Hypertext

Web-Based Information Retrieval

Patrick Alfred Waluchio Ongwen

Knowledge Management Officer

African Research and Resource Forum

Web as Agent of Change

"ICT is not an end in itself or an agent of change by itself but when incorporated into a well managed change process it is a powerful enabler and amplifier."

Bryn Jones,

Effective use of Web therefore, requires content, content management, hyperlinks and navigation tools, retrieval approaches, among others

Introduction

A critical goal of successful information retrieval on the web is to identify which pages are of high quality and relevance to user’s query.

Each search engine index web page, representing it by a set of weighted keywords

A crawler (robot or spider) performs traversal of the web with goal of fetching high quality pages for indexing and retrieval.

Introduction

Search engines must filter the most relevant information matching a user’s query and present retrieved information in a way a user will understand.

Hyperlinks provide a valuable source of information for web retrieval.

Hypertext and hypermedia enables searching the web in non-sequential manner

Challenges of Web Information Retrieval

Management of huge amount of hyperlinked pages

Crawling the web to find appropriate web sites to index

Accessing documents Measuring the quality or authority of available

information

Hypertext Non-Linear arrangements of textual material is called

hypertext. The term hyper means extension to other dimensions.

Converting text into a multidimensional space The term was invented by Ted Nelson in 1965 Hypertext

“non-sequential writing” Nelson, T. 1987. Literary Machines.

Non-linear sequences of information (dictionary, encyclopaedia, newspaper)

Hypertext are systems to manage collection of information that can be accessed non-sequentially.

Hypertext

Consists of a network of nodes and logical links between nodes

Node refer to chunks of content or web page The variety of nodes and links make hypertext a

flexible structure in which information can be provided by what is stored in nodes and links to each node.

Hypertext retrieval systems are the products of emerging technology that specifies alternative approach to the retrieval of information from web

Hypertext and Non-Linearity

Allows a user to follow their own path in a non-sequential manner to access information

Hypertext not usually read linearly Links encourage branching off

History and back button permit backtracking The immediacy of following links by clicking

creates a different experience from traditional non-linearity

Structure of Hypertext

In a hypertext system objects in a database called nodes, are connected to one another by machine-supported links. Users follow links to access information.

Text augmented with links Link: pointer to another piece of text in same or different

document Hypertext systems can accessed by selecting link to icons

and following links from node to node By searching the database for some key word in the

normal way Use of browser to view and navigate hypertext

Hierarchical Structure

Hierarchical Structure

Hierarchy is the basis of almost all websites as well as hypertext

They are orderly and provide ample navigational freedom

Users start at the home page, descend the branch that most interest them, and continue making further choices as the branch divides

Web-like Structures

Web-like structures

Relatively unsystematic and difficult to navigate

Mostly used in works of short stories and fiction in which artistic considerations may override desire for efficient navigation

Multipath Structures

Multipath Structures

Largely linear and to some extent hierarchical but offers alternative pathways hence multipath structures

Hypertext Component

Hypertext model: The run-time layer, which controls the user

interface The storage layer, which is a database

containing a network of nodes connected by links

The within-component layer, which is the content structure inside the node.

Hypermedia

(Hypermedia = Hypertext + Multimedia) Hypermedia integrate text, images, video,

graphics, sound within Web page or node Hyper- representation of textual and non-

textual information in a non-sequential manner.

Allows embedding bitmapped images (GIF, JPEG, PNG)

History of Hypertext

1945: Vannevar Bush describes “memex” (Atlantic Monthly)

1965: Ted Nelson coins the term “hypertext” 1985: Peter Brown, University of Kent,

develops first commercially available hypertext - Guide

1986-1990: More sophisticated hypertext systems developed

History …

1991: Tim Berners-Lee builds IP-based distributed hypertext system at CERN

Develops UDI/URI, HTTP, and HTML… 1993: Mosaic, first graphical Web browser,

released 2002: Work begins on Semantic Web

Hypertext and Hypermedia- In Information retrieval

Browsing – retrieve information by association Follow links, backtrack Maintain history, bookmarks

Searching – retrieve information by content Construct indexes of URLs Search by keyword/description of page

Cont… Hypertext and Hypermedia with new standards,

HTML, XHTML, have brought tremendous revolution in the creation and delivery of content, as well as access and processing of information.

