ΕΠΛ660 Ανάκτηση Πληροφοριών και Μηχανές Αναζήτησης ης Introduction and Boolean Retrieval Slides by Manning, Raghavan, Schutze
Τα Διαδικαστικά Μεταπτυχιακό μάθημα Πληροφορικής Το μάθημα απευθύνεται επίσης σε: προπτυχιακούς φοιτητές (τελειόφοιτους) Πληροφορικής φοιτητές άλλων τμημάτων (στατιστικής, γλωσσολογίας) Ό,τι πληροφορία χρειαστείτε για το μάθημα (συμβόλαιο, πρόγραμμα μαθημάτων, παραπομπές βιβλιογραφίας, επικοινωνία και ανακοινώσεις) ) θατηνβρείστε στην Ιστοσελίδα: http://www.cs.ucy.ac.cy/courses/epl660 / /EPL660 Slides by Manning, Raghavan, Schutze 2
Slides by Manning, Raghavan, Schutze 3
Slides by Manning, Raghavan, Schutze 4
Διδασκαλία Μία 3 ωρη διάλεξη την εβδομάδα (Τρι) Εργαστήριο: Θα γίνονται παρουσιάσεις σχετικά με εργαλεία Ανάκτησης Πληροφοριών (Lucene, Hadoop), συζητήσεις γιαθέματατουμαθήματος(ασκήσεις, εργασίες κλπ). Slides by Manning, Raghavan, Schutze 5
Βιβλιογραφία Βιβλίο μαθήματος: An Introduction to Information Retrieval, Manning, Raghavan, Schütze, Cambridge University Press (υπάρχει στο Διαδίκτυο σε pdf και html μορφότυπο). Άρθρα από την επιστημονική βιβλιογραφία Μαγνητοσκοπημένες διαλέξεις, δημοσιευμένες στο Διαδίκτυο Για περισσότερες πληροφορίες, ανατρέξετε στην ιστοσελίδα Resources του ιστιακού τόπου του μαθήματος. Slides by Manning, Raghavan, Schutze 6
Αξιολόγηση 2 γραπτές εξετάσεις: Ενδιάμεση (25%) Τελική (35%) 3 σειρές ασκήσεων: (15%) Θεωρητικές (από το βιβλίο) Προγραμματιστικές Εργασία 6 μήνου (25%) Slides by Manning, Raghavan, Schutze 7
Search and Information Retrieval Search on the Web is a daily activity for many people throughout the world Search and communication are most popular uses of the computer Applications involving search are everywhere Thefield of computer science that is most involved with R&D for search is information retrieval (IR) Slides by Manning, Raghavan, Schutze
Το Google σήμερα On a worldwide basis, Google employed 19,835 fulltime employees as of December 31, 2009. Google s official i strategy: search, ads, and apps Slides by Manning, Raghavan, Schutze
Ανάκτηση Πληροφοριών (Information Retrieval): Το τυπικό πρόβλημα Δεδομένα Προβλήματος Μια συλλογή από έγγραφα με κείμενο φυσικής γλώσσας D={d1,,dn} Μια επερώτηση Q ενός χρήστη σε μορφή συμβολοσειράς (string) Ζητούμενο Ένα διατεταγμένο σύνολο από έγγραφα που είναι συναφή με την επερώτηση <d5d2d7d9> <d5,d2,d7,d9> Query String IR System Document Corpus Slides by Manning, Raghavan, Schutze d5 d2 d7 d9 Ranked Relevant Documents 10
Μηχανές Αναζήτησης Πολλές μηχανές αναζήτησης είναι Αρκετά αποτελεσματικές Αναγνωρίσιμες και γνωστές Εμπορικά επιτυχημένες (τουλάχιστον μερικές) Τι συμβαίνει όμως στο παρασκήνιο ; Πως δουλεύουν? Πως μπορούμε να κρίνουμε αν δουλεύουν καλά; Πως μπορούμε να τις κάνουμε πιο αποτελεσματικές; Πως μπορούμε να τις κάνουμε να λειτουργούν πιο γρήγορα; Υπάρχει τίποτα παραπάνω από αυτό που βλέπουμε στον Παγκόσμιο Ιστό; Slides by Manning, Raghavan, Schutze 11
Περιγραφή Μαθήματος Σκεπτικό Τα Συστήματα Ανάκτησης Πληροφοριών (Information Retrieval systems) επιτρέπουν την πρόσβαση σε μεγάλους όγκους πληροφοριών αποθηκευμένων με τη μορφή κειμένου, φωνής, video, ή σε σύνθετη μορφή όπως Ιστοσελίδες. Σκοπός των συστημάτων αυτών είναι η ανάκτηση μόνο εκείνων των εγγράφων που είναι συναφή με αυτό που αναζητεί ο χρήστης. Για να το επιτύχουν πρέπει να αντιμετωπίσουν την αβεβαιότητα ως προς το τι πραγματικά αναζητεί ο χρήστης και ποιο το θέμα ενός εγγράφου. Σκοπός του Μαθήματος Εισαγωγή στην περιοχή των συστημάτων ανάκτησης πληροφοριών και των Μηχανών Αναζήτησης. Εξέταση των θεωρητικών και πρακτικών ζητημάτων που σχετίζονται με τη σχεδίαση, υλοποίηση και αξιολόγηση τέτοιων συστημάτων. Slides by Manning, Raghavan, Schutze 12
