Saturday, March 19, 2011

Information Sharing in social network

Few years ago, majority of people located information of interest through their favourite mainstream search engines. However, with the latest evolvement of web technology, the way of finding information now has changed. Today, information searching not just involve search engine but with the influence of social network. This network can be created through latest well known site, Facebook, Twitter and many more. Social networking sites provide information that is highly tailored to an individual and comes from a highly trusted source, while the search engines provide objective data from variety of sources on variety topics.

However, the problem in this current evolution is lack of utility that can support integration between search engine and social functionality. This kind of integration in information finding is referred as social search. The term uses to unify two distinctive information discovery worlds: traditional world of web search and information sharing.

Social search is an umbrella term used to describe search acts that make use of social interactions with others. These interactions may be explicit or implicit, co-located or remote, synchronous or asynchronous (Brynn M Evans)
Social search refers broadly to the process of finding information online the assistance of social resources [Morris, M.R.]. For example, the click-thru of other searcher (Joachisms, T. 2002), the links other people make between information resources (Kleinberg, J., 1999) are used to rank the web search result. Other example is in query suggestion by using the pattern of past queries by other searcher (Cao et al., 2008).

The types of social search can be categorized into two, conduction a search over an existing database of content previously provided by other users, such as search over the collection of public Twitter post. Search over the database of socially-generated content, collaborative filtering and other similar mechanism represent implicit social activities where users of such system are not actively attempting to assist each other in their search efforts and they may not even aware that their action have influenced others’ search result.

Another type is collaborative web search where several users with similar information need, collaborate in order to fulfil the need. This type of social search is more to explicit social activities where people in the group explicitly help the other searchers in their search tasks.

Saturday, March 05, 2011

Serendipitous information discovery

Serendipity refers to an accidental yet desirable discovery. It plays a surprising role in discovery interesting information as the most interesting and valuable search results are not always the most relevant to the query. Personalization, collaborative filtering, recommender system and search engines are still trying to solve the ambiguity of serendipity in information discovery. For this reason, the academic community, were started to put more interest on personalization and recommender systems where it have become an important research area since mid-1990s. Amazon.com is one of the pioneer in e-commerce industry had give big impact to the recommendation system area. Amazon.com site promoted the idea of creating a community of customers. The company fostered this clubby feeling by encouraging readers to write and submit book reviews (which Amazon.com would post on the website) Amazon.com has been the early adopter of personalization technology to recommend products. The success of Amazon, follwoed by other type of e-commerce web site such as, Netflix, ebay etc.

Wednesday, March 02, 2011

Information discovery: from browsing to searching

In 1994, two graduate students from Stanford developed a website which was originally named 'Jerry and David's Guide to the World Wide Web' then they renamed it 'Yahoo!' ('Yet Another Hierarchical Officious Oracle'?). It was a great way for early web surfers to find the websites they were looking for, as all listings on Yahoo! were organized into hierarchical categories.

During that time, search function operated on the web directory, rather than full-text copies of web pages. Information seekers could also browse the directory instead of doing a keyword-based search. Other search engine available in this period includes Excite, Infoseek, Inktomi, Northern Light, and AltaVista. Then in year 1998, came Google, bring new idea of web search. Google make its debut by introducing the new technique from just the keywords based to page linking connectivity, the PageRank technique. They take into account the link connectivity of the Web, seeking to use the interlinking of online documents to identify the more influential and reliable sources of information.

Tuesday, February 15, 2011

Context in Web Search


The notion of context may refer to diverse range of ideas depending on the nature of work being performed. For example, context aware mobile search is a search paradigm in which application can discover and take advantage of contextual information such as user location, time of day, nearby people and devices and user activity. For web search, context can be categorized based on user's information need or tasks. Typical search engine – crawl the WWW, index the page content and search the document based on search query and return the list of relevant document. Which document are valuable depends on the context of the query i.e. education, interest and previous experience along with information about current request.

2 basic ways to recognize context

Explicit - user enter search query and determine the query category

Implicit - recognize the context by observing how user interact with the system



Wednesday, September 15, 2010

Berry Picking Model

Berry picking model is used to describe the basics of information access system. It was proposed by Bates in 1989 [1]. There are two main points in this model.

