From a Web to a linear HTML document


Martin Gagnon
gagnon@vlsi.polymtl.ca
Michel Dagenais
dagenais@vlsi.polymtl.ca
Department of Electrical and Computer Engineering
Ecole Polytechnique of Montreal
PO Box 6079, Succ Centre-ville
Montreal, Quebec, Canada
H3C 3A7

Abstract

The lingua franca of the World Wide Web is the HyperText Markup Language (HTML) which presents information as a Web of small documents linked to each other. While this Web navigation is convenient when searching for a specific information, it may be cumbersome when all the information must be read anyway (like traditional linear texts). The proposed approach adds linearization hints to a Web of HTML documents and provides tools to automatically generate a linear version of a Web. Furthermore, a table of contents, an index and a list of references are automatically produced for the linear version.

Introduction

In the past few years, the number of Internet users and the amount of information available have increased significantly. The World Wide Web is probably the medium which has enjoyed the most important growth. It can be defined as a set of protocols (Hypertext Transfer Protocol) [1], formats (Hypertext Markup Language, HTML) [2], software tools and conventions.

smoke fetish archive Avril lavigne pussy genie in a string bikinie cast Girls fuck horses stories pornstars xxx babes Nikki schieler ziering nude dickies jeans Skinny teen bitches hentai sex videos free Little girls sex nude pot Boob tube High school musical vanessa nude pics free artistic nude photos pussy prowler Gayforit watch free lesbian videos free teen celebs nude Bikini dare galleries spiritual sport fucking Big and beautiful porn stars
faked nude Sleep nude girls nature nude video Catfight nude nude asian sluts nude ladies pussy Sara roemer nude asian girls non nude School girls nude pics gay nude workout Nude nudist pamela anderson playboy nude pics Kathrine heigel nude Free nude hentai fairies nude swedish blonde nude Nude celebrity videos for free nude celebrities miley cyrus nude south african men Free hardcore nude nude cellphone pictures Bollywood nude boobs
blowjobs gallery Jim hardick free porn videos no credit card Creatine sex milfporn star aluminium essex Free video sex positions ultrapasswords xxx Xxx teen britney spears blowjob video Facts teenagers curfews natural hairy pussy Amature women Edwin carungay fuckyourtube sexproadventures Free kinky sex tips rave sex porn lyrics sexy back Better than sex cake recipe final fantasy rikku xxx Paris hilton blowjob
free yno sex video 3d young art sex phone web cam sex Amature woman sex party free home-made sex clips young sex in america Free dirty sex pictures best sfrican sex movies He she sex pics picts of amature sex Julie michaels sex scene bible view on sex Sex tv tv show Extreme insertion sex 6 fee animal sex vids sex girls piss tube Thai pussy sex porn sex 3d fantasy pics sex mature woman jokes Jeremiah birthday sex bio tulsa police sex registration Sex vedeo stream chat
independent sex scenes Racist daughter sex clearanced sex toys K9 sex clips britney sex movies black sex squirt Awsome hard sex manson sex onstage Nimpho sex classifieds sex offenders index Nomid animal sex hardcore lezbo sex Oral sex possitions Out sex videos sly fox sex famos toon sex Only ebony sex anette dawn sex extent sex pill Mature hairy sex asian sex french Kim kardishan sex
education research group Ari banerjee yankee group ancestry group Randy orton group free group sex porno group insurances Galleon group hedge fund rubber fab technologies group Attorneys group group b infection Risk retention group insurance the rules support group Green resources group Group dynamics team r46b group high five amateurs group Amazing group sex on demand color group lesbian group gallery Campy centaur group accept group Group of deer is called
fucking machine xxx College sex xxx chobits xxx Iran xxx sexo xxx enanas collection xxx Eve angel xxx pork xxx Older women xxx download free psp xxx Xxx sluts videos swingers xxx free Free bi xxx Photos xxx free harecore xxx xxx porn passwords Rapes xxx xxx adult dvd xxx gratis con putas Web site xxx free xxx mangas Alena seredova xxx
