We use cookies to give you the best experience possible. By continuing we’ll assume you’re on board with our cookie policy

Over the old ages, the volume of information available through the World Wide Web has been increasing continuously ; unluckily, the unstructured nature and immense volume of information accessible over webs have made it progressively hard to happen the relevant information. The information retrieval techniques normally used are based on keywords, wherein provided keyword list does n’t see the semantic relationship between keywords nor it considers the significance of words and phrases. With such system, users often have jobs showing their information demands and interlingual rendition those demands into petitions.

To get the better of the restrictions of keyword-based information retrieval, one must believe of presenting conceptual cognition to information retrieval which will assist users to explicate their petition. The semantic cognition attached to information is united by agencies of ontologies. The function of constructs in information into conceptual theoretical accounts i.e. ontology appears to be utile method for traveling from keyword based to concept based information retrieval.

Conceptual Framework For Ontology Based Information... TOPICS SPECIFICALLY FOR YOU

We have surveyed assorted attacks and theoretical accounts of ontology based information retrieval, built on assorted techniques in a manner that resources may be retrieved based on the associations, semantic similarity, ranking algorithm, note, burdening algorithm. In the below study we listed out few of the theoretical accounts of ontology based IR, besides we tried to concentrate on assorted techniques used for retrieval efficiency, semantic similarity, ranking, burdening and notes.




Information retrieval system

Information retrieval theoretical accounts

Conceptual model for ontology based information retrieval


The beginning of ontology

What is ontology

Reasons for developing ontologies

Types of ontology

Benefits of ontology

Applications of ontology

Ontology linguistic communications

Ontology-based information retrieval system

Literature study

Ontology attacks of an IR

3.2.1 Association

3.2.2 Semantic Similarity

3.2.3 Relevance

3.2.3 Semantic Indexing

3.2.4 Semantic Note

4. Analysis

5.1 Analysis of different information retrieval theoretical accounts

5.2 Analysis of different attacks of ontology

5. Decision




1.1 Background:

In the recent old ages information retrieval system is based on keyword. The keyword-based information retrieval systems have been used to happen information and to supply entree to big sums of information. For illustration, hunt engines accept keywords as input and return as end product a list of links to paperss incorporating those keywords. However, keyword-based hunt engines have some cardinal drawbacks. Search engines do non “ understand ” the semantic significance of the words the user types into them, and hence the engines may come up with an tremendous figure of false hits and users are less able to happen related information.

To work out above job is to better information retrieval from traditional or keyword based attack to knowledge or concept based attack. Using conceptual cognition to information retrieval which will assist users to explicate their petition. The semantic cognition attached to information is united by agencies of ontologies.

Ontologies appear to be utile method for traveling from keyword- based to concept based- information retrieval.

Ontologies can be general or sphere specific, they can be created automatically or manually, and they can differ in their signifiers of representation and ways of building relationships between the constructs, but they all serve as an expressed specification of a conceptualisation.

1.2 Information retrieval system [ 1 ] :

Information retrieval trades with entree to information every bit good as it ‘s representation, storage and organisation. The overall end of an information retrieval procedure is to recover the information relevant to a given petition. The standard for complete success is the retrieval of all the relevant information points stored in a given system, and the rejection of all the non relevant ones..The consequences of a given petition normally contain both a subset of all the relevant points, plus a subset of irrelevant points, but the purpose remains, of class, to run into the ideal standard for success.

The figure ( 1 ) shows the basic construct of an information retrieval system, representation is defines as the stored information, fiting map as a certain hunt scheme for happening the stored information and questions as the petitions to the system for certain specific information.


Matching map


Figure 1: A simple theoretical account for information retrieval

Representation comprises as abstract description of the paperss in the system.nowadays, more and more paperss are full text paperss, whereas antecedently, representation was normally built on mentions to paperss instead than the paperss themselves similar to bibliographic records in library systems. Mentions to paperss are usually semi structured information with predefined slots for different sorts of information, e.g. rubric, abstract, categorization etc. , while full text paperss typically are unstructed, except for the sentence structure of the natural linguistic communication.

The fiting map in an information retrieval system theoretical accounts the system impression of similarity between paperss and questions, therefore specifying how to compare petitions to the stored description in the representation. Each theoretical account has it advantages and disadvantages with no individual scheme being superior to the others. Also you can read about the advantages and disadvantages of conditional fee agreements.

1.3 Information Retrieval Models [ 2 ] :

A theoretical account of information retrieval predicts and explains what a user will happen relevant given the user question.

A retrieval theoretical account specifies the three basic entities of retrieval:

– Representation R of information resources R,

– Representation Q ( called question ) of users ‘ information demands Q, and,

– Retrieval map M, delegating a set of resources r to each information demand Q.

The undermentioned major theoretical accounts have been developed to recover information: the Boolean theoretical account, the Statistical theoretical account, which includes the vector infinite and the probabilistic retrieval theoretical account, and the Linguistic and Knowledge-based theoretical accounts. The first theoretical account is frequently referred to as the “ exact lucifer ” theoretical account ; the latter 1s as the “ best lucifer ” theoretical accounts.

Questions by and large are less than perfect in two respects: First, they retrieve some irrelevant paperss. Second, they do non recover all the relevant paperss. The undermentioned two steps are normally used to measure the effectivity of a retrieval method. The first 1, called the preciseness rate, is equal to the proportion of the retrieved paperss that are really relevant. The 2nd one, called the callback rate, is equal to the proportion of all relevant paperss that are really retrieved. If seekers want to raise preciseness, so they have to contract their questions. If seekers want to raise callback, so they broaden their question. In general, there is an reverse relationship between preciseness and callback.

1.3.1 Exact lucifer theoretical accounts

Two theoretical accounts of information retrieval that provide exact matching, i.e, paperss are either retrieved or non, but the retrieved paperss are non ranked.

The Boolean theoretical account:

The Boolean theoretical account is based on set theory and Boolean algebra. Queries specifies as Boolean look. The retrieval schemes in the Boolean theoretical account is based upon binary determination standards which denotes either the papers to be relevant or not-relevant to a given question.

In the Boolean theoretical account, paperss questions are Boolean looks of keywords, connected by AND, OR, and NOT, including the usage of brackets to bespeak range [ 3 ] .

