The Collection of
Computer Science Bibliographies
Home About FAQ Browse Add Statistics

Help on the simple search interface for the Computer Science Bibliography Collection

You can search the bibliography collection for references or for bibliographies. The search mechanism is based on glimpse.

The bibliographies are hierarchically structured and a search at any point in the hierarchy will only search the contents of the bibliographies contained in that subtree. Thus, if you are searching at the very top you search the complete collection, whereas if you descend to the bottom of the hierarchy you only search a single bibliography.

Query Syntax

You are searching all the information contained in the references, that is, including author names, titles, journal and conference names, publication dates, keywords, abstracts, ...
Unfortunately there is no way to restrict searches to only one field with this search interface. If you need this capability then try the advanced search interface.


You just enter the phrases you are looking for except for the special characters: '(', ')'. You cannot use parentheses in search phrases.

A search phrase can actually consist of several words with spaces inbetween. The phrase must appear verbatim in any matching reference, even a simple line break in the reference will prevent a match. Therefore phrases should be kept short (max. 2 or 3 words).

Boolean Operators

You can combine search phrases in an arbitrary fashion with the logical AND operator and and/or the OR operator or. By putting matching parentheses ( and ) around expressions you can define the precedences of the boolean operators.

Search Examples

search for "migration"

H{\"o}pfner or Hoepfner
If you are looking for names containing umlauts, accents or digraphs, you must use the corresponding LaTeX expressions and you should also try the common ascii transliteration.

object migration and Lazowska
search for references to publications by "Lazowska" containing "object migration": the two words must appear exactly like that in a matching reference, a line break inbetween the two words would prevent a match.

object and (migrating or migration) and Lazowska
search for references to publications by "Lazowska" containing both "object" and "migration" anywhere in the reference.

(object or process) and migration and 1994
search for references that contain "object" or "process", and "migration" in the year 1994.

embedding and ((mesh and hypercube) or (tree and butterfl))
Search for embeddings of meshes in hypercubes or for embeddings of trees in butterflies. Use with partial word matching to catch plural forms.

@techreport and (CM5 or CM-5 or {CM}-5)
Search for technical report references to the Connection Machine 5. Note the use of braces to account for the BibTeX idiosyncrasy of requiring that proper names containing uppercase letters be surrounded by braces in title fields.

parallel and url
Search for online publications on parallel processing.

You can search for authors using the initals. Just append their initials to their name in capital letters. The search should be made case-sensitive and partial word matching must be set. The above query would match both John Smith and James Thomas Smith.

NEW: You can now enter author names simply as "J. P. Smith" or "Smith, J. P." (without the doublequotes!) and these terms will be automatically converted to queries conforming with the above specification.

Important: Set the exact/partial word matching option carefully:
If you search for "SmithJ" using the exact option, then you will not find the matches where the author is listed as "J. P. Smith".
However, if you use the partial word option then a search for "SmithJ" might matches all authors with "J." as the first inital (e.g. "J. P. Smith", "J. A. Smith", "J.-H. Smith",...),

Searching for bibliographies

When you search for bibliographies you are searching the descriptions of the individual bibliographies in the collection and obtain pointers to bibliographies as results. This search facility may be used instead of hierachically browsing the collection and actually amounts to searching the HTML files in the collection. Generally, queries should be kept very simple, without boolean operators, and search terms should be kept general, describing areas rather than specific topics.

The Search Results

The results page that will be returned to you after a search indicates the search terms of your query in easily readable form. This will show you how your query expression has been interpreted by the query parser.

Then follow the references matching your query, either in BibTeX format or in "citation" format more easily readable for humans.

Any sequence of matches of entries from the same bibliography will be introduced by the name of the bibliography that can be used a direct link to the title page for that bibliography.

The BibTeX entries are rendered in HTML with the matching words highlighted and any URLs in the entry made into a clickable link. That allows you to immediately access that paper if a reference contains a URL to the online version of the paper.

If you save the page to a file in plain format (not HTML) you will obtain perfectly valid BibTeX references that can directly be used with LaTeX or merged with your own bibliographies.

Rendering of results:

The matching entries will be reduced to the bare bibliographic necessities and rendered like an entry in a literature list. Note, that wou will not see any possible abstracts and any other information beyond what is necessary to locate a paper version of the publication.

The advantage of this format is that the list of matching entries is shorter and therefore somewhat easier to browse..

Please note that the links from the citations to the corresponding entrie in BibTeX format are not persistent, that means that at some time later the links might point to BibTeX entries that do not correspond to the references any more. So please, do not save these links (recognizable by the sequences of many strange characters) in another HTML file since they will be invalidated soon.

This format displays the full BibTeX entry, including all the available information that might be missing in the citation format (abstracts, keywords, annotations). The BibTeX entry can be easily transferred into your personal BibTeX bibliography using cut&paste.
Count Only
Do not display any references, only count the number of matches in each bibliography.
Some references also contain forward or backward crossreferences to papers citing the paper or cited by the paper. These crossreferences are rendered as live searches that display the crossreferenced publication.

On the bottom of the page you will find how many matches have been found by the search engine and over how many bibliographies the results spread.

Compression of results

If you choose to obtain compressed result data then the data will be compressed with gzip. The HTTP response header will specify the Content-Encoding x-gzip.

You should configure your browser correctly such that it can uncompress the data transparently on-the-fly. These valuable tips might help you with the configuration.

If you browser cannot handle compressed data transparently you can simply save the results to a file with the ending .html.gz and use gzip manually to decompress the file. Then you can browse the resulting .html file with your browser.

Why compression?

Compression of results helps deliver the results to you quickly and terminates the memory-consuming processes on the server earlier, thus increasing throughput by reduced paging.

How much exactly are the savings? Well, the average BibTeX entry in this collection consumes 550 bytes. The average compression ratio obtained with gzip is 5:1. Therefore if you retrieve 200 matches, you transfer only 22,000 bytes with compression instead of 110,000 bytes uncompressed!

The Database

The search database consists of all the bibliographies in the bibliography collection in slightly modified form (see below). The complete database including indices consumes 400 MBytes of disk space.

The search database differs from the bibliographies in the collection in the following respects:

The database contains only about 75% of the references in the bibliography collection, the other 25% are automatically detected duplicate entries. Therefore you will not get as many redundant references as results of your search.

Many bibliographies in the collection define text macros (@String abbreviations) and use these macros in the fields of references. Those abbreviations have been expanded for the purpose of searching, such that the words in those macros are also indexed and can be searched for. Furthermore, the references returned as search results contain the complete information: there is no need to find the definitions of any macros occurring in a reference.

Some bibliographies in the collection use the BibTeX feature of crossreferences and thereby omit some information in the references. These crossreferences have been expanded for the purpose of searching which improves the search a lot.


Length of query:
If the query contains any non-alphanumerical characters then the query must not be longer than 32 characters.

Server Load:
The number of simultaneous searches is limited. If your search request is not serviced then try again later. Complicated queries might lead to long search times and eventually might cause a timeout for the search. Try to keep your queries neat and simple, that is, use only alphanumerical characters in your search words and search for complete words.

Search Options

Case-sensitive/case insensitive
Here you can choose to make the search case sensitive, e.g. ocr will not match OCR and Fisher will not match fisher. This is only recommended for acronyms or proper names.

partial word(s)/exact
"exact" means that any search phrase must occur in the references as a full phrase, that is, both its ends must fall on word boundaries. By choosing "partial word(s)" you can also have the search phrases match only as partial phrases. "partial word(s)" should e.g. be chosen to include plural forms of search terms in the search, but should be avoided if any short words appear in the query that might frequently appear as syllables in other, unrelated words (for instance, "att" might appear in "Seattle", "attention", "attenuation", ...)

Note, that a multi-word phrase "introduction to neur" will match "introduction to neural nets", but "intro to neur" will not.

This option increases the response time significantly if short search words are used.

Maximum number of returned references
You can adjust the maximum number of matching references returned by the search engine. This number is 40 by default.

How to speed up your search

There are two reasons why searches might take a long time:
You are searching at a very busy time on our server when many users search simultaneously.
The busiest time is between 3pm and 6 pm local time, try to avoid these times. You can also look at the access statistics to find out more about the server load.

The query is not worded optimally or the options are not set optimally.
There are a couple of things you can do to make your query more efficient:
Exact word matching:
Exact word matching usually speeds up the search a lot. Check if you really need to use partial word matching for your particular search.
Use only the most significant words in your query, omit very common words such as of, the and the like.
Advanced search:
If you can formulate your query using a boolean combination of simple words (i.e. no phrases, only alphanumerical characters) and you do not need case sensitivity, then you are probably much better off using the advanced search interface

What to do if you did not find anything (improving your query)

First: If you still did not succeed, you should try to If you tried all of the above and still could not find what you were looking for then the bibliography collection probably does not contain it. The best solution is to create a bibliography relating to the topic you are interested in and to contribute that bibliography to the collection.
Copyright © 1995,1996
Alf-Christian Achilles <>
Last modified: Thu Aug 13 00:42:57 1998