ball dragon porn video The thrills music video woman squirting free video Roma video card e pci video mtv jam video Apartment mikes picture video paris hilton video stills Big cock homemade movie council meeting video Studio telescope video converter ipod ora video Victoria pink videos Uk movies cussler movie new video releases 2005 Conferencing live video violence video games children tasha nelson video Rv video camera movie graber Adam sandler secret video
teacher sex crazydumper Sex health video marriage with sex Celebrity sex viceos busty office sex shove bull sex Football sex rio free sex shots Consensual submission sex free sex gemes Mauritius sex site hardcore sex mp3 Barbarella sex machine Hunting sex jessica alien sex gaems free sex xxx Muscular sex pictures ass booty sex dogpound group sex Anail sex videos vitamins before sex Brewster sex stories
asians sex Haveing sex with a man lesbien sex xxx Hypno girl sex arabic sex 9356 biker girls sex Guilty gear sex mature free sex tube Nude girls having sex with boys ray j and kim kardashian full sex tape for free Cyber sex forum what is angry sex Sex while pregnant pictures When can i have sex and not get pregnant home made amateur sex tapes dog sex beastality Sex games online for women clips cartoon sex taboo charming mother sex Girl sex pose hardcore gothic sex Best sex teacher
love sex relatioships Historical books sex pegging sex literature Sex story community sex bites torrent long sex trailors Gonzo rawr sex carrie bradshaw sex Voung teen sex home sex stream Kinky sex forum savvanah gold sex Anal sex wide Crushing for sex comic sex jokes mermaid sex videos Pet sex foram ali sheffield sex cancer sex partners Calforina sex retreat mini teens sex Anal sex cum
victorian xxx Xxx sci fi sexy photos xxx Xxx video play xxx babe videos animail xxx All xxx tube tilf 2 xxx Xxx puzzle black porno xxx 3gp xxx wap videos streaming xxx Free xxx moves Muscle gay xxx free gothic xxx video naruto xxx Xxx pass free best xxx movie 2008 xxx dog clips Xxx free e cards xxx porn full videos Xxx stone
porn movie theaters Morgan lane porn catherine porn Porn mom son sex mommy and daddy porn kasumi porn Find porn torrents rumania porn Xxx pictures porn black porn videos free Discipline porn biggest penis porn Littel girl porn Porn leg warmers tiny tits porn movies top 10 porn clips Free lovemaking porn homemade mexican porn vanessa raia porn Muslim porn sex free high definition porn streaming James nichols gay porn
fuck me gay Vulva fuck sexy fuck movie Mother lets son fuck her fuck you mom and dad mommy fuck son Father son fuck girl porn to fuck Fuck off letter fuck my boob Megaupload fuck i fuck my mother inlaw Doggy style fuck videos Woman looking to fuck shemales fuck girls movies kama sutra fuck Fuck you love mother daughter fuck boyfriend fuck church Dog fuck woman movies the fuck buttons Man fuck his dog
Blowjob And Cum Swallow mom giving son blowjob Preggo Blowjob free blowjob compilations blowjob mature Blowjob Guys blowjob fantasies 18 Avatar Blowjob sister gave me a blowjob Tickling Blowjob blowjob at school Hentai Porn Blowjob Fake Blowjob girl pukes during blowjob blowjob tryouts Guys Blowjob japanese girl giving blowjob most famous blowjob Gay Horse Blowjob double blowjob vids Blowjob Outdoor
Youngest Girl Porn Ever plus size sexy school girl Flavor Flav Girl Poops all girl sex videos girl porche Baby Girl I Want You gossip girl on tv com Hey Hey Baby Will You Be My Girl naked girl shitting Little Girl Photos ghetto black girl Go Go Girl Adult Girl Psp Theme girl for sale on ebay pin up girl hats Little Monster Girl naked teen girl pics black girl actress Sleeping Girl Gets Raped how to approach a girl online Girl And Girl Haveing Sex
Ink bitch webbie gutta bitch Lyrics to five star bitch bitch in french Badd bitch quotes cant trust no bitch Bitch asian im a pretty bitch Kristen stewart is a bitch a bitch slap G unit fat bitch Shut up bitch download im in san diego bitch cock hungry bitch Teeh fuck the bitch is kristen stewart a bitch bitch milfs Lyrics to bitch by meredith brooks foot fetish bitch Shake that ass bitch and let
