Information Retrieval und Text Mining

Wintersemester 2012/13
Wiltrud Kessler, Thomas Müller, Max Kisselew, Hinrich Schütze
Tu 14:00-15:30, Pfaffenwaldring 5b, V 5.01 & V 5.02
Fr 14:00-15:30, Pfaffenwaldring 5b, V 5.01 & V 5.02

Schedule and Resources

Please sign up for this course in ILIAS for the assignments. You will need these style files to compile the latex sources. Please note that some of the links to slides still refer to last year's slides. They will be updated over time.

	Day	Topic	*	Chapter	Slides	Resources
IIR01	Tu 10/16	Boolean retrieval (WK)	V	pdf html	students instructors source	information retrieval links search Shakespeare
IIR02	Fr 10/19	Term vocabulary & postings lists (WK)	V	pdf html	students instructors source	Porter stemmer
IIR13	Tu 10/23	Text classification, Naive Bayes (HS)	V	pdf html	students instructors source	Weka (includes Naive Bayes) Reuters-21578 vulgarity text classifier fail
	Fr 10/26	Practical exercise	U			Assignment 1 Supporting files for assignment 1 Solutions for assignment 1 (Exercises 1, 3, 4) Solution for assignment 1 - exercise 2 Slides from the practical exercise 1
IIR03	Tu 10/30	Dictionaries & tolerant retrieval (WK)	V	pdf html	students instructors source	trie vs hash vs ternary tree wild card search on Google edit distance demo P. Norvig's spell corrector spelling correction gone wrong freq(misspelling)>freq(correct) soundex demo
IIR12	Fr 11/2	Language models for IR (HS)	V	pdf html	students instructors source	Ponte & Croft paper on LMs in IR Zhai & Lafferty Lemur Toolkit
IIR05	Tu 11/6	Index compression (WK)	V	pdf html	students instructors source	variable byte codes word-aligned binary codes pos/freq compression
	Fr 11/9	Practical exercise	U			Assignment 2 Solutions for assignment 2
IIR06	Tu 11/13	Scores, weights, vector spaces (WK)	V	pdf html	students instructors source	vector space for dummies exploring the similarity space Okapi BM25 Lilian Lee on pivoted document length normalization
IIR09	Fr 11/16	Rel. feedback, query expansion (HS)	V	pdf html	students instructors source	original relevance feedback paper relevance feedback at Excite Justin Bieber: related searches fail WordSpace automatic word sense discrimination
	Tu 11/20	Practical exercise	U			Assignment 3 Solutions for assignment 3 Slides from the practical exercise 3
IIR07	Fr 11/23	Computing scores (WK)	V	pdf html	students instructors source	how Google tweaks ranking interview with Google's Udi Manber Amit Singhal on Google ranking SEO perspective: ranking factors Yahoo BOSS: opening up search compare Google/Yahoo rankings eye tracking at Google
IIR14	Tu 11/27	Vector space classification (HS)	V	pdf html	students instructors source	perceptron example TC overview by Sebastiani FSNLP (decision trees, perceptrons) The elements of statistical learning
IIR08	Fr 11/30	Evaluation & result summaries (WK)	V	pdf html	students instructors source	TREC at NIST v. Rijsbergen's definition of F A/B testing too much A/B testing? early paper on dynamic summaries search quality evaluation at Google
IIR15-1	Tu 12/4	Support vector machines (HS)	V	pdf html	students instructors source	Explanation for distance
	Fr 12/7	Practical exercise	U			Assignment 4 Corpus for exercise 2 Solution for assignment 4 (exercise 1) Solution for assignment 4 (exercise 2) Slides from the practical exercise 4
IIR16	Tu 12/11	Flat clustering (HS)	V	pdf html	students instructors source	van Rijsbergen: Cluster Hypothesis search result clustering: Yippy search result clustering: Carrot2 search result clustering: Bing # clusterings: Stirling number
IIR18	Fr 12/14	Latent semantic indexing (HS)	V	pdf html	students instructors source	Original LSI paper Probabilistic LSI Dimensions of meaning: LSI for words
	Tu 12/18	Practical exercise	U			Assignment 5 Solutions for assignment 5
IIR19	Tu 1/8	Web information retrieval (CS)	V	pdf html	students instructors source	how ads are priced most expensive keywords Geico search ca. 2004 geo-targeted ad size of the web in 2007 size of the web in 2008 ad monitoring at Google fighting webspam
IIR20	Fr 1/11	Crawling (FL)	V	pdf html	students instructors source	Mercator web crawler robots.txt standard Google data centers
	Tu 1/15	Practical exercise	U			Assignment 6 Solutions for assignment 6
IIR21	Fr 1/18	Link analysis (CS)	V	pdf html	students instructors source	more on PageRank math Jon Kleinberg (inventor of HITS) Google bomb (January 2008) defused Google bomb (June 2009)
	Tu 1/22	Semantic Search (WK)	V		students instructors source	CleverSearch Yummly SWSE Ask The Wiki Evi PizzaFinder Semantic Media Wiki
	Fr 1/25	Practical exercise	U			Assignment 7 Solutions for assignment 7
	Tu 1/29	Probeklausur	U			Review questions Review exercises Exam topics
	Fr 2/1	Besprechung Probeklausur, Fragen	U
	Fr 2/8	Klausur	U

* V = Vorlesung / normal class; U = Übung

Exam-related Information

To get admitted to the exam ("Vorleistung"), you will have to

get at least 50 % of the points you can get over all assignment sheets,
vote at least 50 % of all exercises in the course, and
present at least one exercise in class.

For all students (except M.Sc. CL who take the course as part of the concentration StatNLP) there will be a written exam on Fr 2/8/13, your final grade for the course will be the grade you get in the exam.

M.Sc. CL students who take the course as part of the concentration StatNLP do not have to do the written exam, this course will be examined as part of the oral exam for the concentration.

More Resources

Textbook (IIR): Introduction to Information Retrieval. Christopher D. Manning, Prabhakar Raghavan and Hinrich Schütze. Cambridge University Press, 2008. Web publication at http://informationretrieval.org

Old assignments and slides:

	Topic	Chapter	Slides	Resources
IIR04	Index construction	pdf html	students instructors source	SPIMI paper Google data center tour
IIR10	XML retrieval	pdf html	students instructors source
IIR11	Probabilistic information retrieval	pdf html	students instructors source
IIR15-2	Learning to rank (LTR)	pdf html	students instructors source	Microsoft LTR datasets
IIR17	Hierarchical clustering	pdf html	students instructors source	GoogleNews precursor: Newsblaster Bisecting K-means PDDP algorithm

You can find more information on the pages of previous courses: