Introduction to Information Retrieval Introduction to Information Retrieval Adapted from Christopher Manning and Prabhakar Raghavan The term vocabulary,

Slides:



Advertisements
Similar presentations
ITCS 6265 Information Retrieval and Web Mining Lecture 2: The term vocabulary and postings lists.
Advertisements

Introduction to Information Retrieval Introduction to Information Retrieval Document ingestion.
PrasadL4DictonaryAndQP1 Dictionary and Postings; Query Processing Adapted from Lectures by Prabhakar Raghavan (Yahoo and Stanford) and Christopher Manning.
1 Web Search and Text Mining Lecture 2 Adapted from Manning, Raghaven and Schuetze.
Introduction to Information Retrieval (Manning, Raghavan, Schutze) Chapter 2 The term vocabulary and postings lists.
Information Retrieval and Text Mining Lecture 2. Recap of the previous lecture Basic inverted indexes: Structure: Dictionary and Postings Key step in.
Introduction to Information Retrieval Introduction to Information Retrieval Christopher Manning and Prabhakar Raghavan Lecture 2: The term vocabulary and.
Term Vocabulary and Postings Lists 1 Lecture 3: The term vocabulary and postings lists Web Search and Mining.
CES 514 Data Mining Feb 17, 2010 Lecture 2: The term vocabulary and postings lists.
CSE 538 MRS BOOK – CHAPTER II The term vocabulary and postings lists
Introduction to Information Retrieval Introduction to Information Retrieval Modified from Stanford CS276 slides Chap. 2: The term vocabulary and postings.
COMP4201 Information Retrieval and Search Engines Lecture 2: The term vocabulary and postings lists.
CS276: Information Retrieval and Web Search
IR Paolo Ferragina Dipartimento di Informatica Università di Pisa Reading Chapter 1 Many slides are revisited from Stanford’s lectures by P.R.
Information Retrieval and Data Mining (AT71. 07) Comp. Sc. and Inf
Inverted Index Construction Tokenizer Token stream. Friends RomansCountrymen Linguistic modules Modified tokens. friend romancountryman Indexer Inverted.
Preparing the Text for Indexing 1. Collecting the Documents Tokenizer Token stream. Friends RomansCountrymen Linguistic modules Modified tokens. friend.
Information Retrieval and Web Search Lecture 2: Dictionary and Postings.
The term vocabulary and postings lists
Information Retrieval and Web Search Lecture 2: The term vocabulary and postings lists Minqi Zhou 1.
CS276A Information Retrieval
Vocabulary size and term distribution: tokenization, text normalization and stemming Lecture 2.
Information Retrieval Lecture 2. Recap of the previous lecture Basic inverted indexes: Structure: Dictionary and Postings Key step in construction: Sorting.
IR Paolo Ferragina Dipartimento di Informatica Università di Pisa Reading Chapter 1 Many slides are revisited from Stanford’s lectures by P.R.
CS276 Information Retrieval and Web Search Lecture 2: The term vocabulary and postings lists.
Information Retrieval Document Parsing. Basic indexing pipeline Tokenizer Token stream. Friends RomansCountrymen Linguistic modules Modified tokens. friend.
Text Processing & Characteristics
PrasadL4DictonaryAndQP1 Dictionary and Postings; Query Processing Adapted from Lectures by Prabhakar Raghavan (Yahoo and Stanford) and Christopher Manning.
CSCI 5417 Information Retrieval Systems Jim Martin Lecture 2 8/25/2011.
Information Retrieval and Web Search Text properties (Note: some of the slides in this set have been adapted from the course taught by Prof. James Allan.
Introduction to Information Retrieval Introduction to Information Retrieval Lecture 2: The term vocabulary and postings lists Related to Chapter 2:
Information Retrieval Lecture 2: The term vocabulary and postings lists.
Text Pre-processing and Faster Query Processing David Kauchak cs458 Fall 2012 adapted from:
1 Documents  Last lecture: Simple Boolean retrieval system  Our assumptions were:  We know what a document is.  We can “machine-read” each document.
Information Retrieval and Web Search Lecture 2: Dictionary and Postings.
Dan Jurafsky Text Normalization Every NLP task needs to do text normalization: 1.Segmenting/tokenizing words in running text 2.Normalizing word formats.
Information Retrieval Lecture 2: Dictionary and Postings.
Introduction to Information Retrieval Introduction to Information Retrieval CS276: Information Retrieval and Web Search Christopher Manning and Prabhakar.
Information Retrieval and Web Search Text processing Instructor: Rada Mihalcea (Note: Some of the slides in this slide set were adapted from an IR course.
Introduction to Information Retrieval Introduction to Information Retrieval Term vocabulary and postings lists – preprocessing steps 1.
Text Processing & Characteristics Thanks to: Baldi, Frasconi, Smyth M. Hearst W. Arms R. Krovetz C. Manning, P. Raghavan, H. Schutze.
Document Parsing Paolo Ferragina Dipartimento di Informatica Università di Pisa.
CS276 Information Retrieval and Web Search Lecture 2: Dictionary and Postings.
Basic Text Processing Word tokenization.
Information Retrieval and Web Search Lecture 2: The term vocabulary and postings lists Minqi Zhou 1.
Introduction to Information Retrieval Information Retrieval and Data Mining (AT71.07) Comp. Sc. and Inf. Mgmt. Asian Institute of Technology Instructor:
Text Processing & Characteristics
Information Retrieval in Practice
CSE 538 MRS BOOK – CHAPTER II The term vocabulary and postings lists
Information Retrieval
Lecture 2: The term vocabulary and postings lists
Ch 2 Term Vocabulary & Postings List
7CCSMWAL Algorithmic Issues in the WWW
Prof. Paolo Ferragina, Algoritmi per "Information Retrieval"
Lecture 2: The term vocabulary and postings lists
Modified from Stanford CS276 slides
Boolean Retrieval Term Vocabulary and Posting Lists Web Search Basics
Text Processing.
Document ingestion.
Token generation - stemming
CS276: Information Retrieval and Web Search
CS276: Information Retrieval and Web Search
Recap of the previous lecture
PageRank GROUP 4.
Lecture 2: The term vocabulary and postings lists
Lecture 2: The term vocabulary and postings lists
Prof. Paolo Ferragina, Algoritmi per "Information Retrieval"
INFORMATION RETRIEVAL TECHNIQUES BY DR. ADNAN ABID
Basic Text Processing Word tokenization.
INFORMATION RETRIEVAL TECHNIQUES BY DR. ADNAN ABID
Presentation transcript:

Introduction to Information Retrieval Introduction to Information Retrieval Adapted from Christopher Manning and Prabhakar Raghavan The term vocabulary, text statistics

Introduction to Information Retrieval The basic indexing pipeline Tokenizer Token stream. Friends RomansCountrymen Linguistic modules Modified tokens. friend romancountryman Indexer Inverted index. friend roman countryman Documents to be indexed. Friends, Romans, countrymen.

Introduction to Information Retrieval Parsing a document  What format is it in?  pdf/word/excel/html?  What language is it in?  What character set is in use? Each of these is a classification problem, which we will study later in the course. But these tasks are often done heuristically … Sec. 2.1

Introduction to Information Retrieval Tokenization  Input: “Friends, Romans and Countrymen”  Output: Tokens  Friends  Romans  Countrymen  A token is an instance of a sequence of characters  Each such token is now a candidate for an index entry, after further processing  But what are valid tokens to emit? Sec

Introduction to Information Retrieval Tokenization: terms  Issues in tokenization:  Finland’s capital  Finland? Finlands? Finland’s?  Hewlett-Packard  Hewlett and Packard as two tokens?  state-of-the-art: break up hyphenated sequence.  co-education  lowercase, lower-case, lower case ?  It can be effective to get the user to put in possible hyphens  San Francisco: one token or two? Sec

Introduction to Information Retrieval Tokenization: numbers  3/20/91 Mar. 12, /3/91  55 B.C.  B-52  My PGP key is 324a3df234cb23e  (800)  Often have embedded spaces  Older IR systems may not index numbers  But often are very useful: looking up error codes or models on the web Sec

Introduction to Information Retrieval Tokenization: language issues  Chinese and Japanese have no spaces between words:  莎拉波娃现在居住在美国东南部的佛罗里达。  Not always guaranteed a unique tokenization  Further complicated in Japanese, with multiple alphabets intermingled フォーチュン 500 社は情報不足のため時間あた $500K( 約 6,000 万円 ) KatakanaHiraganaKanjiRomaji End-user can express query entirely in hiragana! Sec

Introduction to Information Retrieval Stop words  With a stop list, you exclude from the dictionary entirely the commonest words. Intuition:  They have little semantic content: the, a, and, to, be  There are a lot of them: ~30% of postings for top 30 words  But the trend is away from doing this:  Good compression techniques (lecture!!) means the space for including stopwords in a system is very small  Good query optimization techniques (lecture!!) mean you pay little at query time for including stop words.  You need them for phrase queries or titles Sec

Introduction to Information Retrieval Normalization to terms  We need to “normalize” words in indexed text as well as query words into the same form  We want to match U.S.A. and USA  We most commonly implicitly define equivalence classes of terms by, e.g.,  deleting periods to form a term  U.S.A., USA  USA  deleting hyphens to form a term  anti-discriminatory, antidiscriminatory  antidiscriminatory Sec

Introduction to Information Retrieval Normalization: other languages  Accents: e.g., French résumé vs. resume.  Umlauts: e.g., German: Tuebingen vs. Tübingen  Should be equivalent  Even in languages that commonly have accents, users often may not type them  Often best to normalize to a de-accented term Sec

