized in such a way that exploration of domain becomes rather easy and user can form and refine queries in more assured manner. User is presented with multiple choices to make his selection which enhances the chances of user finding his required information in any of the facets. Text based search is also provided along with faceted search.
2.6 Limitations of Current Search Engines for Supporting Children
Technology is advancing at a rapid pace and it has penetrated into our daily lives so much so that we feel helpless sometimes in its absence. Computers and Internet are now very common and can be found in many households. So children are exposed to Internet and Computers at very early stages of their lives. At early ages from 3-6 they use it mostly for playing games, watching cartoons, and other fun stuff. Then they start using it for searching and browsing information related to their school work, assignments, and other general information. But children face certain problems when they start growing and start using Internet more and more and these problems are caused by their limited skills and abilities due to their early age. Most pages and search engines are made by keeping adults in mind (Jochmann-Mannak 2009). Developer might have some thoughts about children at the back of their minds but most interfaces are developed for general audiences which are mostly adults. Designers don't usually think that children have special needs and will suffer from certain problems while using this interface.
Searching for the required information seems an easy and straight forward task; just go to the search engine of your choice, type in a query, press Enter, and there you are, your have the desired results. But, as simple as it might sound, it is not that easy especially for children. People have to through a complete process to find their desired information. When a child wants to explore a certain piece of information then first of all he needs to know what he wants to search. Since, children have limited vocabulary they find it difficult choose a proper keyword for their information need. Their requirement might be clear in their mind but it is difficult for them to put it in two or three words. So they decide to put a longer query in natural language style when in doubt. Children are not good and typing. They have difficulties in finding the right keys which makes the process of typing very slow. Also they make a lot of spelling and typing mistakes (Druin et al. 2009) which results in wrong keyword being input to the search engine which will obviously return wrong results. When a query is input to a search engine it returns millions of result against that query. Children even adults cannot handle this unnecessary overload of information and it is also a known fact that most relevant results are usually found on the first page. So users really go to second or third result page. So this irrelevant load of information confuses the children and makes the task of comprehending the results difficult. So basically children suffer from following problem while using a search engine to find information:
Choosing appropriate keywords to form queries due to limited vocabulary
Inefficiency in typing which makes the process of typing slow and also results in typing errors
Spelling mistakes because of limited knowledge of grammar
Comprehension of results due to the overload of information
As mentioned above, children suffer from these problems due to negligence of developers to specifically support children in the process of searching. We cannot expect children to follow the same steps for searching the information which are taken by adults. Their age factor demands specific consideration. They require special adjustments to be made to interface to enable them to search in a better way. Faceted search seems to provide the building block to cater to the problems of children in their hunt for information. So it's worth putting some efforts to investigate the impact faceted search can have on the search performance of children. It will be interesting to know the results because if the results are positive then researchers and developers will have a new line of direction to work. They can target specifically children and provide them with the improved interfaces based on faceted search since children are fast becoming a major part of Internet population.
Faceted search has the structure and properties to not only accommodate the above mentioned problems faced by children when searching but also makes the process of searching more interactive and intuitive. The ability of faceted search to split the domain in its attributes or facets automatically solves few of the problems that children have to deal with when they start exploring for information. Since the domain is already divided in various facets children don't have to think about keywords, they don't type so they don't make spelling mistakes. Everything is presented in front of them and they just have to make appropriate selections. Also, the presentation of results is also done in a more instinctive and comprehensible manner and helps children in make connections, and hence enables them to perceive results and make better selections. So faceted looks promising in the forefront but it needs to be analyzed and evaluated to see how much helpful it actually is.
2.7 Related Work
The faceted search has received a lot attention in the recent past. A lot of research is currently in progress to enhance the applications of faceted search. It has become a kind of fashion to provide faceted search along with typical keyword search. There are even talks of supporting faceted for open web (Teevan et al. 2008) but there are still many challenges in doing so as the authors have also identified some of these challenges. We can find various examples of faceted search in our daily use of Internet. Sometimes we notice it and many other times we don't even know that it's faceted search or something different from typical search. EBAY, Amazon, ACM, IEEE, etc. are few examples of websites where we can search for our desired items using the combination of faceted and keyword search. It is generally thought that faceted search can work well in specific domains since these domains have clearly defined metadata and lend them well for the implementation of faceted search. This is the reason why faceted search is very popular in the domains of shopping, online stores, groceries, music, images, etc. because these domains allow defining facets clearly and different items can be searched on the basis of these facets. For example in an online store we can find a mobile phone or any other item of our choice on the basis of different facets like price, manufacturer, operating system, etc. along with the combinations of keyword search. What actually happen is that user generally enters a tentative query and results are presented in front of him on the basis of that query. After that user can refine these results according to different categories or facets provided. Similarly, in digital libraries like ACM user can give a simple query related to the document he wants to find and matching documents are returned. After that, user can refine the results on the basis of different facets like people, publication, conference, etc. The advantage of all this is that user is never really stuck anywhere. He has something to work with all the time. In this way more interaction is provided to play with results of the initial query. It's not like typical search engines where hundreds of results are available but all you can do is to check each of the items manually to see if it fits the requirements. So faceted search not only saves the time and efforts, it's also more effective and efficient if implemented properly.
Researchers have worked on various different projects related to faceted research over the past few years and have generated lot of ideas and directions to work on in the coming periods. Below is a brief overview the work that has been done related to the faceted search.
Faceted Search or faceted browsing both terms point to the same thing and both are used interchangeably throughout the text. According to Wilson faceted browsing falls into two main styles (Wilson & Schraefel 2008)
Traditional faceted search
Directional column faceted search.
In traditional style of faceted search, collections are filtered automatically on the basis of the selections made in different facets and these filtered results can be refined further. FLAMENCO uses this style of faceted browsing.
aa
Fig 2.2 FLAMENCO Faceted Browser on Nobel Prize Winners
In directional column faceted browser the facets are presented column wise and any selection made to the facets in left column filters the result in right columns. Examples of systems which implement this type of faceted search are mSpace and iTunes.
mspace
Fig 2.3 mSpace Faceted Browser on Online Newsfilm Archive
Social networking has become a craze in the past few years and it is now a fashion statement to use facebook, twitter, LinkedIn, etc. It's hard to imagine now that whoever uses Internet doesn't use any social networking website. So it is becoming increasingly important for social networking websites to provide tools to its users to find people they know e.g. their friends, class-mates, coworkers, etc. accurately and speedily.
Fig 2.4 LinkedIn Faceted People Search
LinkedIn has taken up the challenge and realizing the potential of faceted search they have provided the faceted search to support people search. Different facets are provided and user can manipulate the results on the basis of these facets until he/she finally finds the desired target. It is easy to understand and use and gives user more control over the results set and hence is much more interactive than typical static keyword based search.
Mainstream search engines like Google, Yahoo, Bing, etc. are now thinking about providing exploratory search. No one is directly providing it yet but there are talks of and even some efforts exerted on exploratory search for WWW. But there is still a long way to go to get any closer to the aim of providing faceted search for the open web as identified by the researchers at Microsoft (Teevan et al. 2008). The very nature of open web makes it a tough candidate to apply faceted search. The diverse nature of information available on the web and the lack of quality metadata for each document are few of the many hurdles in the way of faceted search for open web. Google tried its stint at faceted search by implementing Google Squared. What Google Squared does is it tries to find simple facts related to any term entered as input query by traversing web and the results are provided in an organized fashion. This organized presentation of information helps user in drawing an image of the domain and this way user get the better idea of the fields and hence is in a better position to perform further actions. Quite recently, Google has launched Google Instant to support users in the process of exploration.
One interested application of faceted navigation can be found on Nuggetize1 webpage.