Ex. Q= ( “ auto ” V “ car ” V “ car ” ) ^ ( “ vacation ” V “ holiday ” )


It gives ( expert ) users a sense of control over the system. It is instantly clear why a papers has been retrieved given a question. If the ensuing papers set is either excessively little or excessively large, it is straight clear which operators will bring forth severally a bigger or smaller set.

Precise, if you know the right schemes

Easy to implement

Precise, if you have an thought of what you ‘re looking for

Efficient for the computing machine


It does non supply a ranking of retrieved paperss & A ; no weighting of index or question footings. Boolean theoretical account is the exact lucifer caused by the binary determination standard. i.e. it retrieves papers may be relevant or non relevant.

Users must larn Boolean logic

Boolean logic insufficient to capture the profusion of linguistic communication

No control over size of consequence set: either excessively many paperss or none

When do you halt reading? All paperss in the consequence set are considered “ every bit good ”

What about partial lucifers? Documents that “ do n’t quite lucifer ” the question may be utile besides

Difficult to show complex user petitions.

Difficult to command the figure of paperss retrieved.

All matched paperss will be returned.

Difficult to rank end product.

All matched paperss logically satisfy the question.

Difficult to execute relevancy feedback. Region theoretical accounts:

Regions theoretical accounts are extensions of the Boolean theoretical account that ground about arbitrary parts of textual informations, called sections, extents or parts. Region theoretical accounts model a papers aggregation as a liberalized twine of words. Any sequence of back-to-back words is called a part. Regions are identified by a start place and an terminal place.

The chief disadvantage of the Boolean theoretical account and the part theoretical accounts is their inability to rank paperss

Statistical Model

The vector infinite and probabilistic theoretical accounts are the two major illustrations of the statistical retrieval attack. Both theoretical accounts use statistical information in the signifier of term frequences to find the relevancy of paperss with regard to a question. Although they differ in the manner they use the term frequences, both produce as their end product a list of paperss ranked by their estimated relevancy. The statistical retrieval theoretical accounts address some of the jobs of Boolean retrieval methods, but they have disadvantages of their ain. Table 2.4 provides sum-up of the cardinal characteristics of the vector infinite and probabilistic attacks. We will besides depict Latent Semantic Indexing and constellating attacks that are based on statistical retrieval attacks, but their aim is to react to what the user ‘s question did non state, could non state, but someway made manifest [ 12 ] . Vector infinite theoretical account [ 7 ] :

The vector infinite theoretical account represents the paperss and questions as vectors in a multidimensional infinite, whose dimensions are the footings used to construct an index to stand for the paperss [ 4 ] . The creative activity of an index involves lexical scanning to place the important footings, where morphological analysis reduces different word signifiers to common “ roots ” , and the happening of those roots is computed. Query and papers alternates are compared by comparing their vectors, utilizing, for illustration, the cosine similarity step. In this theoretical account, the footings of a question alternate can be weighted to take into account their importance, and they are computed by utilizing the statistical distributions of the footings in the aggregation and in the paperss [ 4 ] . The vector infinite theoretical account can delegate a high ranking mark to a papers that contains merely a few of the question footings if these footings occur infrequently in the aggregation but often in the papers. The vector infinite theoretical account makes the undermentioned premises: 1 ) The more similar a papers vector is to a question vector, the more likely it is that the papers is relevant to that question. 2 ) The words used to specify the dimensions of the infinite are extraneous or independent. While it is a sensible first estimate, the premise that words are pair wise independent is non realistic.


Provides term burdening strategy

Simple, mathematically based attack.

Considers both local ( tf ) and planetary ( idf ) word happening frequences.

Provides partial matching and graded consequences.

Tends to work rather good in pattern despite obvious failings.

Allows efficient execution for big papers aggregations


Missing semantic information ( e.g. word sense ) .

Missing syntactic information ( e.g. phrase construction, word order, propinquity information ) .

Premise of term independency ( e.g. ignores synonymy ) .

Lacks the control of a Boolean theoretical account ( e.g. , necessitating a term to look in a papers ) .

Probabilistic theoretical account:

The probabilistic retrieval theoretical account is based on the Probability Ranking Principle, which states that an information retrieval system is supposed to rank the paperss based on their chance of relevancy to the question, given all the grounds available [ 4 ] . The rule takes into history that there is uncertainness in the representation of the information demand and the paperss. There can be a assortment of beginnings of grounds that are used by the probabilistic retrieval methods, and the most common one is the statistical distribution of the footings in both the relevant and non-relevant paperss.

We will now depict the state-of-art system developed by Turtle and Croft ( 1991 ) that uses Bayesian illation webs to rank paperss by utilizing multiple beginnings of grounds to calculate the conditional chance P ( Info.need|document ) that an information demand is satisfied by a given papers. An illation web consists of a directed acyclic dependence graph, where borders represent conditional dependence or causal dealingss between propositions represented by the nodes. The illation web consists of a papers web, a construct representation web that represents indexing vocabulary, and a question web stand foring the information demand. The construct representation web is the interface between paperss and questions. To calculate the rank of a papers, the illation web is instantiated and the resulting chances are propagated through the web to deduce a chance associated with the node stand foring the information demand. These chances are used to rank paperss.

1.3.3 Latent Semantic Indexing

In LSI the associations among footings and paperss are calculated and exploited in the retrieval procedure. The premise is that there is some “ latent ” construction in the form of word use across paperss and that statistical techniques can be used to gauge this latent construction. An advantage of this attack is that questions can recover paperss even if they have no words in common. The LSI technique captures deeper associative construction than simple term-to-term correlativities and is wholly automatic. The lone difference between LSI and vector infinite methods is that LSI represents footings and paperss in a decreased dimensional infinite of the derived indexing dimensions. As with the vector infinite method, differential term weighting and relevancy feedback can better LSI public presentation well.

