Basic crawler operation. Crawler (AKA Spider) (AKA Robot) (AKA Bot) Basic Crawler. Downloading Web Pages. Crawling picture

Transcription

1 Basic crawler operation Sec Crawler (AKA Spider) (AKA Robot) (AKA Bot) Begin with known seed s and parse them Extract s they point to Place the extracted s on a queue each on the queue and repeat Some of these slides are based on Stanford IR Course slides at Basic Crawler Downloading Web Pages Frontier Will this terminate? remove( ) getpage( ) Suppose our robot wants to download: getlinksfrompage( ) 3 4 HTTP Request Web Server HTTP Response Webdata/index.html File System Got it! Now, lets parse it and continue on Actually more complicated; Server replies with redirection 5 Web s crawled and parsed Seed pages Crawling picture s frontier Unseen Web Sec

2 Structure of the Web Simple picture complications Sec What does this mean for seed choice? 7 Web crawling isn t feasible with one machine All of the above steps distributed Malicious pages Spam pages Spider traps dynamically generated Even non-malicious pages pose challenges Latency/bandwidth to remote servers vary Webmasters stipulations How deep should you crawl a site s hierarchy? Site mirrors and duplicate pages Politeness don t hit a server too often 8 What any crawler must do Be Polite: Respect implicit and explicit politeness considerations Only crawl allowed pages Sec What any crawler should do Be capable of distributed operation: designed to run on multiple distributed machines Sec Respect.txt (more on this shortly) Be Robust: Be immune to spider traps and other malicious behavior from web servers Be scalable: designed to increase the crawl rate by adding more machines Performance/efficiency: permit full use of available processing and network resources 9 10 What any crawler should do Sec Updated crawling picture Sec pages of higher quality first Continuous operation: Continue fetching fresh copies of a previously fetched page Extensible: Adapt to new data formats, protocols 11 s crawled and parsed Seed Pages Crawling thread frontier Unseen Web 12 2

3 Processing steps in crawling Basic crawl architecture Pick a from the frontier the document at the the Extract links from it to other docs (s) Check if has content already seen If not, add to indexes For each extracted Ensure it passes certain tests Which one? E.g., only crawl.edu, obey.txt, etc. Check if it is already in the frontier (duplicate ination) 13 Frontier 14 Basic crawl architecture frontier Sec Can include multiple pages from the same host Must avoid trying to fetch them all at the same time Must try to keep all crawling threads busy Frontier Basic crawl architecture (Domain Name Server) Sec A lookup service on the internet Given a, retrieve its IP address Service provided by a distributed of servers thus, lookup latencies can be high (even seconds) Common OS implementations of lookup are blocking: only one outstanding request at a time Solutions caching Frontier Batch resolver collects requests and sends them 17 out together 18 3

4 Basic crawl architecture lication is widespread on the web If the page just fetched is already in the index, do not further process it This is verified using document fingerprints or shingles Second part of this lecture Frontier Basic crawl architecture Explicit and implicit politeness Sec Explicit politeness: specifications from webmasters on what portions of site can be crawled.txt Implicit politeness: even with no specification, avoid hitting any site too often Frontier Robots.txt Robots Exclusion Protocol Protocol for giving spiders ( ) limited access to a website, originally from Website announces its request on what can(not) be crawled For a server, create a file /.txt This file specifies access restrictions Put the file.txt at the root directory of the server Crawlers are expected to follow these instructions

5 .txt Format A.txt file consists of several records Each record consists of a of some crawler id s and a of s these crawlers are not allowed to visit User-agent lines: Names of crawlers Disallowed lines: Which s are not to be visited by these crawlers (agents)? User-agent: * Disallow: / User-agent: * Disallow: User-agent: Google Disallow: User-agent: * Disallow: / Examples User-agent: * Disallow: /cgi-bin/ Disallow: /tmp/ Disallow: /junk/ User-agent: BadBot Disallow: / Robots Meta Tag Robots Meta Tag A Web-page author can also publish directions for crawlers These are expressed by the meta tag with name, inside the HTML file Format: <meta name="" content="options"/> Options: index or noindex: index or do not index this file follow or nofollow: follow or do not follow the links of this file 27 <html> <head> <meta name="" content="noindex,follow"> <title>...</title> </head> <body> An Example: How should a crawler act when it visits this page? 28 Revisit Meta Tag Stronger Restrictions Web page authors may want Web applications to have an up-to-date copy of their page Using the revisit meta tag, page authors can give crawlers some idea of how often the page is being updated For example: <meta name="revisit-after" content="10 days" /> 29 It is possible for a (non-polite) crawler to ignore the restrictions imposed by.txt and meta directions Therefore, if one wants to ensure that automatic do not visit her resources, she has to use other mechanisms For example, password protections 30 5

6 Interesting Questions Basic crawl architecture Can you be sued for copyright infringement if you don t follow the.txt protocol? Can you be sued for copyright infringement if you follow the.txt protocol? Can the site owner determine who is actually crawling their site? Can the site owner protect against unauthorized access by specific? Frontier Parsing: normalization licate ination When a fetched document is parsed, some of the extracted links are relative s E.g., has a relative link to /wiki/wikipedia:general_disclaimer which is the same as the absolute aimer For a non-continuous (one-shot) crawl, test to see if an extracted+ed has already been passed to the frontier For a continuous crawl see details of frontier implementation During parsing, must normalize (expand) such relative s Distributing the crawler Communication between nodes Run multiple crawl threads, under different processes potentially at different nodes Output of the at each node is sent to the Eliminator of the appropriate node Geographically distributed nodes Partition hosts being crawled into nodes To other nodes Hash used for partition How do these nodes communicate and share s? 35 Frontier Host splitter From other nodes 36 6

7 frontier: two main considerations Politeness challenges Politeness: do not hit a web server too frequently Freshness: crawl some pages more often than others E.g., pages (such as News sites) whose content changes often These goals may conflict each other. (E.g., simple priority queue fails many links out of a page go to its own site, creating a burst of accesses to that site.) 37 Even if we restrict only one thread to fetch from a host, can hit it repeatedly Common heuristic: insert time gap between successive requests to a host that is >> time for most recent fetch from that host 38 frontier: Mercator scheme Mercator frontier s Prioritizer K front queues Biased front queue selector Back queue router B back queues Single host on each Back queue selector s flow in from the top into the frontier Front queues manage prioritization Back queues enforce politeness Each queue is FIFO Crawl thread requesting Front queues Front queues Prioritizer 1 K Prioritizer assigns to an integer priority between 1 and K Appends to corresponding queue Biased front queue selector Back queue router 41 Heuristics for assigning priority Refresh rate sampled from previous crawls Application-specific (e.g., crawl news sites more often ) 42 7

8 Biased front queue selector Back queues When a back queue requests a (in a sequence to be described): picks a front Biased front queue selector Back queue router queue from which to pull a This choice can be round robin biased to 1 B queues of higher priority, or some more sophisticated variant Can be randomized 43 Back queue selector Heap 44 Back queue invariants Back queue heap Each back queue is kept non-empty while the crawl is in progress Each back queue only contains s from a single host Maintain a table from hosts to back queues One entry for each back queue The entry is the earliest time at which the host corresponding to the back queue can be hit again This earliest time is determined from Last access to that host Any time buffer heuristic we choose Back queue processing A crawler thread seeking a to crawl: Extracts the root of the heap es at head of corresponding back queue q (look up from table) Checks if queue q is now empty if so, pulls a v from front queues If there s already a back queue for v s host, append v to q and pull another from front queues, repeat Else add v to q When q is non-empty, create heap entry for it 47 Number of back queues B Keep all threads busy while respecting politeness Mercator recommendation: three times as many back queues as crawler threads 48 8

9 licate documents Near licate ument Detection The web is full of duplicated content Strict duplicate detection = exact match Not as common But many, many cases of near duplicates E.g., Last modified date the only difference between two copies of a page 49 licate/near-licate Detection lication: Exact match can be detected with fingerprints (= Hash of content) Near-lication: Approximate match Compute syntactic similarity with an editdistance measure Use similarity threshold to detect near-duplicates E.g., Similarity > 80% => uments are near duplicates Too Expensive!! Not transitive though sometimes used transitively Detecting approximate duplicates Create a sketch for each page that is much smaller than the page Extract a of features for each page Compare pages by comparing s of features In the following, assume: we have converted page into a sequence of tokens Eliminate punctuation, HTML markup, etc 52 Shingling Given document D, a w-shingle is a contiguous subsequence of w tokens The w-shingling S(D,w) of D, is the of all w- shingles in D Example: D=(a rose is a rose is a rose) a rose is a rose is a rose is a rose is a rose is a rose is a rose Resemblance The resemblance of docs A and B is defined S(D,4) = {(a,rose,is,a),(rose,is,a,rose), (is,a,rose,is)} as: In general, 0 r w (A,B) 1 Note r w (A,A) = 1 But r w (A,B)=1 does not mean A and B are identical! (What is a good value for w?) Remember this? 9

10 Sketches Comparing uments Set of all shingles is large Bigger than the original document We cannot compute resemblance! We create a document sketch by sampling only a few shingles Requirement Sketch resemblance should be a good estimate of document resemblance Notation: From now on, w will be a fixed value, so we omit w from r w and S(A,w) r(a,b)=r w (A,B) and S(A) = S(A,w) 55 Basic structure of document comparison process Note that sketches of documents that have already been seen are stored by crawler Shingle A Jaccard Shingle B 56 Choosing a sketch Random sampling does not work! Suppose we have identical documents A, B each with n shingles Let m A be a shingle from A, chosen uniformly at random; similarly m B For k=1: E[{m A } {m B } ] = 1/n But r(a,b) = 1 So the sketch overlap is an underestimate Choosing a sketch Assume the of all possible shingles is totally ordered Define: m A is the smallest shingle in A m B is the smallest shingle in B r (A,B) = {m A } {m B } For identical documents A & B r (A,B) = 1 = r(a,b) Problems With a sketch of size 1, there is still a high probability of error Especially if we choose minimum shingle, e.g., in alphabetical order Sketch of a document Create a sketch vector (of size ~200) for each document uments that share t (say 80%) corresponding vector elements are deemed near duplicates For doc D, sketchd[ i ] is as follows: Size of each shingle is still large e.g., if w = 7, about bytes 200 shingles = 8K-10K 59 Let f map all shingles in the universe to 0.. (e.g., f = fingerprinting) Let i be a random permutation on 0.. sketchd[i] = min { (f(s)) s shingle in D} 10

11 Computing Sketch[i] for 1 ument 1 Test if 1.Sketch[i] = 2.Sketch[i] ument 1 ument 2 Repeat 200 times! Compute shingles of 1 Take fingerprints f(shingles) Permute on the number line with i Pick the min value A B Are these equal? Test for 200 random permutations: 1, 2, 200 However ument 1 ument 2 A A = B iff the shingle with the MIN value in the union of 1 and 2 is common to both (i.e., lies in the intersection) Why? Claim: This happens with probability Size_of_intersection / Size_of_union B Set Similarity of s Ci, Cj View s as columns of a matrix A; one row for each element in the universe. aij = 1 indicates presence of item i in j Example Ci C j Jaccard(C i,c j) C C C 1 C 2 i Jaccard(C 1,C 2 ) = 2/5 = j Key Observation Min Hashing For columns Ci, Cj, four types of rows Ci Cj A 1 1 B 1 0 C 0 1 D 0 0 Overload notation: A = # of rows of type A Claim A Jaccard(C i,c j) A B C Randomly permute rows Hash h(ci) = index of first row with 1 in column Ci Surprising Property Why? P h(c ) h(c ) Jaccard C, C i Both are A/(A+B+C) Look down columns Ci, Cj until first non-type-d row h(ci) = h(cj) iff type A row j i j 11

12 Similar uments Remember: We test with 200 random permutations So, probability that D1 and D2 are declared similar is Jaccard(S(D1),S(D2)) 200 Computing Random Permutations Compute 200 different fingerprint(s) for each shingle For example, use MD5 hashing function over shingle%i, where 1<=i<=200 For each fingerprint function, choose lexicographically first shingle in each document Near duplication likelihood is computed by checking how many of the 200 smallest shingles are common to both documents Solution (cont) Observe that this gives us: An efficient way to compute random orders of shingles Significant reduction in size of shingles stored (40 bits is usually enough to keep estimates reasonably accurate) 69 12