Nuggetize traverses the web to find the most relevant information related to the query provided, analyzes each piece of information comprehensively, and organizes and presents the information in a categorized manner called nuggets. The idea is to help user get started in their quest of information by providing them exact and to the point information. It gives users more options to interact with the results and reformulate their queries so users are not stuck at a point and don't feel helpless, as they don't have to think about what to do now. Users can also provide feedback about whether a result matches their requirements or not and the results are refreshed accordingly.
Nuggetize
Fig 2.5 Nuggets for the query "faceted search"
http://www.nuggetize.com
Elastic lists (Stefaner and Muller 2007) are used to improve already existing faceted datasets by providing a dynamic and more visually appealing view. It assigns weights to the items and various comparisons can be made at a glance. Animations are used when different filters are applied so that users get the idea of what is going on as compared to an abrupt change in view which happens in faceted interfaces without elastic lists.
ElasticLists
Fig 2.6 Elastic Lists on Nobel Prize Dataset
2.8 Interfaces for Children
When it comes to the web searching, children are never the central figure. Search engines are rarely developed keeping the special needs of children in mind. Grown ups almost always play the role of target users by default. It is not that children have been ignored altogether. There are few notable researchers who are very active in the field of designing interfaces specifically for children considering all their special needs and demands. Even in the past people have made efforts in this direction but this work is few and far between when compared with the work done for adults. Despite, search engines which have been designed specifically for children are not comprehensive and do not help children much more than any traditional search engine will do. All they do is filtering of the results according to the needs and age of children, and objectionable pages are omitted. These search engines do not help children much in formulating queries or avoiding mistakes (van Kalsbeek and de Wit 2007).
The problem with traditional search engines is that they are designed implicitly for adults keeping their needs and abilities in mind. But there is a vast difference between the cognitive and analytical skills of children and adults. Both neither think nor act alike. Kids have limited set of mental and other expertise as they are still in the process of learning and growth and their skills are not yet developed to the full potential.
Researchers believe that children feel more comfortable in using browsing based search interfaces rather than keyword based search engines. The reason they give is that browsing tools are based on the concept of recognition and kids feel at ease to use such interfaces because it does not require a lot of effort on the part of children to look for information (Borgman et al. 1995).
Why children face problems in using keyword based or traditional search engines? People have explored and observed the searching habits of children and various difficulties they face and have found out the key problem areas. According to Borgman et al. children face problems in retrieving information from information retrieval systems due to their lack of skills in the following disciplines:
Typing
Spelling
Vocabulary
Alphabetizing
Boolean Logic (Borgman et al. 1995)
Bilal and Kirby suggested that children need to learn more and adapt themselves to situations. They need to understand each task after careful analysis and evaluation and should plan accordingly to put them in a position to make better use of information retrieval system (Bilal and Kirby 2002).
As mentioned above, one of the main reasons of why children face many problems while using the search engines to find information is that search engines are designed by adults for adults mainly. On the other hand, children are still in a phase where their developmental growth consisting of many aspects like cognitive, physical, emotional, and social is still in progress (Cooper 2005). It is not yet reached at a point where it is comparable to the abilities possessed by adults at a much older age. So it is unfair to assume that children will be able to utilize the facilities provided by the search engines in the same way as adults. Even many adults face various problems in using the search engines even though they are much more experienced and skilled. So, how can we expect children to be good at it?
Druin et al. (Druin et al. 2009) studied the keyword searching behavior and patterns of children aging 7-11 and found vital areas where the research is lacking. They closely worked with children and observed the problem areas for children where they face difficulties and commit errors. They made valuable and important revelations. The established that despite all the advancement in technology children still are hampered by the very basic problems i.e. typing mistakes, spelling errors, and keyword selection due to lack of vocabulary. Even the aids provided by search engines to minimize such problem such as auto complete and spelling suggestions features are ignored by children since children are busy looking at keyboard to avoid mistakes hence cannot focus on the screen. Children are also not good at complex queries when they have to search more than one item at a time. Children are not comfortable with so many results returned against a single query which confuses them and they don't usually go beyond first few results or the first page. All these things create a sense of frustration in children and they want more help by search engine designers. Druin et al. reported that children and their parents want the search interfaces to be more innovative and interactive. They want interfaces which have prefixed and predefined categories so that they can make choices and search for information by clicking and don't have to type. Children also want that their queries should return limited set of results which cater to their needs and irrelevant results are minimized.