Allows users to navigate within or across a range of documents from several computer networks.

Allows browsers and other software to interpret and process information for different purposes

Search engines use the links among pages to select information resources from the Internet.

Google use the link data to rank pages in order of their relevance to query.

Underlying Principles and Challenges of IR Information retrieval is a complex task Query-based IR system must be able to accept

a query about any topic and find texts that contain the specified information of query.

IR systems are required to operate in real-time, which demand they should be fast and efficient.

Most searches are conducted on the natural language text, which inherently have all the ambiguities and imprecision.

The following are some of the challenges IR systems face in natural language processing:

Synonyms Synonym occurs when different words of phrases

mean essentially the same thing. For example, the words: “finance”, “fund”,

“support”, may be related depending on context of inquiry.

Natural language is filled with many words and phrases that have similar meanings, and it is often impossible for users to provide all the words which might be relevant to the query.

To address this problem, some IR systems expand the query to include all the synonymous words for a given word with the help of thesaurus.

Polysemy

Polysemy occurs when a single word has more than one meaning. For example, the word “shot” can refer to following meanings:

A shooting, in - He shot at a tiger. An attempt, in - I took a shot at playing. A photograph, in - He took a nice shot

Phrases in Information Retrieval

Expressions consisting of multiple words often have a meaning that is substantially different from the meaning of the individual words.

The phrase “Artificial Intelligence” is different from the individual word “Artificial” and “Intelligence”, and “Operating System” is different from: “Operating” and “System”.

One method for phrase-based indexing is to use proximity measures to specify the acceptable distance between the words. WITH or NEAR

Object Recognition

Certain types of information require special procedures to identify them. For example, dates come in various forms such as: July 3, 2001, 3.7.2001, as well as 7.3.2001 (American System). (greater and less than logic <>)

Semantics and Role- Relationships

Some information can only be identified through semantics.

If a user is interested in finding out the names of lecturers teaching the courses in the area of: “Artificial Intelligence”.

First, the system must be able to know the courses related to Artificial Intelligence. These can be AI, Fuzzy logic, Genetic Algorithms, Neural Networks, Machine learning.

This should then be linked to lecturers allocated the courses.

Computable Values

Determining whether information is relevant some times depends on a specific calculation.

Suppose a user is interested in news paper article about merger in corporation that occurred after January, 1995.

The IR system must identify the documents using search logic operator LESS THAN or GREATER THAN <>

Text Representation Techniques

The purpose of IR system is to search the text database for relevant documents in real-time.

Consequently, the text database is preprocessed and stored in a structure which helps in fast searching.

This preprocessed form is called text representation

Inverted File Approach

The inverted file approach is used in text representation. It allows an IR system to quickly determine what documents contain a given set of words, and how often each word appears in the document.

In inverted file system, each database contains two files Text file –normal form in which documents appear in a

database and, inverted file- which contain all index terms drawn

automatically from the document records. Provides indirect file access

Using Probability Methods

All IR systems draw conclusions about the content of a document by examining source representation

IR must base its conclusions about the document features, such as the present or absence of particular word or phrases.

IR system must take into account these uncertain relationships to determine the strength of the relevance of a document/s to a particular request.

Relevance feedback

Relevance feedback is a technique used by some IR systems to improve performance on query by asking the user for feedback about retrieved texts.

Evaluation forms given to users to seek their views on performance of information retrieval systems.

CAT TWO- Search Engines

Compare the assigned search engines and competently comment on the following:

Structure of the search engines Indexing techniques used Information resources offered Links between nodes Search facilities and retrieval approaches used Ease of use by novice and experienced users Interface design and display of screen layout

Search Engines

Group 1: Yahoo and Lycos Group 2: Google and Alta Vista Group 3: Ask Jeeves and Excite Group 4: All the Web and HotBot Group 5: Web crawler and MSN Group 6: Dogpile and EBay