Linguistic and Knowledge-based Approachs

In the simplest signifier of automatic text retrieval, users enter a twine of keywords that are used to seek the upside-down indexes of the papers keywords. This attack retrieves paperss based entirely on the presence or absence of exact individual word strings as specified by the logical representation of the question. Clearly this attack will lose many relevant paperss because it does non capture the complete or deep significance of the user ‘s question. The Smart Boolean attack and the statistical retrieval attacks, each in their specific manner, seek to turn to this job. Linguistic and knowledge-based attacks have besides been developed to turn to this job by executing a morphological, syntactic and semantic analysis to recover paperss more efficaciously [ Lancaster and Warner 1993 ] . In a morphological analysis, roots and affixes are analysed to find the portion of address ( noun, verb, adjectival etc. ) of the words. Following complete phrases have to be parsed utilizing some signifier of syntactic analysis. Finally, the lingual methods have to decide word ambiguities and/or generate relevant equivalent word or quasi-synonyms based on the semantic relationships between words. The development of a sophisticated lingual retrieval system is hard and it requires complex cognition bases of semantic information and retrieval heuristics. Hence these systems frequently require techniques that are normally referred to as unreal intelligence or adept systems techniques.

1.4 Conceptual frame work for Ontology based information retrieval system:

Fig. Conceptual Framework for ontology based information retrieval system

The stairss involved in these beds are

Question parsing

To recover the information for the demand of user, acquire the question from them. Divide the question into meaningful words and use the word stemming procedure.

Word stemming

Linguistically, words follow morphological regulations that allow a individual to derive discrepancies of a same thought to arouse an action ( verb ) , an object or construct ( noun ) or the belongings of something ( adjectival ) . For case, the undermentioned words are derived from the same root and portion an abstract significance of action and motion.

Activate – & gt ; Activates, Activated

The word “ Activate ” is used to stand for the words “ Activates and Activated ” . Steming does the contrary procedure: it deduces the root from a to the full suffixed word harmonizing to its morphological regulations. These regulations concern morphological and inflectional postfixs. The former type normally changes the lexical class of words whereas the latter indicates plural and gender and it besides removes the unwanted words like a, an, the etc. ( Porter Stemmer [ 6 ] )

For illustration, a list of root words

Stop_words = ( “ the ” , “ and ” , “ a ” , “ to ” , “ of ” , “ in ” , “ I ” , “ is ” , “ that ” , “ it ” , “ on ” , “ you ” , “ this ” , “ for ” , “ but ” , “ with ” , “ are ” , “ have ” , “ be ” , “ at ” , “ or ” , “ as ” , “ was ” , “ so ” , “ if ” , “ out ” , “ non ” ) ;

Ontology Matching

After dividing the question into meaningful words, each word should be checked against the ontology. All the combination of words is taken for processing. Specific sphere ontology is taken to verify whether the word is present in that ontology. If yes so the relationship of the words are taken into the consideration.

Weight Assignment

The weight is assigned to each word with regard to other word harmonizing to the relationship in ontology like superclass, immediate subclass, subclass etc based on improved matching [ 4 ] algorithm.

Standards 1: If the two stemmed words are non present in the ontology or any one word is non

nowadays in the ontology so the weight is assigned as nothing.

Standards 2: If the root word is a direct superclass of another word and the so the

weight assigned is 1

Standards 3: If the root word is a direct subclass of another word and the so the weight

assigned is 0.5

Standards 4: If the root word is a subclass of another word and the so the weight assigned is

1/level of relationship.

Standards 5. If the root word is a superclass of another word and the so the weight

assigned is ( 1/2+ ( 1/level of relationship ) ) .

Rank Calculation and Information retrieval

The cumulative weight is calculated for each combination of words based on improved duplicate algorithm. The best papers gets the lower limit mark. The paperss are arranged in go uping order harmonizing to their cumulative mark.

For illustration, Ontology for Academic service

Academic establishment needs sums of paperss to be maintained. Once if the papers care is made machine-controlled so it is really easy for the academician to recover the relevant paperss. Any academic establishment has to keep the paperss like ( I ) Admission Details ( two ) Course Details ( three ) Department Details ( four ) List of programmes conducted by each section ( V ) Student inside informations ( six ) Staff Details ( seven ) Accounts ( eight ) Conferences and workshops organized ( nine ) Placement inside informations ( ten ) Examination inside informations etc.

The Fig. shows the ontology for academic services [ 8 ] . The Ontology is created by holding the root node as “ thing ” so follows the assorted classs like Administration, Controller of scrutiny, Academic sections, and arrangement. Each bomber class has many other nodes related to it


Administration Controller of scrutiny Departments Placement

Account subdivision office

Engineering engineering school

CSE EE EXTC Engineering

Circulars category stuff pupils staff

Meeting trial conference

Academicians ontology

Sample Query Processing

This subdivision shows how the above described conceptual frame work helps in efficient retrieval of paperss for the question “ I want to cognize the Communications Security Establishment section inside informations of psgtech ”

QUERY Processing








1 0

2 0

3 1

4 1.5

5 0

6 2


The minimal mark is 1.the inside informations that are present in the CSE ( subclass pf section ) node are retrieved. The following minimal mark is 1.5. All the section inside informations are retrieved.






Communications security establishment






Communications security establishment


0 0 0 0



Communications security establishment







Communications security establishment








0.5 0

Detailss 0






Communications security establishment






Communications security establishment




0 0

0 0

0 1

0 0 1


The above figure shows the diagrammatic representation of the proposed model. The end product of each stage is besides shown in Fig. In weight assignment stage for each combination of words the mark is calculated depending upon the above specified standards. In ranking the aggregative weight is calculated for each combinations and it is sorted in go uping order. In our illustration the lower limit mark is 1. So the papers which comes under department- & gt ; Communications Security Establishment is retrieved. The user besides needs the same.

This method provides the new manner of seeking the contents on the domain/web. It finds the relevant papers for the user question utilizing the techniques called word stemming, ontology matching, weight assignment, rank computation etc. In the ontology fiting stage an improved matching algorithm is used to better the relevance of retrieval. The Query parsing and word stemming stage is farther extended by including query enlargement technique, remainder of the stages are improved farther by adding societal notes and give voice degree matching.



2.1 The beginning of Ontology:

The term “ ontology ” has been used for a figure of old ages by the Artificial Intelligence & A ; knowledge representation community but is now going portion of the standard nomenclature of a much wider community including information system modeling.

The term is borrowed from doctrine, where ontology intend a systematic history of being.