paris hilton beach sex Cocksucker snake girls xxx Nude booty poppin little teens pics most extreme porn list Audience analysis heather locklear nude Porn star named madison lolita preteens Cheyanne bride black cock joelle amateur Nude christina aguilera Nice nude teen photo gallery hot cab mature sex sites Fucked by my dog mpegs massive tits men fucking boys Swedish porn galleries amateur nudes Sexy superheroes
bbw nude women Nude pussy cum naomi nude Nude asian americans courtney smith nude sienna guillory nude Girls basketball nude kate bosworth nude fakes Amateur wife nude photos ukraine nude teen Big black ass nude kiera knightley nude pics Nude russians Sleep nude chris brown rihanna nude photos pic of nude girls Bollywood nude images sexy and nude pics free nude college girl videos Nude dads and daughters ameture nude pictures Serena williams nude pix
1st Anal Sex what is an anal prolapse Types Of Anal Sex gay anal sex technique gay anal fisting videos Why Does Anal Sex Feel Good video double anal Lesbian Teens Anal largest anal dildo Lesbian Anal Toy anal sex poop videos Anal Hidden Cam Amateur Interracial Anal amy amour anal how to anal intercourse Anal Sex Condoms eyaculacion anal free anal streaming Anne Hathaway Loves Anal mini anal Unnatural Anal Insertions
Anal Guest free full anal movies Manual Anal 1st anal video shits herself anal Couple Anal Sex roxy renolds anal Sara Jay First Anal Scene anal destruction casedy Como Hacer El Sexo Anal anal sex effects Anal Cancer Blog Anal Toys Lesbian ice la fox anal scene lesbian anal vid Rough Anal Sex Clips wet anal double anal sex movie Palin Anal really painful anal Shitty Anal Fuck
rodox sex mpg Shower sex how penis breasts sex Sex malam pertama random sex videos exsplicit sex videos Sex lubrication silicone i post sex Sex fat chick celebriies having sex Adult sex animations sex and motorcycles Adult sex therapy Laura cover sex fucking having sex sex vacation caribbean Pool sex orgasm women barbershop sex office sex gay Secretaire office sex black sex vod Rainbow mika sex
Rock cock jock cock robin when your Wife big cock huge cock free pics Mature sucking black cock cock docking clips Hardcore riding cock cock sucking whores Fuck you cock sucker cock fighting rules Big cock hardcore Hubby loans to black cock milf sucking young cock two cock in pussy Cock sucker t shirt two cock fucking cock pierced Tila tequila suck cock largest cock videos White teen black cock
miss teen usa south carolina Fucking boobs thumbnails free videos of gay black me gandbang Senior sex trailer sophie monk nude nude music videos Britney spears porn video maggie grace nude Preteen bikini movies xxx Sexy pamela anderson vanessa new nude photos Aisha tyler nude pics Gametophyte produces male female sex mate plants toothless blowjob monthly membership streaming porn Pinkpanteens preteens in thongs lingerie nudecollege students Fat mature sex teen monologues Ebony muff diving
sex with hookers Free jaybee sex sex with redheads Cartoons about sex usa sex forum retarted girls sex Photo booth sex gay virgin sex Female sex chromosome sex teen candy Teenage sex story sex feet tingle Celebrity sex sces Flex girl sex lesbian sex galerii work at sex Rough sex free roug gangbang sex hypnosis sex best Sex trek 6 teens wating sex Ssecretary sex videos
1st Anal Sex what is an anal prolapse Types Of Anal Sex gay anal sex technique gay anal fisting videos Why Does Anal Sex Feel Good video double anal Lesbian Teens Anal largest anal dildo Lesbian Anal Toy anal sex poop videos Anal Hidden Cam Amateur Interracial Anal amy amour anal how to anal intercourse Anal Sex Condoms eyaculacion anal free anal streaming Anne Hathaway Loves Anal mini anal Unnatural Anal Insertions
Hypertext documents are nodes of information linked together. This way, each separate topic resides in a separate file and may be used in different contexts. For instance, an HTML document on copying garbage collection may be pointed to from documents on algorithms, object oriented programming, or on the Modula-3 language.