There are examples of few interfaces which are designed specifically for children such as KidsClick, International Children Digital Library (ICDL), Ask for Kids, Yahoo! Kids, etc. Some of these are good while some are not quite to the mark. But none of these systems or any other system fully captures the needs of children and provides the best solution. Below is a brief description of some of these systems.
KidsClick! is a search engine designed by librarians especially for children. Its specialty is that some children related material and pages are assigned to relevant categories which are further divided into sub categories. So that children can find the required information by exploring the appropriate categories and subcategories.
kidsclick
Fig 2.7 KidClick! Web Search for Kids
When a child clicks a specific category the links of relevant pages assigned to that category are shown. It's a neat, simple and easy to understand interface and a keyword search is also provided along with categorical search. Its disadvantage is that these choices are not exclusive and do not cover a lot of information. Also it is a static interface with no updates with no changes to predefined categories or pages overtime.
search results for kidsclick
Fig 2.8 Search results for category "Internet" at KidsClick
Safe Search for Kids powered by Google is keyword based search engines and returns only those pages which are relevant to the needs of children by using the safe filter.
SafeSearch
Fig 2.9 Google Safe Search for Kids
It has a very simple interface just like Google. It provides a ranked list of results depending on the keyword. Google directory for kids and teens is also provided along with search engine. It also has the disadvantage of being static and not much of interaction is provided for children.
results of safesearch
Fig 2.10 Result of query "science" at Google Safe Search for Kids
Search engines mentioned above and many other designed for children are mainly static and lack the quality of interaction and refinement of search results. Children themselves have to find the appropriate information from the list of ranked result which is not an easy task for them. International Children Digital Library is designed specially for children and consists of books for kids in many languages and related to many subjects. Its search engine is devised keeping the special needs of children in mind and the thought process they follow. It is based on different facets or attributes of the books so children can search a book on the basis of its properties. Children can search a book on the basis of its color, its length, its subject, and age group. It is more visually appealing to children.
simple search ICDL
Fig 2.11 Simple search for Red Cover and Six to Nine age group books at ICDL
An advanced search interface is also provided where many other facets or options are provided along with the keyword search. It provides the chance to children to be more interactive with the results and they can refine their queries on the basis of many other choices available. But there is a drawback that it takes some time for children to getting used to such type of interface because they are accustomed to using traditional search engines. But there is no doubt that this interface does a good job of facilitating children in their information search.
advance search ICDL
Fig 2.12 Advance search for books of ten to thirteen age group and rectangle shape at ICDL
2.9 Conclusion
Many researchers have evaluated the usefulness and effectiveness of the different exploratory and faceted search systems. Evaluation typically involves around the different tasks defined to test the efficiency of the system or improvements made to the interface. Some studies show many systems have the potential to be very successful exploratory search systems and worth the effort and attention exploratory search is getting. Some studies have failed to show the desired results due to various reasons e.g. the lack of growth of technology, improper implementation, etc. Even though faceted search might be more effective and useful in certain situations, it is not preferred over typical search by users (Kules et al. 2009). The reasons could be many depending upon the context and situation e.g. the lack of familiarity of users, partiality towards typical style of searching, or the absence of the elements of an interface for searchers.
There is no doubt that faceted navigation has potential. But interfaced developed so far do not make use of this potential to a full extent. It has been hit and miss kind of journey so far. Also, faceted search is still limited to particulars domains like item selection, stores, shopping, etc. and works best for them. So if we want to take the faceted search to the next level then there is still a lot of work to be done as there is a great room for improvement in faceted search interfaces.