Introduction to Information Retrieval Normalization: other languages  Tokenization and normalization may depend on the language and so is intertwined with language detection  Crucial: Need to “normalize” indexed text as well as query terms into the same form Morgen will ich in MIT … Is this German “mit”? Sec

Introduction to Information Retrieval Normalization to terms  An alternative to equivalence classing is to do asymmetric expansion  An example of where this may be useful  Enter: windowSearch: window, windows  Enter: windowsSearch: Windows, windows, window  Enter: WindowsSearch: Windows  Potentially more powerful, but less efficient Sec

Introduction to Information Retrieval Case folding  Reduce all letters to lower case  exception: upper case in mid-sentence?  e.g., General Motors  Fed vs. fed  SAIL vs. sail  Often best to lower case everything, since users will use lowercase regardless of ‘correct’ capitalization… Sec

Introduction to Information Retrieval Thesauri and soundex  Do we handle synonyms and homonyms?  E.g., by hand-constructed equivalence classes  car = automobile color = colour  We can rewrite to form equivalence-class terms  When the document contains automobile, index it under car- automobile (and vice-versa)  Or we can expand a query  When the query contains automobile, look under car as well  What about spelling mistakes?  One approach is soundex, which forms equivalence classes of words based on phonetic heuristics

Introduction to Information Retrieval Lemmatization  Reduce inflectional/variant forms to base form  E.g.,  am, are, is  be  car, cars, car's, cars'  car  Lemmatization implies doing “proper” reduction to dictionary headword form Sec

Introduction to Information Retrieval Stemming  Reduce terms to their “roots” before indexing  “Stemming” suggest crude affix chopping  language dependent  e.g., automate(s), automatic, automation all reduced to automat. for example compressed and compression are both accepted as equivalent to compress. for exampl compress and compress ar both accept as equival to compress Sec

Introduction to Information Retrieval Porter’s algorithm  Commonest algorithm for stemming English  Results suggest it’s at least as good as other stemming options  Conventions + 5 phases of reductions  phases applied sequentially  each phase consists of a set of commands  sample convention: Of the rules in a compound command, select the one that applies to the longest suffix. Sec

Introduction to Information Retrieval Typical rules in Porter  sses  ss  ies  i  ational  ate  tional  tion  Weight of word sensitive rules  (m>1) EMENT →  replacement → replac  cement → cement Sec

Introduction to Information Retrieval Other stemmers  Other stemmers exist, e.g., Lovins stemmer   Single-pass, longest suffix removal (about 250 rules)  Full morphological analysis – at most modest benefits for retrieval  Do stemming and other normalizations help?  English: very mixed results.  Definitely useful for Spanish, German, Finnish, …  30% performance gains for Finnish! Sec

Introduction to Information Retrieval Language-specificity  Many of the above features embody transformations that are  Language-specific and  Often, application-specific  These are “plug-in” addenda to the indexing process  Both open source and commercial plug-ins are available for handling these Sec

Introduction to Information Retrieval Dictionary entries – first cut ensemble.french 時間. japanese MIT.english mit.german guaranteed.english entries.english sometimes.english tokenization.english These may be grouped by language (or not…). More on this in ranking/query processing. Sec. 2.2

Introduction to Information Retrieval

Dmoz Directory

Introduction to Information Retrieval Yahoo! Directory

Introduction to Information Retrieval STATISTICAL PROPERTIES OF TEXTS

Introduction to Information Retrieval Statistical properties of texts  Tokens are not distributed uniformly. They follow the so called “Zipf Law”  Few tokens are very frequent  A middle sized set has medium frequency  Many are rare  The first 100 tokens sum up to 50% of the text, and many of them are stopwords

Introduction to Information Retrieval An example of “Zipf curve”

Introduction to Information Retrieval A log-log plot for a Zipf’s curve

Introduction to Information Retrieval  k-th most frequent term has frequency approximately 1/k; or the product of the frequency (f) of a token and its rank (r) is almost a constant  Scale-invariant: f(br) = b  s * f(r) The Zipf Law, in detail f = c |T| / r s  s r * f = c |T| f = c |T| / r General Law

Introduction to Information Retrieval Distribution vs Cumulative distr Sum after the k-th element is ≤ f k k/(s-1) Sum up to the k-th element is ≥ f k k Power-law with smaller exponent

Introduction to Information Retrieval Consequences of Zipf Law  Stop words are frequent but do not discriminate  Do exist many tokens that are NOT frequent and do not discriminate (error?).  English: Calpurnia  Italian: Precipitevolissimevolmente (o, paklo) Words with medium frequency Words that discriminate

Introduction to Information Retrieval Other statistical properties of texts  The number of distinct tokens grows as  The so called “Heaps Law” (|T|  where  <1, tipically 0.5 )  The average token length grows as  (log |T|)  Interesting words are the ones with medium frequency (Luhn)