Διάρθρωση Μαθήματος Μπούλειος Ανάκτηση ηπληροφοριών ρ Κωδικοποίηση κειμένου, λημματοποίηση, στελέχωση κειμένων Λεξικά και ανάκτηση ανεκτική σε σφάλματα Κατασκευή και συμπίεση ευρετηρίων Διαβάθμιση όρων Ανάκτηση διανυσματικού χώρου Αξιολόγηση ανάκτησης πληροφοριών Μηχανισμοί ανάδρασης και διαστολή επερωτήσεων Ταξινόμηση ηκειμένου και απλοϊκές τεχνικές Bayes Ταξινόμηση διανυσματικού χώρου Επίπεδη ομαδοποίηση Ιεραρχική ομαδοποίηση Βασικές έννοιες αναζήτησης στον Ιστό Ιχνηλασία και ευρετηριασμός Ιστού Ανάλυση υπερσυνδέσμων Slides by Manning, Raghavan, Schutze 13
Γιατί χρειαζόμαστε ΑΠ σήμερα; Πόσο εύχρηστος θα ήταν ο Ιστός χωρίς μηχανές αναζήτησης; Ο Ιστός περιέχει δισεκατομμύρια σελίδες (η μηχανή Google έχειευρετηριάσειπάνω από 1 τρισεκατομμύριο μοναδικά άurl URLs) 25 7 2008: http://googleblog.blogspot.com/2008/07/we knewweb was big.html Ο κόσμος παράγει περίπου 2 exabytes νέας πληροφορίας το χρόνο, 90% της οποίας είναι σε ψηφιακή μορφή και με 50% ετήσια αύξηση Slides by Manning, Raghavan, Schutze 14
Το πρόβλημα είναι σημαντικό και επίκαιρο Μέσα έ σε μόλις επτά χρόνια μια παγκόσμια αυτοκρατορία εξαπλώθηκε. Όχι δεν έχει στρατό και πλοία. Είναι μια εξουσία της γνώσης: η μεγαλύτερη μηχανή διύλισης για την ακρίβεια των πληροφοριών που κυκλοφορούν στο Διαδίκτυο. Είναι δωρεάν και προσφέρει απλόχερα τις αγαθοεργούς υπηρεσίες της εν είδει καθολικής και αποστολικής εκκλησίας της γνώσης. Και όπως κάθε παγκόσμια εκκλησία, έχει θησαυρίσει. Με δεδομένη την καχυποψία μας για κάθε αυτοκρατορική εξουσία και με τη φθονερή βεβαιότητα ότι ουδέν καλόν αμιγές κακού, ας δούμε ποια είναι και που το πάει η Google (http://www.tovima.gr/print_article.php?e=b&f=14670&m=h04&aa=1) Τάσος Καφαντάρης Εφημερίδα ΤΟ ΒΗΜΑ (22/1/2006) / Slides by Manning, Raghavan, Schutze 15
Το πρόβλημα είναι σημαντικό και επίκαιρο Ο πρόεδρος της Γαλλίας (Ζακ Σιράκ) σήμανε προσκλητήριο για μια ευρωπαϊκή μηχανή αναζήτησης που θα απέκρουε τον αγγλοσαξονικό πολιτισμικό ιμπεριαλισμό. Εξήγγειλε ως βασική προτεραιότητα ρ του για το 2006 το Project Quaero ( Ερευνώ στα λατινικά), την υλοποίηση δηλαδή μιας ευρωπαικής μηχανής αναζήτησης 30/8/2005: Βρισκόμαστε στο μέσον ενός παγκόσμιου ανταγωνισμού για τεχνολογική υπεροχή. Στη Γαλλία, στην Ευρώπη, διακυβεύεται η αυτοκυριαρχία μας. 1/1/2006: Σήμερα χαράσσεται η νέα γεωγραφία της γνώσης και των πολιτισμών. Αύριο εκείνο που δεν είναι ευρέσιμο στο Διαδίκτυο κινδυνεύει να είναι αθέατο από τον κόσμο. Project Quaero Συνεταίροι: Thomson, FranceTelecom, DeutscheTelekom, CNRS, RWTH(Aachen), INRIA, Bertelsmann,,... Project Search Computing http://www.search computing.net Search computing focuses onbuilding the answers to complex search queries like "Where can I attend an interesting conference in my field close to a sunny beach?" Slides by Manning, Raghavan, Schutze 16
Definitions: Information Retrieval Information retrieval (IR) deals with the representation, storage, organization of, and access to information items. [Baeza-Yates, Ribeiro-Nieto, 1999] The term information retrieval has different meanings E.g.: getting a credit card out of your wallet so that you can type in the card number Our meaning: Given the user query, the key goal of an IR system is to retrieve information which might be useful or relevant to the user. Slides by Manning, Raghavan, Schutze