The term ontology has been applied in many different ways, but the nucleus significance is a theoretical account for depicting the universe that consists of a set of types, belongingss, and relationship types ( Garshol, 2004 ) . In the context of cognition sharing, Gruber ( 1993a ) uses the term ontology to intend a specification of a conceptualisation. Gruber ( 1993b ) defines conceptualisation as “ an abstract, simplified position of the universe that we wish to stand for for some intent. ” Every cognition direction system is committed to some conceptualisation, explicitly or implicitly ( Gruber, 1993b ) . Ontology can assist specify the relationships among resources and happen related resources. It is of import to stress that there are multiple relationships between specific words and constructs. This means that in pattern: 1 ) different words may mention to the same construct, and 2 ) a word may mention to several constructs.

2.2 What is Ontology?

An ontology is “ the specification of conceptualisation, used to assist plans and worlds portion cognition ” .

An ontology is a set of concepts- such as things, events, & A ; dealingss that are specified in some manner in order to make an in agreement upon vocabulary for interchanging information.

In information direction countries and cognition sharing countries, ontology can be defined as follows:

An ontology is a vocabulary of constructs and dealingss rich plenty to enable us to show cognition and connotation without semantic ambiguity.

Ontology describes sphere cognition and provides an in agreement -upon apprehension of a sphere.

Ontologies are aggregations of statements written in a linguistic communication such as RDF that defines the dealingss between constructs & A ; stipulate logical regulations for concluding about them.

Main Definition of ontology:

“ An ontology is a formal, expressed specification of a shared conceptualisation “

“ explicit ” means that “ the type of constructs used & A ; the restraints on their usage are explicitly defined ” ;

“ formal ” refers to the fact that “ it should be machine clear ” ;

“ shared ” refers to the fact that “ the cognition represented in ontology are agreed upon and accepted by a group ” ;

“ conceptualisation ” refers to an a abstract theoretical account that consists the relevant constructs and the relationships that exists in a certain state of affairs.

The conceptualisation consists of,

The identified constructs ( objects, events, etc )

For ex: Concepts: disease, symptoms, therapy

The conceptual relationships that are assumed to be and to be relevant

For ex: Relationships: “ disease causes symptoms ” , “ therapy dainties disease ”

2.3 Reasons for developing ontologies?

An ontology defines a common vocabulary for research workers who need to portion information in a sphere. It includes machine-interpretable definitions of basic constructs in the sphere and dealingss among them.

Why would person desire to develop ontology? Some of the grounds are:

aˆ? To portion common apprehension of the construction of information among people or

package agents

aˆ? To enable reuse of sphere cognition

aˆ? To do sphere premises explicit

aˆ? To divide sphere cognition from the operational cognition

2.4 Types of ontologies

( 1 ) Top-level ontology,

( 2 ) Domain ontology,

( 3 ) Task ontology, and

( 4 ) Application ontology.

First, top-level ontology describes really general constructs like infinite, clip, and events, which are independent of peculiar jobs or spheres. Second, sphere ontology describes the vocabulary related to a generic sphere by specialising the constructs introduced in the top- degree ontology. Third, undertaking ontology describes the vocabulary related to a generic undertaking or activity in the top-level ontologies.

Finally, application ontology is the most specific of ontologies. Concepts in application ontologies frequently correspond to functions played by sphere entities while executing a certain activity.

Depending on the broad scope of undertaking to which the ontologies are put ontologies can change in their complexness. Ontologies range from simple taxonomy to extremely embroil webs including restraints associated with constructs and dealingss.

Light Weight Ontology


‘is-a ‘ hierarchy among constructs

Relationss between constructs

Heavy Weight ontology

Cardinality restraints

Taxonomy of dealingss

Axioms ( limitations )

In practical footings, developing an ontology includes:

specifying categories in the ontology,

set uping the categories in a systematic ( subclass-superclass ) hierarchy,

specifying slots and depicting allowed values for these slots,

make fulling in the values for slots for cases.

We can so make a cognition base by specifying single cases of these categories make fulling in specific slot value information and extra slot limitations.

2.5 Benefits of ontology:

To ease communicating among people and administrations: assistance to human communicating and shared apprehension by stipulating significance

To ease communicating among systems without semantic ambiguity: i.e. to accomplish inter-operability

To supply foundations to construct other ontologies ( reuse )

To salvage clip and attempt in constructing similar cognition systems ( sharing )

To recycle sphere cognition.

To do sphere premises explicit: ontological analysis clarifies the construction of cognition & A ; let sphere to be explicitly defined and described.

2.6 Application countries of ontologies:

Information Retrieval:

As a tool for intelligent hunt through illation mechanism alternatively of keyword matching

Easy irretrievability of information without utilizing complicated Boolean logic

Cross linguistic communication information retrieval

Improve callback by query enlargement through synonymy relation.

Improve preciseness through word sense disambiguation ( designation of the relevant significance of a word given context among all its possible significances )

Natural linguistic communication processing:

Better machine interlingual rendition.

Questions utilizing natural linguistic communication.

Knowledge direction:

As a cognition direction tools for selective semantic entree ( intending oriented entree ) .

2.7 Ontology linguistic communications:


Resource Description Framework: RDF is a model for depicting Web resources, such as the rubric, writer, alteration day of the month, content, and copyright information of a Web page. RDF was designed to supply a common manner to depict information so it can be read and understood by computing machine applications.


Jenaa„? is a Java model for constructing Semantic Web applications. Jena provides a aggregation of tools and Java libraries to assist you to develop semantic web and linked-data apps, tools and waiters.