  1. User’s information need continually shift as a result of reading and learning from the information gathered during search process. Information from one point lead to another new point of information. The original information need become partly fulfilled thus lowering the priority of one goal favor of another. This to contrast the assumption of standard information retrieval that users’ information needs remain the same through out the search process.
  2. Users’ information needs are satisfied by a series of selection and information retrieve during search process. This to contrast the assumption that the search goal is focused only to the set of retrieved document into a perfect match of the original information need

This model also been supported by several others information seeking studies including [2]. They found that information seeking process consist of series of interconnected but diverse searches on one problem-based theme. The result item can lead to another new goals and search to the new direction.

We can see the new technologies in Google search engine, such as query suggestion, Google Instant and spelling correction, leading to a new dimension of web search.

  1. BATES, M. J. 1993. The design of browsing and berrypicking techniques for the online search interface. Online Information Review, 13, 407 - 424.
  2. O'DAY, V. L. & JEFFRIES, R. 1993. Orienteering in an information landscape: how information seekers get from here to there. Proceedings of the INTERACT '93 and CHI '93 conference on Human factors in computing systems. Amsterdam, The Netherlands: ACM.

Tuesday, September 14, 2010

Information Access Process - using search engine

Usually when we want to search information about something we normally have one of more goals in mind, and we will use search engine like Google to do the searching. We either asking specific question or do exhaustive research in order to get the information. There are 3 main types of information seeking tasks [1]

1. Monitor the topic over time
2. Follow the planned search activities to achieve particular goals
3. Exploring the topic in an undirected fashion


Information Searching model


The standard steps involved in search process as below. This process also been used by current web search engine [2,3]

1. Start with Information need
2. Select a system and collection to search on
3. Formulate the query
4. Send the query to system
5. Receive search result in form of result-list
6. Scan, evaluate and interpret the result.
7. If not satisfy, return to step 3 – reformulate the query
8. In satisfy with the result - stop

Users learn during search process. When they receive the research result, firstly they will scan the result item, read the title, snippet, and URL. Read the retrieved page, viewing if any of the content in the page related to the submitted query, navigating within the hyperlinked web site in the page. Normally, users will scan the information structure, titles, thesaurus terms, hyperlinks, category labels or result of clustering and then either select the result item to read in detail, use as input to formulate another query, and navigate to a new page of information. Then if still not satisfy with the information, they will formulate a query – either from the already known information or from the combination with the new information.

1. O'DAY, V. L. & JEFFRIES, R. 1993. Orienteering in an information landscape: how information seekers get from here to there. Proceedings of the INTERACT '93 and CHI '93 conference on Human factors in computing systems. Amsterdam, The Netherlands: ACM.
2. SALTON, G. 1989. Automatic text processing: the transformation, analysis, and retrieval of information by computer, Addison-Wesley Longman Publishing Co., Inc.
3. SHNEIDERMAN, B., BYRD, D. & CROFT, W. B. 1998. Sorting out searching: a user-interface framework for text searches. Commun. ACM, 41, 95-98.

Friday, August 13, 2010

Collaborative Web Search

Search engine is a central part of information seeking in the internet. Searching the information using search engine for most people almost a common daily activity. As the amount of information on the web also growing rapidly this will make search result full with numerous irrelevant information.

Searching proces in the web much easier if a person has knowledge and experience to formulate good query so that the time spent to find the right information can be reduced. Time spent to get the information is longer if the searcher know what kind of information they searching for. For instance, let say a new undergraduate student looking information about social network, maybe facing problem to get the right information in short time. In this situation the collaborative web search one of the the solution. The less experince user can utilize the collaborative funtionality in a web search as they can get recommendation from other user based on the query they submit to the search engine.

The idea of collaboration in web searching and browsing have been discussed in Twidale et al.(1997), Morris (2008), Evans and Chi (2008),Chi (2009). A study by Smyth et al.(2004) shows that collaborative web search (CWS) can exploits query repetition and selection regularity among communities of like-minded searchers in order to recommend result from underlying search engine that are likely to be especially relevant for a particular communitiy.