Definitions (Really) Short history Old discipline: for approximately 4000 years, man has organized information for later retrieval and usage Ancient Romans and Greeks recorded information on papyrus rolls Some of them had tags attached, containing a short summary in order to save time T.O.C. first appeared in Greek scrolls from the second century BC Early representative of computerized in electronic document repositories search: Cornel SMART System (1960s) [CornelSmart] Slides by Manning, Raghavan, Schutze
Definitions: (Really) Short history In the past, IR systems were used only by expert librarians as reference retrieval systems in batch modalities Many libraries still use categorization hierarchies to classify their volumes Novel computers and the birth of WWW (1989) changed forever the concepts of storage, access and searching of documents collections Link analysis in IR (1998) Current history. Slides by Manning, Raghavan, Schutze
Definitions: Information Retrieval As an academic discipline, information retrieval might be defined as follows: Information retrieval (IR) is devoted to finding relevant documents, not finding simple match to patterns. Grossman - Frieder, 2004 Information retrieval (IR) is finding material (usually documents) of an unstructured nature (usually text) that satisfy an information need from within large collections (usually stored on computers). Slides by Manning, Raghavan, Schutze Manning et al., 2007
Some core concepts of IR Slides by Manning, Raghavan, Schutze 21
What is a Document? Examples: web pages, email, books, news stories, scholarly papers, text messages, Word, Powerpoint, PDF, forum postings, patents, IM sessions, etc. Common properties Significant text content Some structure (e.g., title, author, date for papers; subject, sender, destination for email) Slides by Manning, Raghavan, Schutze
Documents vs. Database Records Database records (or tuples in relational databases) are typically made up of well defined fields (or attributes) e.g., bank records with account numbers, balances, names, addresses, social security numbers, dates of birth, etc. Easy to compare fields with well defined semantics to queries in order to find matches Text is more difficult Slides by Manning, Raghavan, Schutze
Documents vs. Records Example bank database query Find records with balance > $50,000 in branches located in Amherst, MA. Matches easily found by comparison with field values of records Example search engine query bank scandals in western mass This text must be compared to the text of entire news stories Slides by Manning, Raghavan, Schutze
Comparing Text Comparing the query text to the document text and determining what is a good match is the core issue of information retrieval Exact matching of words is not enough Many different ways to write the same thing in a natural language like English e.g., does a news story containing tii the text t bank director in Amherst steals funds match the query? Somestories willbe better matchesthanothers than Slides by Manning, Raghavan, Schutze
Dimensions of IR IR is more than just text, and more than just web search although these are central People doing IR work with different media, different types of search applications, and different tasks Slides by Manning, Raghavan, Schutze
Other Media New applications increasingly involve new media e.g., video, photos, music, speech Like text, content is difficult to describe and compare text may be used to represent them (e.g. tags) g) IR approaches to search and evaluation are appropriate Slides by Manning, Raghavan, Schutze
Dimensions of IR Content Applications Tasks Text Web search Ad hoc search Images Vertical search Filtering Video Enterprise search Classification Scanned docs Desktop search Question answering Audio Forum search Music P2P search Literature search Slides by Manning, Raghavan, Schutze