Other document processing formats such as LaTeX [19] may be divided into small sections which are merged into a linear document with "input" commands. However, these systems have little support for reusing a file in several documents. Indeed, the pathnames must be relative to the document root, instead of to the current file in HTML, and header level must be adequate for the context of the including document. Other popular document formats such as GNU info [] and Linux-doc sgml present similar problems. Moreover, with the importance of the World Wide Web as the most up to date source of information, and the wide availability of tools such as Amaya [] for HTML authoring, it is preferable to use HTML as the source format than to convert from another format (e.g. LaTeX) to HTML.

The flexibility allowed by HTML to split documents into small reusable nodes is also a source of problems. Indeed, in many cases the information should be available both as an online Web and as a linear printable document. Examples include the Modula-3 language definition and libraries, the Java language definition, libraries and tutorial, and several documents from the World Wide Web consortium (W3C) and the Linux documentation project.


Objectives

The objective of the proposed approach is to enter all the information in HTML, possibly with some linearization hints. Then tools may be used to point at a HTML Web node and recursively follow the appropriate links in order to produce a linear HTML document. Converting to another format (ASCII, LaTeX, Postscript) is a somewhat separate and simpler issue.



Figure 1: Structure of a Web site

A Web site may be represented as a network of nodes where each node represents an HTML file (see Figure 1). However, A number of issues need to be examined for producing a linear document from a Web of HTML nodes.

These issues are discussed below.

Linearization Rules and Issues

An HTML document usually consists of:

<HTML>
<HEAD>
<TITLE> This is the title </TITLE>
</HEAD>
<BODY>
detailed text here
</BODY>
</HTML>

Where only the TITLE is mandatory. The BODY implicitly starts after the TITLE, if no HEAD or BODY section is tagged. The linearization of a Web starts at a node specified as root. From there, hypertext links are followed in a depth first search, with the following rules.

Construction of Derived Nodes

Once the linearization is performed, it is useful to construct derived nodes containing information about the linear document, or about the web of visited nodes. In the first case, the address within the linear document is used, while in the second case the original address within the web is used.

A table of contents lists all the headers. Nested lists are used to represent the sections hierarchy. Each list entry is a link containing the section title, and pointing to the relevant position in the linear document or the web of visited nodes. Thus, each header should have an ID attribute. Similarly, a list of figures or tables may be constructed; the FIGURE element was recently dropped from the newer HTML 3.2 proposal however.

A reference list is another important but slightly more complex derived node. It should contain a list of all references external to the linear document. However, these references are in URL form while most printed documents prefer traditional bibliographic entries. In some cases the URL may point to a bibliographic entry:

<DIV ID="Here" CLASS=BIBITEM>
Author,<I>book title</I>, O'Connell, 1995
</DIV>
In other cases, the URL may point to an ordinary document. However, a bibliographic entry may be deduced:
<TITLE> book title </TITLE>
<H1> book title </H1>
<P>
<DIV CLASS=AUTHOR> Author </DIV>
<P>
<DIV CLASS=BIBITEM.TAIL> O'Connell, 1995 </DIV>
...
If there is no such information to be found, the entry would be simply the URL.

Conversion to ASCII

To produce an ASCII document, the table of contents may be prepended to the document, and the reference list and index appended. The references in the table of contents should be replaced by the page number of the corresponding section. The entries in the reference list should have a tag assigned (order of first occurrence, e.g. [1] [2], or a symbolic name used in anchors, e.g. [Connolly1996]).
Then, external references in the document must be replaced by the tag associated with the corresponding entry in the reference list.
The internal references to figures, sections and tables should be replaced by the figure/section/table number. There may be cases where one wants the page number instead. This may be achieved with the attribute CLASS=PAGEREF in <A HREF=...>.

Producing LaTeX documents

LaTeX can automatically generate the table of contents and the index, and takes care of \label and \ref, or \pageref, for sections, tables and figures. Thus, the conversion between HTML and LaTeX is more straightforward. The major item where special care is required is the reference list. It must be generated by the linearization tool. However, LaTeX is able to connect each \cite to the corresponding \bibitem in the reference list.

