Overview of Information Retrieval (CS598-CXZ Advanced Topics in IR Presentation) Jan. 18, 2005 ChengXiang Zhai Department of Computer Science University of Illinois, Urbana-Champaign

What is Information Retrieval (IR)? Narrow-sense: –IR= Search Engine Technologies (IR=Google, library info system) –IR= Text matching/classification Broad-sense: IR = Text Information Management: –Gneral problem: how to manage text information? –How to find useful information? (info. retrieval) (e.g., google) –How to organize information? (text classification) (e.g., automatically assign to different folders) –How to discover knowledge from text? (text mining) (e.g., discover correlation of events)

Why is IR Important? More and more online information in general (Information Overload) Many tasks rely on effective management and exploitation of information Textual information plays an important role in our lives Effective text management directly improves productivity

Elements of Text Info Management Technologies Search Text Filtering Categorization Summarization Clustering Natural Language Content Analysis Extraction Mining Visualization Retrieval Applications Mining Applications Information Access Knowledge Acquisition Information Organization

A Quick Tour of the State of the Art….

Component Technology 1: Natural Language Processing

What is NLP? … يَجِبُ عَلَى الإنْسَانِ أن يَكُونَ أمِيْنَاً وَصَادِقَاً مَعَ نَفْسِهِ وَمَعَ أَهْلِهِ وَجِيْرَانِهِ وَأَنْ يَبْذُلَ كُلَّ جُهْدٍ فِي إِعْلاءِ شَأْنِ الوَطَنِ وَأَنْ يَعْمَلَ عَلَى مَا … How can a computer make sense out of this string ? Arabic text - What are the basic units of meaning (words)? - What is the meaning of each word? - How are words related with each other? - What is the “combined meaning” of words? - What is the “meta-meaning”? (speech act) - Handling a large chunk of text - Making sense of everything Syntax Semantics Pragmatics Morphology Discourse Inference

An Example of NLP A dog is chasing a boy on the playground DetNounAuxVerbDetNounPrepDetNoun Noun Phrase Complex Verb Noun Phrase Prep Phrase Verb Phrase Sentence Dog(d1). Boy(b1). Playground(p1). Chasing(d1,b1,p1). Semantic analysis Lexical analysis (part-of-speech tagging) Syntactic analysis (Parsing) A person saying this may be reminding another person to get the dog back… Pragmatic analysis (speech act) Scared(x) if Chasing(_,x,_). + Scared(b1) Inference

What we can do in NLP A dog is chasing a boy on the playground DetNounAuxVerbDetNounPrepDetNoun Noun Phrase Complex Verb Noun Phrase Prep Phrase Verb Phrase Sentence Semantics: some aspects - Entity/relation extraction - Word sense disambiguation - Anaphora resolution POS Tagging: 97% Parsing: partial >90%(?) Speech act analysis: ??? Inference: ???

What We Can’t Do in NLP 100% POS tagging –“He turned off the highway.” vs “He turned off the fan.” General complete parsing –“A man saw a boy with a telescope.” Deep semantic analysis –Will we ever be able to precisely define the meaning of “own” in “John owns a restaurant.”? Robust & general NLP tends to be “shallow” … “Deep” understanding doesn’t scale up …

Component Technology 2: Search (ad hoc retrieval)

What is Search (Ad hoc IR)? Retrieval System User “robotics applications” query Robotics others relevant docs non-relevant docs database/collection text docs

What we can do in Search Search in a pure text collection is well studied –Many different methods –Equally effective when optimized Basic search techniques (e.g., vector space, prob. models) are good enough for commercialization –All implementing TF-IDF style heuristics –Some new models have more potential for further optimization

What we can’t do in Search Basic retrieval models –No single model is the best on all test collections –Automatic parameter optimization Lack of interactive search support Lack of personalization Search context modeling Retrieval with more than pure text –With structures –Multi-media

Component Technology 3: Information Filtering

What is Information Filtering? Stable & long term interest, dynamic info source System must make a delivery decision immediately as a document “ arrives ” Filtering System … my interest:

State of the Art: Filtering Content-based adaptive filtering –Basic techniques, though not perfect, are there –We haven’t seen many (any?) filtering applications Collaborative filtering (recommender systems) –Simple methods can be (are being) commercialized –Real applications exist –More applications are possible

Component Technology 4: Text Categorization

What is Text Categorization? Pre-given categories and labeled document examples (Categories may form hierarchy) Classify new documents A standard supervised learning problem Categorization System … Sports Business Education Science … Sports Business Education

State of the Art: Categorization Many supervised learning methods have been developed –SVM is often the best in performance –Other methods are also competitive –Commercial applications exist, but not at a large-scale –More applications can be developed Feature selection/extraction is often more important than the choice of the learning algorithm Applications have been developed Relatively well explored

Component Technology 5: Clustering

The Clustering Problem Discover “ natural structure ” Group similar objects together Object can be document, term, passages Example

State of the Art: Clustering Many methods have been developed, applicable in different situations Difficult to predict which method is the best When patterns are clear, most methods work well In difficult situations –Special clustering bias must be incorporated –Properties of clustering methods need to be considered

End of State of the Art Tour…

Where is IR Going? IR and related areas Current trends How would this course fit to the picture?

Related Areas Information Retrieval Databases Library & Info Science Machine Learning Pattern Recognition Data Mining Natural Language Processing Applications Web, Bioinformatics… Statistics Optimization Software engineering Computer systems Models Algorithms Applications Systems

Current Trends Information Retrieval Databases Library & Info Science Machine Learning Pattern Recognition Data Mining Natural Language Processing Applications Web, Bioinformatics… Statistics Optimization Software engineering Computer systems Models Algorithms Applications Systems Web/ Bioinformatics/… Literature/Digital Library Structured + Unstructured Data Human-Computer Interactions High-Performance Computing More Powerful Content Analysis More Principled Models/Algorithms

Publications/Societies ACM SIGIR VLDB, PODS, ICDE ASIS Learning/Mining NLP Applications Statistics ?? Software/systems ?? COLING, EMNLP, ANLP HLT ICML, NIPS, UAI RECOMB, PSB JCDL Info. Science Info Retrieval ACM CIKM, TREC Databases ACM SIGMOD ACL ICML AAAI ACM SIGKDD ISMB WWW

Let Users Lead the Way… The underlying driving force has always been real world applications The ultimate impact of research in IR is to benefit people in accessing and using information in the real world Research on many component technologies is reaching a stage of “diminishing return”; the challenge is how to make use of such imperfect techniques Think more about complete solutions (as opposed to component technologies) as well as new applications

How would this Course Fit to the Picture? Identify novel application problems Identify new research topics Examine existing research work in these directions Design and carry out new projects in some of the directions We will broadly look at 3 application domains: Web, , and Literature