IR Tasks Ad hoc search Find relevant documents for an arbitrary text query Filtering Identify relevant user profiles for a new document Classification Identify relevant labels for documents Question answering Give a specific answer to a question Slides by Manning, Raghavan, Schutze
Σχετικές Περιοχές Slides by Manning, Raghavan, Schutze 30
RDBMS vs. IR Relational Database Management Systems (RDBMS): Semantics of each object are well defined Complex query languages (e.g., SQL) Exact retrieval for what you ask Emphasis onefficiency Information Retrieval (IR): Semanticsof object are subjective, not well defined Usually simple query languages (e.g., natural language query) You should get what htyou want, even the query is bd bad Effectiveness is primary issue, although efficiency is important Slides by Manning, Raghavan, Schutze 31
DB and IR: Two Parallel Universes Database Systems Information Retrieval canonical application: data type: accounting libraries numbers, short strings foundation: Sigmod algebraic 2007 / logic based text For more details see Weikum s Keynote search paradigm: market l d Boolean retrieval (exact queries, result sets/bags) probabilistic / statistics based ranked retrieval (vague queries, result lists) Oracle, IBM DB2, Google, Yahoo!, MSN, leaders: MS SSQL Server, etc. Verity, Fast, etc. Slides by Manning, Raghavan, Schutze 32
Κατηγορίες Δεδομένων Η ΑΠ χρησιμοποιείται και για την ανεύρεση πληροφοριών μέσα σε συλλογές πληροφοριών που περιέχουν διαφορετικές κατηγορίες δεδομένων: Δομημένα δεδομένα (structured data) Αδόμητα δεδομένα (unstructured data) Ημιδομημένα δεδομένα (semistructured data) Slides by Manning, Raghavan, Schutze 33
Unstructured (text) vs. structured (database) data in 1996 Slides by Manning, Raghavan, Schutze 34
Unstructured (text) vs. structured (database) data in 2009 Slides by Manning, Raghavan, Schutze 35
Sec. 1.1 Unstructured data in 1680 Which plays of Shakespeare contain the words Brutus AND Caesar but NOT Calpurnia? One could grep all of Shakespeare s plays for Brutus and Caesar, then strip out lines containing Calpurnia? Why is that not the answer? Slow (for largecorpora) NOT Calpurnia is non trivial Other operations (e.g., find the word Romans near countrymen) not feasible Ranked retrieval (best documents to return) Later lectures Slides by Manning, Raghavan, Schutze 36
Sec. 1.1 Term document incidence Antony and Cleopatra Julius Caesar The Tempest Hamlet Othello Macbeth Antony 1 1 0 0 0 1 Brutus 1 1 0 1 0 0 Caesar 1 1 0 1 1 1 Calpurnia 0 1 0 0 0 0 Cleopatra 1 0 0 0 0 0 mercy 1 0 1 1 1 1 worser 1 0 1 1 1 0 Brutus AND Caesar BUT NOT Calpurnia 1 if play contains word, 0 otherwise Slides by Manning, Raghavan, Schutze
Sec. 1.1 Incidence vectors So we have a 0/1 vector for each term. To answer query: take the vectors for Brutus, Caesar and Calpurnia (complemented) bitwise AND. 110100 AND 110111 AND 101111 = 100100. Slides by Manning, Raghavan, Schutze 38
Sec. 1.1 Answers to query Antony and Cleopatra, Act III, Scene ii Agrippa [Aside to DOMITIUS ENOBARBUS]: Why, Enobarbus, When Antony found Julius Caesar dead, He cried almost to roaring; and he wept When at Philippi i he found Brutus slain. Hamlet, Act III, Scene ii Lord Polonius: I did enact Julius Caesar I was killed i' the Capitol; Brutus killed me. Slides by Manning, Raghavan, Schutze 39
Sec. 1.1 Basic assumptions of Information Retrieval Collection: Fixed set of documents Goal: Retrieve documents with information that is relevant to the user s information need and helps the user complete a task Slides by Manning, Raghavan, Schutze 40