The Jena Framework includes:

an API for reading, processing and composing RDF informations in XML, N-triples and Turtle formats ;

an ontology API for managing OWL and RDFS ontologies ;

a rule-based illation engine for concluding with RDF and OWL informations beginnings ;

shops to let big Numberss of RDF three-base hits to be expeditiously stored on disc ;

a question engine compliant with the latest SPARQL specification

waiters to let RDF informations to be published to other applications utilizing a assortment of protocols, including SPARQL

2.8 An attack for Ontology-based Information Retrieval [ 20 ]

The logic based IR provides a sound platform to ground about the significance of an information resources ‘ content in the retrieval procedure, i.e. about the relevancy of that significance for the user ‘s information demand. In that manner, the user can happen resources that are relevant for his question even if there are no syntactical similarities between them. It is clear that the quality of the retrieval depends on the measure and quality of the sphere cognition that is available to the logical thinking procedure. Indeed, a logical system can recover a papers about autos for a question for vehicle, if and merely if there is a officially described statement that a auto is a type of vehicle

Therefore, in order to enable retrieval of all semantic relevant resources for a question, the cognition about sphere has to be consistently acquired and described in the signifier of a sphere theory. Furthermore, in order to decide the “ anticipation job, ” the sphere theory as to be normally shared, i.e. a sort of common understanding about the used vocabulary should be. Since ontologies represent explicit and formal specifications of the conceptualization of a sphere of involvement, they seem to be really suited for the extension of the logic-based IR systems in the above mentioned manner.

Fig.2.8.1 ontology description [ 20 ]

Ontology-based Information Retrieval Model [ 20 ]

The Retrieval Model

The ontology-based theoretical account for information retrieval redefines the undertaking of IR as an extraction from a given depository of information resources, of those resources r that, given question Q, makes the formula O|- R a†’ Q valid, where R and Q are expression of the chosen logic, “ a†’ ” denotes the trade name of logical deductions formalized by the logic in the inquiry and O is a set of logical sentences called sphere cognition ( ontology ) . A derivability relationship |- is defined between a set of expression and a expression, if there exists a finite sequence of the illation regulations that leads the set of expression to that expression.

Fig.2.8.2 ontology based retrieval theoretical account [ 20 ]

For the ontology-based IR, we have the undermentioned reading of the basic retrieval theoretical account

presented in subdivision 1.2:

– LRes = IsB ( O ) , i.e. a resource is modelled as a set of relation cases ( facts ) from the corresponding cognition base. This set can be treated as one of case averments ; so the dealingss ( constructs ) of which a fact is asserted to be an case constitute wholly the description of the resource ;

– LQuery = I© ( O ) , i.e. a question is modelled as an ontology-based question Q ( O ) ; the intuitive significance of this pick is that all resources represented by facts retrieved for question Q ( O ) , i.e. the set of facts F ( Q ( O ) ) , should be retrieved ;

– Iridium = I ( O ) aS† LRes, i.e. a depository ( aggregation ) of information resources represents a set of all concept instantiations ;

– Meter ( I ( O ) , Q ( O ) ) , the fiting map between the depository and the given question, is implemented through logical illation defined by the logical linguistic communication used for stand foring ontology O.

In this theoretical account we used bottom-up fix-point rating process. It means that some functions in M are defined implicitly through the maxims from set A ( O ) .They can let the

specification of lexical, “ thesaural ” cognition every bit good, i.e. they contribute to the specification of the significance of the footings used in both document representation and question preparation. In the inferring procedure this sort of cognition is brought to bear ( and therefore serves as ) “ background cognition ” harmonizing to which questions are to be interpreted. The positive consequence is that maxims are in fact a recall-enhancing mechanism, because they support the find of resources relevant to the question that would hold otherwise gone undetected.

Given a retrieval theoretical account, the interaction with a simple ontology-based retrieval system may be described as follows ( see Figure 2. ) .The set of information resources and their belongingss is represented as a set of cases in the cognition base IsB ( O ) . A user ‘s information demand is conceptualised in an ontology-based question Q ( O ) . This question is matched against the set of information resources, M ( I ( O ) , Q ( O ) ) and the set of replies F ( Q ( O ) ) is returned to the user.


Ontology-based attacks

Ontology-based attacks are characterized by the usage of extremely elaborate conceptualisations in the signifier of ontologies and KBs. They provide formal descriptions of the significances involved in user demands and contents. Therefore, these theoretical accounts have better opportunities to accomplish the alleged semantic hunt paradigm.

4.1 Semantic Association Analysis in Ontology-based Information Retrieval

This system is based on semantic web. Semantic Web. RDF and SPARQL linguistic communications do non adequately supply a question mechanism to detect the composite and inexplicit relationships between the resources. Such complex relationships are called semantic associations [ 8 ] . The procedure of detecting semantic associations is besides referred to as semantic analytics.

4.1.1.Semantic Association Analysis

The conventional and semantic supported hunt attacks typically respond to user questions by returning a aggregation of links to assorted resources. Users have to verify each papers to happen out the information they need, in most instances the reply is a combination of information from different resources. Relationss are at the bosom of Semantic Web [ 10 ] . Concentrating on Semantic Web engineerings, the accent of hunt will switch from seeking for

paperss to happening facts and practical cognition. Relation searching is a particular category of hunt methods which is concerned with stand foring, detecting and construing complex relationships or connexions between resources.

Sheth et Al ( 2005 ) discourse an algorithm developed to treat different sorts of semantic associations utilizing graph traverse algorithms at the ontology degree. The relationships between two entities in the consequences of a semantic question could be established through one or more semantic associations. In this instance the semantic associations could be represented by a graph which shows the connexion between entities. It is besides of import to treat and prioritise the semantic association based on user penchants and the context of hunt. There are besides ranking algorithms proposed based on different prosodies to rate the semantic association [ 10 ] .

The semantic association analysis consists of several cardinal procedures and constituents like discuss ontology development, informations set building, semantic association find, semantic association ranking, consequences presentation, and public presentation rating s severally. However, there are besides other of import issues such as entity disambiguation, informations set care and so on. Here Ontology Development: utilizing protege tool

4.1.2. Data Set Construction:

The information should be selected from extremely dependable Web sites which provide informations in structured, semi-structured, parseable unstructured signifier or with database backend. Structured information is preferred ( i.e. RDF or OWL ) . Semi-structured or parseable unstructured informations ( i.e. XML ) can be transformed to structured informations utilizing xPath or XSLT. Data with rich metadata and dealingss is preferred. For illustration, for a “ Computer Scientist ” category, the beginning besides provides “ reference ” , “ state ” attributes every bit good as some dealingss with other categories such as “ Research Area ” , “ Publication ” , “ Organization ” . The information set should hold rich dealingss and big sum of cases which are extremely connected.

