Information retrieval

Information retrieval (IR) is the science of searching for information in documents, searching for documents themselves, searching for metadata which describe documents, or searching within databases, whether relational stand-alone databases or hypertext networked databases such as the Internet or intranets, for text, sound, images or data. There is a common confusion, however, between data retrieval, document retrieval, information retrieval, and text retrieval, and each of these has its own bodies of literature, theory, praxis and technologies.

The term "information retrieval" was coined by Calvin Mooers in 1948-50.

IR is a broad interdisciplinary field, that draws on many other disciplines. Indeed, because it is so broad, it is normally poorly understood, being approached typically from only one perspective or another. It stands at the junction of many established fields, and draws upon cognitive psychology, information architecture, information design, human information behaviour, linguistics, semiotics, information science, computer science, librarianship and statistics.

Automated information retrieval (IR) systems were originally used to manage information explosion in scientific literature in the last few decades. Many universities and public libraries use IR systems to provide access to books, journals, and other documents. IR systems are often related to object and query. Queries are formal statements of information needs that are put to an IR system by the user. An object is an entity which keeps or stores information in a database. User queries are matched to documents stored in a database. A document is, therefore, a data object. Often the documents themselves are not kept or stored directly in the IR system, but are instead represented in the system by document surrogates.

In 1992 the US Department of Defense, along with the National Institute of Standards and Technology (NIST), cosponsored the Text Retrieval Conference (TREC) as part of the TIPSTER text program. The aim of this was to look into the information retrieval community by supplying the infrastructure that was needed for such a huge evaluation of text retrieval methodologies.

Web search engines such as Google and Lycos are amongst the most visible applications of information retrieval research.

Performance measures

There are various ways to measure how well the retrieved information matches the intended information: The formulas for precision, recall and fall-out are translated from the german Wikipedia-article "Recall und Precision". See also this nice intuitive, graphical depiction.

Precision

The proportion of retrieved and relevant documents to all the documents retrieved:

{\mbox{precision}}={\frac {|\{{\mbox{relevant documents}}\}\cap \{{\mbox{retrieved documents}}\}|}{|\{{\mbox{retrieved documents}}\}|}}

In binary classification, precision is analogous to positive predictive value. Precision can also be evaluated at a given cut-off rank, denoted P@n, instead of all retrieved documents.

Note that the meaning and usage of "precision" in the field of Information Retrieval differs from the definition of accuracy and precision within other branches of science and technology.

Recall

The proportion of relevant documents that are retrieved, out of all relevant documents available:

{\mbox{recall}}={\frac {|\{{\mbox{relevant documents}}\}\cap \{{\mbox{retrieved documents}}\}|}{|\{{\mbox{relevant documents}}\}|}}

In binary classification, recall is called sensitivity.

Fall-Out

The probability to find an irrelevant among the retrieved documents.

{\mbox{fall-out}}={\frac {|\{{\mbox{irrelevant documents}}\}\cap \{{\mbox{retrieved documents}}\}|}{|\{{\mbox{retrieved documents}}\}|}}

F-measure

The weighted harmonic mean of precision and recall, the traditional F-measure or balanced F-score is:

F=2\times \mathrm {precision} \times \mathrm {recall} /(\mathrm {precision} +\mathrm {recall} ).\,

This is also known as the $F_{1}$ measure, because recall and precision are evenly weighted.

The general formula for non-negative real α is:

F_{\alpha }=(1+\alpha )\times \mathrm {precision} \times \mathrm {recall} /(\alpha \times \mathrm {precision} +\mathrm {recall} ).\,

Two other commonly used F measures are the $F_{0.5}$ measure, which weights precision twice as much as recall, and the $F_{2}$ measure, which weights recall twice as much as precision.

Mean average precision

Over a set of queries, find the mean of the average precisions, where Average Precision is the average of the precision after each relevant document is retrieved.

Where r is the rank, N the number retrieved, rel() a binary function on the relevance of a given rank, and P() precision at a given cut-off rank:

\operatorname {Ave} P={\frac {\sum _{r=1}^{N}(P(r)\times \mathrm {rel} (r))}{\mbox{number of relevant documents}}}\!

This method emphasizes returning more relevant documents earlier.

Model types

classification of IR-models (translated from German entry, original source Dominik Kuropka)

For successful IR, it is necessary to represent the documents in some way. There are a number of models for this purpose. They can be classified according to two dimensions like shown in the figure on the right: the mathematical basis and the properties of the model. (translated from German entry, original source Dominik Kuropka)

First dimension: mathematical basis

Set-theoretic Models represent documents by sets. Similarities are usually derived from set-theoretic operations on those sets. Common models are:

Algebraic Models represent documents and queries usually as vectors, matrices or tuples. Those vectors, matrices or tuples are transformed by the use of a finite number of algebraic operations to a one-dimensional similarity measurement.
- Vector space model
- Generalized vector space model
- Topic-based vector space model (literature: [1], [2])
- Extended Boolean model
- Enhanced topic-based vector space model (literature: [3], [4])
- Latent semantic indexing aka latent semantic analysis

Probabilistic Models treat the process of document retrieval as a multistage random experiment. Similarities are thus represented as probabilities. Probabilistic theorems like the Bayes' theorem are often used in these models.
- Binary independence retrieval
- Uncertain inference
- Language models
- Divergence from randomness models

Second dimension: properties of the model

Models without term-interdependencies treat different terms/words as not interdependent. This fact is usually represented in vector space models by the orthogonality assumption of term vectors or in probabilistic models by an independency assumption for term veriables.

Models with immanent term interdependencies allow a representation of interdependencies between terms. However the degree of the interdependency between two terms is defined by the model itself. It is usually directly or indirectly derived (e.g. by dimensional reduction) from the co-occurrence of those terms in the whole set of documents.

Models with transcendent term interdependencies allow a representation of interdependencies between terms, but they do not allege how the interdependency between two terms is defined. They relay an external source for the degree of interdependency between two terms. (For example a human or sophisticated algorithms.)

Open source information retrieval systems

ASPseek
DataparkSearch, search engine written in C
Fluid Dynamics Search Engine (FDSE) An open source search engine written in Perl, freeware and shareware versions are available
GalaTex XQuery Full-Text Search (XML query text search)
ht://dig Open source web crawling software
Glimpse and Webglimpse [5] advanced site search sowfware
iHOP Information retrieval system for the biomedical domain
EBIMed Information retrieval (and extraction) system over Medline
Information Storage and Retrieval Using Mumps (Online GPL Text)
Lemur Language Modelling IR Toolkit
Lucene [6] Apache Jakarta project
mnoGoSearch the renowned SQL search engine
MG full-text retrieval system Now maintained by the Greenstone Digital Library Software Project
SMART Early IR engine from Cornell University
Sphinx Free open-source SQL full-text search engine
Terrier Information Retrieval Platform
Tiny Search Engine written in C++
Wumpus multi-user information retrieval system
Xapian Open source IR platform based on Muscat
Zebra GPL structured text/XML/MARC boolean search IR engine supporting Z39.50 and Web Services
Zettair, compact and fast search engine written in C, able to handle large amounts of text

Major information retrieval research groups

Major figures in information retrieval

Awards in this field: Tony Kent Strix award.

ACM SIGIR Gerard Salton Award

1983 - Gerard Salton, Cornell University: "About the future of automatic information retrieval"
1988 - Karen Sparck Jones, University of Cambridge: "A look back and a look forward"
1991 - Cyril Cleverdon, Cranfield Institute of Technology: "The significance of the Cranfield tests on index languages"
1994 - William S. Cooper, University of California, Berkeley: "The formalism of probability theory in IR: a foundation or an encumbrance?"
1997 - Tefko Saracevic, Rutgers University: "Users lost: reflections on the past, future, and limits of information science"
2000 - Stephen E. Robertson, City University, London: "On theoretical argument in information retrieval"
2003 - W. Bruce Croft, University of Massachusetts, Amherst: "Information retrieval and computer science: an evolving relationship"
2006 - C. J. van Rijsbergen, University of Glasgow, UK: "Quantum haystacks"

External links