The classic search model TASK Info Need Verbal form Query Misconception? Mistranslation? Misformulation? Gt Get rid of mice in a politically correct way Info about removing mice without killing them How do I trap mice alive? mouse trap SEARCH ENGINE Query Refinement Slides by Manning, Raghavan, Schutze Results Corpus
Sec. 1.1 How good are the retrieved docs? Precision : Fraction of retrieved docs that are relevant to user s information need Recall : Fraction of relevant docs in collection that are retrieved More precise definitions and measurements to follow oo in later ae lectures es Slides by Manning, Raghavan, Schutze 42
Sec. 1.1 Bigger collections Consider N = 1 million documents, each with about 1000 words. Avg 6 bytes/word including spaces/punctuation 6GB of data in the documents. Say there are M = 500K distinct terms among these. Slides by Manning, Raghavan, Schutze 43
Sec. 1.1 Can t build the matrix 500K x 1M matrix has half a trillion 0 s and 1 s. But it has no more than one billion 1 s. matrix is extremely sparse. What s a better representation? We only record the 1 positions. Why? Slides by Manning, Raghavan, Schutze 44
Sec. 1.2 Inverted index For each term t, we must store a list of all documents that contain t. Identify each by a docid, a document serial number Can we used fixed size arrays for this? Brutus 1 2 4 11 31 45 173174 Caesar 1 2 4 5 6 16 57 132 Calpurnia 2 31 54101 What happens if the word Caesar is added to document 14? Slides by Manning, Raghavan, Schutze 45
Sec. 1.2 Inverted index We need variable size postings lists On disk, a continuous run of postings is normal and best In memory, can use linked lists or variable length arrays Some tradeoffs in size/ease of insertion Posting Brutus 1 2 4 11 31 45 173174 Caesar 1 2 4 5 6 16 57 132 Calpurnia 2 31 54101 Dictionary Postings Sorted by docid (more later on why). Slides by Manning, Raghavan, Schutze 46
Sec. 1.2 Inverted index construction Documents to be indexed. Friends, Romans, countrymen. Tokenizer Token stream. Friends Romans Countrymen More on Linguistic h l modules Modified tokens. friend roman countryman these later. Inverted index. Slides by Manning, Raghavan, Schutze Indexer friend roman countryman 2 4 1 2 13 16
Sec. 1.2 Indexer steps: Token sequence Sequence of (Modified token, Document ID) pairs. Doc 1 Doc 2 I did enact Julius Caesar I was killed i' the Capitol; Brutus killed me. So let it be with Caesar. The noble Brutus hath told you Caesar was ambitious Slides by Manning, Raghavan, Schutze
Sec. 1.2 Indexer steps: Sort Sort by terms And then docid Core indexing step Slides by Manning, Raghavan, Schutze
Sec. 1.2 Indexer steps: Dictionary & Postings Multiple term entries in a single document are merged. Split into Dictionary and Postings Doc. frequency information is added. Why frequency? Will discuss later. Slides by Manning, Raghavan, Schutze
Sec. 1.2 Where do we pay in storage? Terms and counts Pointers Lists of docids Later in the course: How do we index efficiently? i How much storage do we need? Slides by Manning, Raghavan, Schutze 51
Sec. 1.3 The index we just built How do we process a query? Later what kinds of queries can we process? Today s focus Slides by Manning, Raghavan, Schutze 52
Sec. 1.3 Query processing: AND Consider processing the query: Brutus AND Caesar Locate Brutus in the Dictionary; Retrieve its postings. Locate Caesar in the Dictionary; Retrieve its postings. Merge the two postings: 2 4 8 16 32 64 1 2 3 5 8 13 21 128 34 Brutus Caesar Slides by Manning, Raghavan, Schutze 53
