Chapter 19: Information Retrieval

Chapter 19: Information Retrieval

اسلاید 1: 1Chapter 19: Information Retrieval

اسلاید 2: Chapter 19: Information RetrievalRelevance Ranking Using TermsRelevance Using HyperlinksSynonyms., Homonyms, and OntologiesIndexing of DocumentsMeasuring Retrieval EffectivenessWeb Search EnginesInformation Retrieval and Structured DataDirectories

اسلاید 3: Information Retrieval SystemsInformation retrieval (IR) systems use a simpler data model than database systemsInformation organized as a collection of documentsDocuments are unstructured, no schemaInformation retrieval locates relevant documents, on the basis of user input such as keywords or example documentse.g., find documents containing the words “database systems”Can be used even on textual descriptions provided with non-textual data such as imagesWeb search engines are the most familiar example of IR systems

اسلاید 4: Information Retrieval Systems (Cont.)Differences from database systemsIR systems don’t deal with transactional updates (including concurrency control and recovery)Database systems deal with structured data, with schemas that define the data organizationIR systems deal with some querying issues not generally addressed by database systemsApproximate searching by keywordsRanking of retrieved answers by estimated degree of relevance

اسلاید 5: Keyword SearchIn full text retrieval, all the words in each document are considered to be keywords. We use the word term to refer to the words in a documentInformation-retrieval systems typically allow query expressions formed using keywords and the logical connectives and, or, and notAnds are implicit, even if not explicitly specifiedRanking of documents on the basis of estimated relevance to a query is criticalRelevance ranking is based on factors such asTerm frequencyFrequency of occurrence of query keyword in documentInverse document frequencyHow many documents the query keyword occurs in Fewer  give more importance to keywordHyperlinks to documentsMore links to a document  document is more important

اسلاید 6: Relevance Ranking Using TermsTF-IDF (Term frequency/Inverse Document frequency) ranking:Let n(d) = number of terms in the document dn(d, t) = number of occurrences of term t in the document d.Relevance of a document d to a term t The log factor is to avoid excessive weight to frequent termsRelevance of document to query Qn(d)n(d, t)1 +TF (d, t) = logr (d, Q) =TF (d, t)n(t)tQ

اسلاید 7: Relevance Ranking Using Terms (Cont.)Most systems add to the above modelWords that occur in title, author list, section headings, etc. are given greater importanceWords whose first occurrence is late in the document are given lower importanceVery common words such as “a”, “an”, “the”, “it” etc are eliminatedCalled stop wordsProximity: if keywords in query occur close together in the document, the document has higher importance than if they occur far apartDocuments are returned in decreasing order of relevance scoreUsually only top few documents are returned, not all

اسلاید 8: Similarity Based RetrievalSimilarity based retrieval - retrieve documents similar to a given documentSimilarity may be defined on the basis of common wordsE.g. find k terms in A with highest TF (d, t ) / n (t ) and use these terms to find relevance of other documents.Relevance feedback: Similarity can be used to refine answer set to keyword queryUser selects a few relevant documents from those retrieved by keyword query, and system finds other documents similar to theseVector space model: define an n-dimensional space, where n is the number of words in the document set.Vector for document d goes from origin to a point whose i th coordinate is TF (d,t ) / n (t )The cosine of the angle between the vectors of two documents is used as a measure of their similarity.

اسلاید 9: Relevance Using HyperlinksNumber of documents relevant to a query can be enormous if only term frequencies are taken into accountUsing term frequencies makes “spamming” easyE.g. a travel agency can add many occurrences of the words “travel” to its page to make its rank very highMost of the time people are looking for pages from popular sitesIdea: use popularity of Web site (e.g. how many people visit it) to rank site pages that match given keywordsProblem: hard to find actual popularity of siteSolution: next slide

اسلاید 10: Relevance Using Hyperlinks (Cont.)Solution: use number of hyperlinks to a site as a measure of the popularity or prestige of the siteCount only one hyperlink from each site (why? - see previous slide)Popularity measure is for site, not for individual pageBut, most hyperlinks are to root of siteAlso, concept of “site” difficult to define since a URL prefix like cs.yale.edu contains many unrelated pages of varying popularityRefinementsWhen computing prestige based on links to a site, give more weight to links from sites that themselves have higher prestigeDefinition is circularSet up and solve system of simultaneous linear equationsAbove idea is basis of the Google PageRank ranking mechanism

اسلاید 11: Relevance Using Hyperlinks (Cont.)Connections to social networking theories that ranked prestige of peopleE.g. the president of the U.S.A has a high prestige since many people know himSomeone known by multiple prestigious people has high prestigeHub and authority based rankingA hub is a page that stores links to many pages (on a topic)An authority is a page that contains actual information on a topicEach page gets a hub prestige based on prestige of authorities that it points toEach page gets an authority prestige based on prestige of hubs that point to it Again, prestige definitions are cyclic, and can be got by solving linear equationsUse authority prestige when ranking answers to a query

اسلاید 12: Synonyms and HomonymsSynonymsE.g. document: “motorcycle repair”, query: “motorcycle maintenance”need to realize that “maintenance” and “repair” are synonymsSystem can extend query as “motorcycle and (repair or maintenance)”HomonymsE.g. “object” has different meanings as noun/verbCan disambiguate meanings (to some extent) from the contextExtending queries automatically using synonyms can be problematicNeed to understand intended meaning in order to infer synonymsOr verify synonyms with userSynonyms may have other meanings as well

اسلاید 13: Concept-Based QueryingApproachFor each word, determine the concept it represents from contextUse one or more ontologies:Hierarchical structure showing relationship between conceptsE.g.: the ISA relationship that we saw in the E-R modelThis approach can be used to standardize terminology in a specific fieldOntologies can link multiple languagesFoundation of the Semantic Web (not covered here)

اسلاید 14: Indexing of DocumentsAn inverted index maps each keyword Ki to a set of documents Si that contain the keywordDocuments identified by identifiersInverted index may record Keyword locations within document to allow proximity based rankingCounts of number of occurrences of keyword to compute TFand operation: Finds documents that contain all of K1, K2, ..., Kn.Intersection S1 S2 .....  Snor operation: documents that contain at least one of K1, K2, …, Knunion, S1 S2 .....  Sn,.Each Si is kept sorted to allow efficient intersection/union by merging “not” can also be efficiently implemented by merging of sorted lists

اسلاید 15: Measuring Retrieval EffectivenessInformation-retrieval systems save space by using index structures that support only approximate retrieval. May result in:false negative (false drop) - some relevant documents may not be retrieved.false positive - some irrelevant documents may be retrieved.For many applications a good index should not permit any false drops, but may permit a few false positives.Relevant performance metrics:precision - what percentage of the retrieved documents are relevant to the query.recall - what percentage of the documents relevant to the query were retrieved.

اسلاید 16: Measuring Retrieval Effectiveness (Cont.)Recall vs. precision tradeoff:Can increase recall by retrieving many documents (down to a low level of relevance ranking), but many irrelevant documents would be fetched, reducing precisionMeasures of retrieval effectiveness: Recall as a function of number of documents fetched, orPrecision as a function of recall Equivalently, as a function of number of documents fetchedE.g. “precision of 75% at recall of 50%, and 60% at a recall of 75%”Problem: which documents are actually relevant, and which are not

اسلاید 17: Web Search EnginesWeb crawlers are programs that locate and gather information on the WebRecursively follow hyperlinks present in known documents, to find other documentsStarting from a seed set of documentsFetched documentsHanded over to an indexing systemCan be discarded after indexing, or store as a cached copyCrawling the entire Web would take a very large amount of timeSearch engines typically cover only a part of the Web, not all of itTake months to perform a single crawl

اسلاید 18: Web Crawling (Cont.)Crawling is done by multiple processes on multiple machines, running in parallelSet of links to be crawled stored in a databaseNew links found in crawled pages added to this set, to be crawled laterIndexing process also runs on multiple machinesCreates a new copy of index instead of modifying old indexOld index is used to answer queriesAfter a crawl is “completed” new index becomes “old” indexMultiple machines used to answer queriesIndices may be kept in memoryQueries may be routed to different machines for load balancing

اسلاید 19: Information Retrieval and Structured DataInformation retrieval systems originally treated documents as a collection of wordsInformation extraction systems infer structure from documents, e.g.:Extraction of house attributes (size, address, number of bedrooms, etc.) from a text advertisementExtraction of topic and people named from a new articleRelations or XML structures used to store extracted dataSystem seeks connections among data to answer queriesQuestion answering systems

اسلاید 20: DirectoriesStoring related documents together in a library facilitates browsingusers can see not only requested document but also related ones.Browsing is facilitated by classification system that organizes logically related documents together.Organization is hierarchical: classification hierarchy

اسلاید 21: A Classification Hierarchy For A Library System

اسلاید 22: Classification DAGDocuments can reside in multiple places in a hierarchy in an information retrieval system, since physical location is not important.Classification hierarchy is thus Directed Acyclic Graph (DAG)

اسلاید 23: A Classification DAG For A Library Information Retrieval System

اسلاید 24: Web DirectoriesA Web directory is just a classification directory on Web pagesE.g. Yahoo! Directory, Open Directory projectIssues:What should the directory hierarchy be?Given a document, which nodes of the directory are categories relevant to the documentOften done manuallyClassification of documents into a hierarchy may be done based on term similarity

اسلاید 25: 25End of Chapter