4.1.3. Semantic Association Discovery Algorithms

Semantic association find can be seen as a particular category of semantic hunt taking to happen out complex relationships between entities. The job can be generalized as reciting all possible waies between any two nodes in a semantic graph. The hunt is performed utilizing ontologies and semantic informations sets. The construction of the ontology constrains the possible waies that one can take from one node to another. Typically the construction of the ontology or relation between categories is simple ; nevertheless, the dealingss between cases in the knowledge-base ( i.e. cases ) might be really complicated depending on the connection of the graph.

4.1.4 Semantic Association Ranking

Ranking mechanism is an of import portion of a hunt engine. Ranking algorithm reflects the cognitive idea of human existences towards the ranking of existent universe objects harmonizing to their perceived importance. The PageRank algorithm contributes to Google ‘s success and it is one of the most of import grounds that most people prefer to utilize it. Most of the current hunt engines rank paperss based on vector infinite theoretical account. In semantic association analysis, an of import undertaking is integrating the most meaningful associations out of all detected dealingss. However, new ranking algorithms demand to be developed in order to use the advantages of Semantic Web engineerings.

4.1.5 Presentation:

The identified semantic associations could be presented to users in a meaningful manner which is able to assist users understand the significance of entities. We have implemented an machine-controlled interactive multimedia presentation coevals engine, called MANA, to build interactive multimedia presentations based on paperss associating to entities in semantic associations.

Different constituents in a semantic enhanced information hunt and retrieval system [ 17 ]

The entities are hyperlinked to those paperss which are able to supply external accounts that help users to research relevant information sing a submitted question. Figure shows different constituents and beds in an ontology-based information hunt, retrieval and presentation.

Finally we conclude that, semantic analytics country and in peculiar find and reading of complex dealingss between entities in a knowledge-base. In Semantic Web, semantic analytics demonstrates important importance in assorted application spheres by enabling hunt mechanism to detect and treat meaningful dealingss between information resources.

4.2 Relevance Information Retrieval based on ontology: ( ONTOBROKER SYSTEM ) [ 20 ]

4.2.1 Relevance

Relevance is one of the most of import constructs in the theory of IR. The construct arises from the consideration that if the user of an IR system has an information demand, so some information stored in some resources in the information depository may be “ relevant ” to this demand. In other words, the information to be considered relevant to a user ‘s information demand is the information that might assist the user to fulfill his information demand. Any information that is non considered relevant to a user ‘s information demand is to be considered “ irrelevant ” to that information demand. This is a effect of accepting construct of relevancy.

Therefore, given a set of information resources and a question, the undertaking of the retrieval procedure is to recover those resources, and merely those whose information content is relevant to the information content of the question. Because relevancy is of import in the graded retrieval for the quality of the retrieval procedure. The importance of relevancy is the chief ground why the logical formalization of information retrieval is a non fiddling job:

foremost, in finding the relevancy of a resource to a question, the success or failure of an deduction associating the two is non plenty. It is necessary to take into history the uncertainness built-in in such an deduction,

2nd, the debut of uncertainness can besides be motivated by the consideration that a aggregation of resources can non be considered as a consistent and complete set of statements. In fact, resources in the aggregation could and frequently do contradict each other in any peculiar logic and non all the necessary cognition is available, and,

eventually, what is relevant is decided by the user from session to session and from clip

to clip, and is so to a great extent dependent on judgements where extremely subjective and barely consistent factors are brought to bear.

Due to its conceptual nature, ontologies provide an ideal abstraction degree, on the top of conditional logical thinking, for specifying this flexible impression of relevancy, which we will name conceptual relevancy. This relevancy will be explained in item in the following subdivision. ontologies represent a conceptual theoretical account of a sphere, they seem to be an ideal beginning for specifying this epistemic position on relevancy. Furthermore, the conceptual impression of relevancy is one of cardinal features ( advantages ) of the ontology-based information retrieval

Sing conceptual degree, there are two positions on the relevancy of a resource R for an information demand expressed in question Q, we define in following two definitions.

4.2.2 Conceptual Relevance

There are different readings of chance that can be used for ciphering relevancy [ 163 ] . Traditionally, one can understand chance from the frequence point of position. That is, chance is a statistical impression, refering itself with the statistical Torahs of opportunity. On the other manus, chance can be interpreted as the grade of belief-the epistemic position. This position concerns the assignment of beliefs in propositions. Different readings of the theory of chance lead to different attacks for patterning relevancy in information retrieval.

Therefore, in the traditional relevancy theoretical accounts, relevances are obtained merely by numbering the figure of resources incorporating a peculiar form or index term. The statement for utilizing the statistical impression of relevancy is that chances should be viewed as a step of opportunity at the execution degree. However, the disregard of other accounts of relevancy at the conceptual degree is possibly the beginning of troubles in the conventional relevancy theoretical account. Since ontologies represent a conceptual theoretical account of a sphere, they seem to be an ideal beginning for specifying this epistemic position on relevancy. Furthermore, the conceptual impression of relevancy is one of cardinal features ( advantages ) of the ontology-based

information retrieval.

Sing conceptual degree, there are two positions on the relevancy of a resource R for an information demand expressed in question Q, we define in following two definitions.

Collection relevancy:

It represents relevancy of the resource sing the given information depository ( so called aggregation relevancy, in the impression ColRel ) :

ColRel: I ( O ) x KB ( O ) a†’ Roentgen

Explanation relevancy

It represents relevancy of the retrieval procedure M ( see old subdivision ) in which a resource ( i.e. consequence of a question ) is retrieved ( so called account relevancy, in the impression ExpRel ) : ExpRel: M x I ( O ) a†’ Roentgen

Fig.4.2.2 The superior procedure in the Ontobroker retrieval [ 20 ]

we presented the formal theoretical account for the ontology-based information retrieval, as an extension of the bing logic-based IR theoretical accounts, particularly in specifying the impression of the relevancy. We proposed a comprehensive theoretical account for the relevancy, the alleged Conceptual relevancy that theoretical accounts non merely whether an information resource is relevant for a question, but furthermore why ( and accordingly, how strong ) a resource is relevant for a question. In that manner, by uniting the relevancy sing how ( i.e. why ) an information is retrieved in a retrieval system ( the alleged Explanation relevancy ) and how semantically is this information related to other relevant information ( the alleged Collection relevancy ) , we tried to mime the relevancy concluding found in human existences.