Sec. 1.3 The merge Walk through the two postings simultaneously, in time linear in the total number of postings entries 2 8 2 4 8 16 32 64 128 1 2 3 5 8 13 21 34 Brutus Caesar If the list lengths are x and y, the merge takes O(x+y) operations. Crucial: postings sorted by docid. Slides by Manning, Raghavan, Schutze 54
Intersecting two postings lists (a merge algorithm) Slides by Manning, Raghavan, Schutze 55
Sec. 1.3 Boolean queries: Exact match The Boolean retrieval model is being able to ask a query that is a Boolean expression: Boolean Queries are queries using AND, OR and NOT to join query terms Views each document as a set of words Is precise: document matches condition or not. Perhaps the simplest model to build an IR system on Primary commercial retrieval tool for 3 decades. Many search systems you still use are Boolean: Email, library catalog, Mac OS X Spotlight Slides by Manning, Raghavan, Schutze 56
Sec. 1.4 Example: WestLaw http://www.westlaw.com/ Largest commercial (paying subscribers) legal search service (started ted 1975; ranking added 1992) Tens of terabytes of data; 700,000000 users Majority of users still use boolean queries Example query: What is the statute of limitations in cases involving the federal tort t claims act? LIMIT! /3 STATUTE ACTION /S FEDERAL /2 TORT /3 CLAIM Slides by Manning, Raghavan, Schutze /3 = within 3 words, /S = in same sentence 57
Sec. 1.4 Example: WestLaw http://www.westlaw.com/ Another example query: Requirements for disabled people to be able to access a workplace disabl! /p access! /s work site work place (employment /3 place Note that SPACE is disjunction, not conjunction! Long, precise queries; proximity operators; incrementally developed; not like web search Many professional searchers still like Boolean search You know exactly what htyou are getting But that doesn t mean it actually works better. Slides by Manning, Raghavan, Schutze
Boolean queries: More general merges Sec. 1.3 Exercise: Adapt the merge for the queries: Brutus AND NOT Caesar Brutus OR NOT Caesar Can we still run through the merge in time O(x+y)? What can we achieve? Slides by Manning, Raghavan, Schutze 59
Sec. 1.3 Merging What about an arbitrary Boolean formula? (Brutus OR Caesar) AND NOT (Antony OR Cleopatra) Can we always merge in linear time? Linear in what? Can we do btt better? Slides by Manning, Raghavan, Schutze 60
Sec. 1.3 Query optimization What is the best order for query processing? Consider a query that is an AND of n terms. For each of the n terms, get its postings, then AND them together. Brutus 2 4 8 16 32 64 128 Caesar 1 2 3 5 8 16 21 34 Calpurnia 13 16 Query: Brutus AND Calpurnia AND Caesar Slides by Manning, Raghavan, Schutze 61
Sec. 1.3 Query optimization example Process in order of increasing freq: start with smallest set, then keep cutting further. This is why we kept document freq. in dictionary Brutus 2 4 8 16 32 64 128 Caesar 1 2 3 5 8 16 21 34 Calpurnia 13 16 Execute the query as (Calpurnia AND Brutus) AND Caesar. Slides by Manning, Raghavan, Schutze 62
Sec. 1.3 More general optimization e.g., (madding OR crowd) AND (ignoble OR strife) Get doc. freq. s for all terms. Estimate the size of each OR by the sum of its doc. freq. s (conservative). Process in increasing order of OR sizes. Slides by Manning, Raghavan, Schutze 63
Exercise Recommend a query processing order for (tangerine OR trees) AND (marmalade OR skies) AND (kaleidoscope OR eyes) Term Freq eyes 213312 kaleidoscope 87009 marmalade 107913 skies 271658 tangerine 46653 trees 316812 Slides by Manning, Raghavan, Schutze 64