SearchTogether (Morris and Horvitz, 2007), CoSearch (Amershi and Morris, 2008) and Cerchiamo (Pickens et al. 2008 (Pickens et al., 2008) are example of available application
that support collaboration in web search explicitly, it can support joint search tasks for two or more searchers that have a common goal.

Thursday, August 12, 2010

Collaborative Searching

This activity involves a group of like-minded people searching together in collaborative environment. Consider a group of student planning a holiday at one of the European country. In order to plan the holiday they need to work together to find the accommodation, flight ticket, plan the itinerary, and etc. One solution is they can delegate the task to each member. Then each member will do the searching and made decision by their own then combine all the information. In collaborative search system it would allow student to search the information together so that every group member could contribute the ideas and make the decision together. Collaborative searching is depending on the location where the searcher is. For co-located collaborative search, all searchers are in the same location or sitting in front of the same computer. The other type is remote collaborative searching. This scenario happens if the searcher physically located in different places. It is either is different building, different computer.

Wednesday, August 11, 2010

Transaction log analysis

Transaction log and transaction log analysis is one approach to data collection and a research method for both system performance and user behavior. Jansen. et (2009) define transaction log as an electronic record of interaction that have occurred between a system and the user of that system. Transaction log analysis is broadly used in various information system applications. Transaction log analysis enables macro-analysis of aggregate user data and pattern and micro-analysis of individual search pattern.

Tuesday, August 10, 2010

Classification and Clustering

These two terms is a normal terms that being used in our daily routine. For example when we are in the supermarket, and want to buy fruits, we will categorize the fruits based on its type for example citrus, melons and etc.

What’s the different between classification and clustering? We use classification or categorization if we already know the label of category that we what to use. For example, in email application “spam” and “not spam”. While for clustering, it is only use when the labels for cluster not define at the beginning of the process. It allows the data to speak for itself.

Classification and clustering process are a part of machine learning problem. As in machine learning the learning process is characterized into two types, supervised and unsupervised. In supervised learning the model will learn using a set of training data then it will be applied to test data where this technique very useful in classification process. On the other hand, in unsupervised learning learn from unlabeled data and this is used in clustering technique.

Saturday, August 07, 2010

Two different views of IR problems

There are 2 different views in IR problems

1. Computer-centered
How to build efficient indexes, process user's query, ranking algorithms to improve quality of result
2. Human centered
The study of user behavior, what is their information needs

Friday, August 06, 2010

Are we browsing or searching?

Sometime when we has an interest to search about something in the net, we will Google the thing that we want to know. This 'thing' usually not clearly defined and sometime very broad.. Let say we want to get recipe from the internet, use the search engine, get the webpage, and maybe we don't know one of the ingredient, the go back to the search engine looking what is the ingredient, where to get it and etc, and at the end we end up with Tesco website. In this situation we are not do the searching but browsing the information. it is still a process of information retrieval but the main objective is not clearly defined in the beginning and the purpose might change during the interaction with the system

Thursday, August 05, 2010

Ranking algorithm

One central problem regarding IR system is how to predict the relevance of the document to the given query. Here we can use ranking algorithm to put the document in order of relevance. the document at the top is more likely to be relevant to the query. Thus, ranking algoritms are the core of IR system

Wednesday, August 04, 2010

Problems in modern information retrieval


3 main questions

1. Despite the high interactivity, people still find it is difficult to retrieve relevant information that can fulfill their information needs. Thus, in the dynamic WWW, which techniques will allow retrieval of high quality information that satisfy the needs of information to the various type of users?

2. with the increasing number of demand for information access, response time become a crucial problem. Thus, which techniques will yields faster indexes and smaller query response time?

3. The quality of retrieval task in greatly affected by user interaction with the system. Thus, how will the web search engineer can better understand the user behavior in designing and deployment the new technology in information retrieval?

Data retrieval vs information retrieval

What's the different between data retrieval and information retrieval?

Data retrieval only consider the document which can satisfy the given query while information retrieval go in depth where it also consider the semantics of the query. So the outcome from information retrieval in broader content. Result from data retrieval is more precise than information retrieval as data is aim to retrieve all object which satisfy clearly the given query. In information retrieval also dealing with natural language where the query is not in well structure text and semantically ambiguous.