4.3.Ontology based Information Retrieval by Semantic Similarity ( SSRM ) [ 18 ]

Measures of semantic similarity and relatedness for usage in ontology-based information retrieval. The implicit in hypothesis is that by widening the classical information retrieval theoretical accounts to include the cognition contained in ontologies covering the sphere of the information base, we obtain agencies for bring forthing better replies to user questions.

Better replies are, in this context, chiefly a more powdered ranking of information base objects, which is obtained by working better methods for calculating the similarity or relatedness between a question and objects from the information base.

Semantic similarity between constructs is typically calculated utilizing merely the information available in a construct inclusion hierarchy. However, semantic relatedness between constructs can be viewed as the sum of the overall interconnectedness between the constructs in inquiry, sing a wider figure of semantic dealingss.

In retrieval undertaking, a user poses a question stand foring an information demand to the system. The information retrieval system must fulfill the user ‘s information demand by analyzing both the question and the paperss and so showing a list of paperss to the user that are found relevant to that peculiar question. This list of paperss is the consequence of a duplicate procedure that compares each papers with the question. The chief map of the analysis of the question is to deduce a representation that can be matched with the papers representation. One manner of including the cognition contained in the ontology is to take a representation formalism where questions and objects are described utilizing a construct linguistic communication and where the looks can be straight mapped into the ontology. We can so cipher the similarity between the description of the question and the descriptions of the objects, based on the nearness rule derived from the ontology.

One attack to a nearness rule is concluding over the ontology. The attack in this is based on a relatedness step between constructs, derived from the construction and dealingss of the ontology, which is so used to execute query enlargement. By making so, we can replace semantic matching from direct logical thinking over the ontology with numerical similarity computation by agencies of a general collection rule. This has at least two advantages in an information retrieval context. The first is that it allows for partial matching of questions, and the 2nd is that it is less clip consuming.

4.3.1 Representation of ontologies:

The chief construct is steps of similarity derived from the construction and dealingss of an ontology for usage in information retrieval. The purpose is hence to place the type of ontology formalism, every bit good as the chief ontology constituents needed to organize the footing for deducing and ciphering similarity.

Ontologies have typically been represented utilizing frames [ 11 ] , conceptual graphs [ 12 ] , first-order logic or description logics [ 13 ] . Dominant in the last five old ages are new representation strategies based on description logic linguistic communications, such as OIL, and OWL [ 14 ] .

4.3.2 Introduction to Description Logic:

Description logic is a cognition representation formalism that represents the cognition of an application sphere by specifying the relevant constructs and functions of the sphere, and so utilizing these constructs and functions to stipulate belongingss of objects and persons happening in that sphere. Concepts are sets of persons and functions are binary relationships between persons. The atomic constructs and functions can, by agencies of construct builders be combined into complex descriptions.

Apart from the representation formalism, description logic offers concluding capablenesss that allow for the illation of inexplicit cognition, means for limited querying, and support for the designation of contradictory constructs. A cognition base system based on description logic consists of two constituents, a TBox and an ABox.

The TBox describes the construction of a sphere in footings of categories ( constructs ) and belongingss ( functions ) . The description consists of a set of terminological maxims, which are statements about how the constructs and functions are related to each other. This means that in description logic, constructs are defined deliberately in footings of descriptions that specify what belongingss objects must hold to belong to a certain category.

The ABox, consists of averments about named persons, utilizing the constructs and functions defined in the TBox.

4.3.3 The description linguistic communication:

The name of a description logic denotes the construct builders available. AL is a Attribute Language, as introduced Schmidt-ScauB and Smolka [ 15 ] . Concept description in A L are formed harmonizing to the undermentioned sentence structure regulations, where A denotes atomic constructs, C and D denote complex constructs, and R denotes atomic functions

Syntax regulations for AL are,

C, D – & gt ; A ( Atomic construct )

T ( cosmopolitan construct ) etc.

4.3.4 Semantic Similarity Measures:

A manner of mensurating semantic similarity in a semantic web is to measure the distance between the constructs being compared, where shorter distance means higher Shortest Path Length.The semantic similarity steps are

Weighted Shortest Way

Depth-relative Scaling Approaches

Information Content

Hierarchical Concept Graphs Leaden Shortest Way:

Another simple edge-counting attack was presented in Bulskov et Al. [ 16 ] for usage in information retrieval. We argued that construct inclusion ( ISA ) intuitively implies strong similarity in the opposite way from inclusion ( specialisation ) . In add-on, the way of the inclusion ( generalisation ) must lend some grade of similarity. fraction of an. With mention to this undermentioned ontology, the atomic construct Canis familiaris has high similarity to the constructs poodle and Alsatian.

The step respects the ontology in the sense that every construct subsumed by the construct Canis familiaris by definition bears the relation ISA to chase. The intuition is that, to a question on Canis familiaris, an reply including cases poodle is satisfactory ( a particular reply to a general question ) . Because the ISA relation evidently is transitive, we can by the same statement include farther specialisations, e.g. to include poodle in the extension of animate being. However, similarity working the taxonomy should besides, as was the instance in Rada ‘s attack, reflect distance ” in the relation. Intuitively greater distance ( longer way in the relation graph ) corresponds to smaller similarity.

Specialization Property

Concept inclusion implies strong similarity in the opposite way of the inclusion. Furthermore, generalisation should lend to similarity. Of class, it is non purely right, but because all Canis familiariss are animate beings, animate beings are to some grade similar to Canis familiariss. Therefore, the belongings of generalisation similarity should be exploited. However, for the same grounds as in the instance of specialisations, transitive generalisations should lend a reduced grade of similarity.

Generalization Property

Concept inclusion implies reduced similarity in the way of the inclusion. A concept inclusion relation can be mapped into a similarity map in conformity with the two belongingss described above and the minimum distance belongings as follows. Assume an ontology given as a sphere cognition relation.

The above fig. can be viewed as such an illustration. To do “ distance ” influence similarity, we assume the ISA relation to be transitively reduced.