Query processing exercises Exercise: If the query is friends AND romans AND (NOT countrymen), how could we use the freq of countrymen? Exercise: Extend the merge to an arbitrary Boolean query. Can we always guarantee execution in time linear in the total postings size? Hint: Begin with the case of a Boolean formula query: in this, each query term appears only once in the query. Slides by Manning, Raghavan, Schutze 65
Exercise Try the search feature at http://www.rhymezone.com/shakespeare/ Write down five search features you think it could do better Slides by Manning, Raghavan, Schutze 66
What s ahead in IR? Beyond term search What about phrases? Stanford University Proximity: Find Gates NEAR Microsoft. Need index to capture position information in docs. Zones in documents: Find documents with (author = Ullman) AND (text contains automata). Slides by Manning, Raghavan, Schutze 67
Evidence accumulation 1 vs. 0 occurrence of a search term 2 vs. 1 occurrence 3 vs. 2 occurrences, etc. Usually more seems better Need term frequency information in docs Slides by Manning, Raghavan, Schutze 68
Ranking search results Boolean queries give inclusion or exclusion of docs. Often we want to rank/group results Need to measure proximity from query to each doc. Need to decide whether docs presented to user are singletons, or a group of docs covering various aspects of the query. Slides by Manning, Raghavan, Schutze 69
IR vs. databases: Structured vs unstructured data Structured data tends to refer to information in tables Employee Manager Salary Smith Jones 50000 Chang Smith 60000 Ivy Smith 50000 Typically allows numerical range and exact match (for text) t) queries, es, e.g., Salary < 60000 AND Manager = Smith. Slides by Manning, Raghavan, Schutze 70
Unstructured data Typically refers to free text Allows Keyword queries including operators More sophisticated concept queries e.g., find all web pages dealing with drug abuse Classic model for searching text documents Slides by Manning, Raghavan, Schutze 71
Semi structured data In fact almost no data is unstructured E.g., this slide has distinctly identified zones such as the Title and Bullets Facilitates semi structured structured search such as Title contains data AND Bullets contain search to say nothing of linguistic structure Slides by Manning, Raghavan, Schutze 72
More sophisticated semi structured search Title is about Object Oriented Programming AND Author something like stro*rup where * is the wild card operator Issues: how do you process about? how do you rank results? The focus of XML search (IIR chapter 10) Slides by Manning, Raghavan, Schutze 73
Clustering, classification and ranking Clustering: Given a set of docs, group them into clusters based on their contents. Classification: Given a set of topics, plus a new doc D, decide which topic(s) D belongs to. Ranking: Can we learn how to best order a set of documents, e.g., a set of search results Slides by Manning, Raghavan, Schutze 74
The web and its challenges Unusual and diverse documents Unusual and diverse users, queries, information needs Beyond terms, exploit ideas from social networks link analysis, clickstreams... How do search engines work? And how can we make them better? Slides by Manning, Raghavan, Schutze 75
More sophisticated information retrieval Cross language information retrieval Question answering Summarization Text mining Slides by Manning, Raghavan, Schutze 76
Resources for today s lecture Introduction to Information Retrieval, chapter 1 Shakespeare: http://www.rhymezone.com/shakespeare/ Try the neat browse by keyword sequence feature! Slides by Manning, Raghavan, Schutze 77