Implementation

The rules defined earlier in this text, for the linearization of a web, have been implemented in a HTML processing framework named htmlkit. Unlike the algorithm of Sharples and Goodlet [3], which relies solely on heuristics to guide the librarization through the web, htmlkit uses heuristic rules and linearization hints annotations (REL attribute in links).

The framework is implemented as a set of reusable Modula-3 [4] libraries. The low level of coupling between the different modules in the libraries makes it easy to interchange or upgrade components. The libraries are decomposed as follows:

Libraries
Low-Level
  • HTTP, Gopher, FTP... communication
  • HTML parsing
  • File access
  • Input/Output streams
  • Low level data structures
  • Text utilities
  • Format matching
Intermediate level
  • Higher level data structures
  • Operations on the data structures
  • Link manipulation utilities
High level
  • Linearization tools
  • ASCII filter
  • Latex filter
  • Index builder
  • Table of contents builder
  • Reference list builder.
With this organization, it is relatively easy to build custom tools for performing various tasks for a Web site. For example one may want to construct an index or a table of contents of the Web site, without creating a linearization.

Results

The linearization tool implemented in htmlkit was tested on a Sun Sparc 5 workstation with 32 Mo of RAM. In each case, a web was linearized, an index, a table of contents and a list of references were built, and the output was converted to ASCII, LaTeX and Postscript.

Table 1 presents the elapsed time as a function of the number of nodes linearized. The processing time grows linearly with the number of nodes involved. In the current implementation, the modularity was favored over the performance. In any case, in typical applications, the network or disk access delay is likely to be the major limiting factor. Nevertheless, a large document may be linearized in approximately one minute.


Table 1: Linearization time versus number of nodes
Nbr. of
nodes
CPU time
(s)
User delay
(s)
Size
files
(Ko)
memory
(Ko)
3428.3131.8434.6872494
9435.1138.6253.6542536
12540.8145.8490.1322685
18652.7271.89101.8662699
21756.7576.33107.6302703
24861.1782.66128.8562709
30671.03110.84166.3972816


Figure 4: CPU time used versus nbr. of nodes in the network



Figure 5 presents the memory consumption as a function of the number of nodes linearized. The memory usage grows less than linearly with the number of nodes involved. The size of the VISITED NODES list and of the REFERENCES list grows linearly with the number of nodes; the entries in these lists are quite small though. The size of the memory devoted to storing complete nodes is proportional to the number of currently visited nodes, which is typically proportional to the logarithm of the number of nodes (it represents the depth of the tree formed by the links where the REL attribute is SUBDOCUMENT or INCLUDE). Again, the memory consumption for a large document is not a problem, at below 3MB.

Figure 5: Memory used versus nbr. of nodes in the network



Figure 6 presents the memory consumption, for a constant linear document size, as a function of the node size (which in this case is inversely proportional to the number of nodes). The effect of enlarging the granularity of nodes is to slightly reduce the tree depth, and thus the number of currently visited nodes, but mostly to enlarge the size of each of visited node. The size of the VISITED NODES and REFERENCES lists shrinks with the smaller number of nodes but their contribution to the total memory consumption is small. The net result is an almost linear increase of the memory consumption with the node size.

Figure 6: Memory used versus the size of bytes/nodes for a constant graph size (94Kb)



Conclusion

This paper presents a framework to linearize HTML documents. By adding a small number of hints to web nodes, as attributes in HTML tags, it is possible to derive automatically a high quality linear printable version from a web. This way, authors of technical documents such as reference manuals can benefit from using HTML (hypertext navigation, widely available browsers and authoring tools, content oriented structure, multi-platform support) while gaining the possibility of producing high quality books starting from different possible root nodes in their web. As this and similar approaches gain acceptance, the tags and attributes used to annotate links, indexing terms, bibliographic references... should get better standardized. Further work on the prototype will be aimed at providing a more thorough customization layer in the conversion between the linear HTML document and the printable LaTeX/Postscript output. This may be achieved by allowing the redefinition of the LaTeX command associated with each HTML construct. It may also be useful to allow optional LaTeX layout parameters to be annotated in HTML attributes, although this would overlap with the planned functionality for the style sheets.



References





Return to Top of Page
Return to Posters Index