An illustration: ontology with relation ISA covering pets [ 18 ]

Similarity reflecting “ distance ” can so be measured from way length in the graph matching to the ISA relation. A similarity map “ sim ” based on “ distance ” , dist ( X, Y ) in ISA should hold following belongingss:

Sim: U x U – & gt ; [ 0,1 ] , where U is the existence of constructs

Sim ( x, Y ) =1 merely if x=y

Sim ( x, y ) & lt ; sim ( x, omega ) if dist ( x, y ) & gt ; dist ( x, omega )

Properties 2 and 3 correspond to the Identity belongings and the Minimal Distance Property severally. By parameterized with two factors I? and I? , showing similarity of immediate specialisation and generalisation severally, we can specify a simple similarity map as follows. A way between nodes ( constructs ) ten and Y utilizing the ISA relation.

P= ( p1, . . , pn )


Pi ISA Pi+1 or p i+1 ISA pi

for each I with x=p1 and y= pn.

Given a way P= ( p1, aˆ¦. , pn ) , set s ( P ) to the figure of specialization and g ( P ) to the figure of generalization alog the way P, as follows,

s ( P ) = | { i|Pi ISA Pi+1 } |


g ( P ) = | { i|Pi+1 ISA Pi } |

If P1 ; .. ; Pm are all waies linking ten and Y, so the grade to which Y is similar to x can be defined as follows.

simWSP ( x, y ) = max { I?s ( pj ) I?g ( Pj ) }

We denote this step sim ( x ; y ) WSP ( Weighted Shortest Path ) , as similarity between two constructs ten and Y is calculated as the maximum merchandise of weights along the waies between x and y. This similarity can be considered as derived from the ontology by transforming the ontology into a directional leaden graph, with I? as downwards and I? as upwards weights, and with similarity derived as the merchandise of the weights on the waies. Figure shows the graph matching to the above.

As such, the step is in conformity with the Specialization Property, the Generalization Property, and the Identity Property, because there is an border with weight 1 from every construct to itself. Furthermore, it conforms with the Minimal Distance Property, whereminimal ” is interpreted as the maximum amount of the merchandise of weights along all possible waies between two constructs. A widely acknowledged job with the shortest-path attacks is that they typically rely on the impression of unvarying distance in the taxonomy. This

implies, as mentioned antecedently, that non all borders ( nexus ) denote the same distance and hence non the same similarity. There have hence been assorted efforts at scaling the web by integrating the place in the taxonomy of the constructs being compared

[ 18 ] The ontology transformed into a directed leaden graph, with the immediate specialisation and generalisation similarity values I? = 0.9 and I? = 0.4 as weights. Similarity is derived as the maximal ( multiplicative ) weighted way length, and therefore sim ( poodle ; Alsatian ) = 0:4 A¤ 0:9 = 0:36.

4.4 Semantic cognition note

All the significances and information conveyed by content in unstructured signifier ( such as text or audiovisual content ) can non in general be to the full translated to a clear and formal semantic representation, for both matter-of-fact ( cost ) and intrinsic ( jobs for the formalisation of the universe ) grounds. However, it is possible to officially depict parts of the conveyed information, albeit to an uncomplete extent, as metadata. Metadata is informations about other informations ( e.g. , the ISBN figure and the writer ‘s name are metadata about a book ) . For the same ground that it is by and large utile to maintain both parts of information ( information and metadata ) in the system, it is besides relevant to hold a nexus that connects the two of them, normally known as note.

Different syntactic supports and criterions have been proposed for the representation of metadata and notes. Markup languages like HTML and XML are widespread presents, but they have restrictions in their expressiveness and portion ability ( Passin, 2004 ) . Ontology-based engineerings have been developed in the last few old ages to turn to and get the better of these restrictions. For illustration, conceive of a papers that contains the keyword “ panther ” . This keyword is equivocal because it might mention to the animate being or to the auto. An ontology-based note can associate the word “ panther ” , looking in the papers, to an ontology construct that defines “ jaguar ” as the abstract construct “ carnal ” , therefore taking any ambiguity.

A study of ontology-based engineerings for semantic note is reported in ( Uren, et al.,2006 ) . This work proposes a papers centric theoretical account for ontology-based semantic note that manages three elements: ontologies ( metadata ) , paperss ( informations, or content in unstructured signifier ) and notes ( links between the informations and the metadata ) . They identify seven demands for ontology-based semantic note systems:

aˆ? Standard formats: utilizing standard formats is preferred whenever possible because the investing in doing up resources is considerable and standardisation physiques in future proofing because new tools, services, etc.

User centered/collaborative design: in the instance of manual note tools, it is important to supply users with easy to utilize interfaces that simplify the note procedure and topographic point it in the context of their every twenty-four hours work.

aˆ? Multiple Ontology support: note tools need to be able to back up multiple ontologies. For illustration, in a medical context, there may be one ontology for general metadata about a patient and other proficient ontologies that deal with diagnosing and intervention.

aˆ? Support of heterogenous papers formats: criterions for note tend to presume that the paperss being annotated are in Web-native formats such as HTML and XML. However, with the outgrowth of new multimedia content in the Web, paperss will be in many different formats ( sound, picture, etc ) .

aˆ? Document development: Ontologies and paperss change continuously, which means that the

note procedure should non be fixed.

aˆ? Annotation storage: The ontology-based semantic note theoretical account assumes that notes will be stored individually from the original paperss. However, many tools store the notes as built-in portion of the paperss and hence they do non uncouple informations and metadata.

aˆ? Automation: an of import facet of easing the cognition acquisition constriction is the proviso of installations for automatic grade up of papers aggregations. To accomplish this, the integrating of cognition acquisition engineerings into the note environment is critical.

The work in ( Uren, et al. , 2006 ) besides analyzes different note tools sing this seven note demands. Fig 3.8 shows a comparing sing the first six demands while Fig 3.9 represents merely the mechanization demand. As we can see in Fig 3.9, many systems have some sort of automatic and semi-automatic support for notes

4.4.1 TAP is proposed as a Web-based hunt system where paperss and constructs are nodes likewise in a semantic web [ ] . This work views the Semantic Web as a large web incorporating resources matching non merely to media objects ( such as Web pages, images, audio cartridge holders, etc. ) as the current Web does, but besides domain objects like people, topographic points, organisations, and

Share this Post!