LIS618 lecture 1 Thomas Krichel 2004-02-01. structure of talk Recap on Boolean (aurally) Before online searching Working with DIALOG –Overview –Search.

LIS618 lecture 1

Thomas Krichel

2004-02-01

structure of talk

• Recap on Boolean (aurally)• Before online searching • Working with DIALOG

– Overview– Search command

• Boolean exercise (on the fly)

before a search I

• What is the purpose of the query?– brief overview– comprehensive search

• What perspective on the topic is required?– scholarly– technical– business– popular

before search II

• What type of information does the patron want?– fulltext– bibliographic– directory– numeric

• Are there any known sources?– authors– journals– papers– conferences

before search III

• What are the language restrictions?

• What, if any, are the cost restrictions?

• How current need the data to be?

• How much of each record is required?

concept analysis

• This is the art/science of taking the topic to search for and develop facets. Example “Internet filtering in Libraries”– Internet filter– Libraries– Controversy not technical issues

• We may also need the think about the aim of the search.

search aims

• a known needle in a known haystack

• a known needle in an unknown haystack

• an unknown needle in an unknown haystack

• any needle in a haystack

• the sharpest needle in a haystack

• most of the sharpest needles in a haystack

search aims

• all the needles in a haystack

• affirmation of no needles in a haystack

• things like needles in a haystack

• is there a new needle in the haystack

• where are the haystacks

• needles, haystacks, anything

types of searches

• known-item searches

• negative searches

• selective dissemination of information

• topical or subject searches

• passage searching, where the user is only interested in part of the item

search strategies I

• Building block approach– Do a number of elementary searches– Combine the resulting sets with Boolean

operators

• This is what I did in the example in the previous lecture

• Works only with the Boolean model

search strategies II

• Snowballing approach– Start with a very specific query– Think of other term that can be added to get

more results– Stop when a reasonable number of results are

achieved.

• Not sure this really works well in practice.

search strategies III

• The successive fraction approach is the opposite of the snowballing approach– First search for a broad concept– Then repeat the query by adding various

limiting factors.

• Can work well if the IR system allows to repeat and edit queries.

• But queries can become unwieldy.

search strategies IV

• Most specific facet first– Conduct concept analysis– Look for the most specific facet– Search that first, add others later

• Presupposes that you have done a decent concept analysis.

two steps in DIALOG

• step one: select databases (aka files) to look at

• step two: perform searches on the selected databases

• You may wonder why one does not have one single step like in a search engine. Discuss.

• today we concentrate on the second step

working on selected files

• We assume that we have selected database that we know and we look at the search interface on the selected database.

• The database selection process is a bit more complicated, covered next week.

• First, let us login and look at the command prompt.

• Then we select the first database (file) with the begin command

the ‘begin’ command

• As its name suggests, usually the first command.

• begin number, number,…

• selects files with numbers number

• Once they are selected they can be searched.

• Now select the ERIC "begin 1"

• "Begin 1" can be abbreviated as "b 1"

substeps in the second step

• Identify search terms

• Use Dialog basic commands to conduct a search

• View records online or print the results

the 's' (select) command

• Once issued the "begin" command to select a database, we issue the "s" command on the database.

• "s query_expression" where query_expression is a query expression.

• This will search the index of selected database in full-text view for the query issued

• It will not find any of the following: "an and by for from of the to with". They are stop words.

query expression

• A query expression contains search terms expressed in special ways– You can truncate search terms. – You can build an elementary expression by

putting several keywords together. This is achieved by DIALOG's connectors.

– You can combine several expressions with the use of Boolean operators

• We will cover this is in turn now.

truncation of terms I

• Open Truncation– "select path?" retrieves all words that begin

with path: paths, pathos, pathway, pathology

• Controlled-Length Truncation– "select path??" retrieves the root and up to

two additional characters: paths, pathos

truncation of terms II• Embedded Character truncation can be used

for variant spellings:– "select organi?ation" -> organization

organisation – "select fib??board" -> fiberboard fibreboard 

• This truncation feature is also useful for searching for unusual plural forms:– "select wom?n" -> woman women

• Apparently you can also do prefixes by putting the ? in the beginning. – "?mobile" -> automobile metamobile

use of connectors

• Connectors are used to put several words together.

• One instance where this is useful is when you have words that on their own mean different things.

• For example "mate" is a herbal beverage consumed in South America. Looking for mate on the Internet retrieves a lot of singles' pages.

example: terms related to "mate"

What other terms to be used? – matear (drink mate)– matero (mate drinker)– cebar (prepare mate)– cebador (mate preparer) – yerba (mate herb)– bombilla (mate straw)

connectors I

• '(W)' requires terms to appear one after the other next to each other e.g. 'yerba(W)mate?' matches "yerba mate".

• '(i W)' where i is an integer, means followed by at most i words, e.g. 'ceba?(3W)mate?' matches "cebar un maravilloso mate" but not "cebador guapo mirando un buen mate"

connectors II

• '(N)' requires terms to be next to each other e.g. 'yerba(N)mate?' matches "yerba mate" or "mate yerba".

• '(i N)' where i is an integer, means proximity by at most i words, e.g. 'ceba?(3N)mate?' matches "cebar mate" or "matear con la cebadora".

• '(S)' searches for the occurrence of connected terms in the same paragraph.

using Boolean operators

• In your query, you can combine several expressions with Boolean operators

• Example: "S LIBRARY(W)SCHOOL? AND DISTANCE(W)EDUCATION"

• But I usually do not issue such fancy queries.

executing several searches

• There can be several searches done sequentially, and the results sets are saved by the system.

• Each time the system assigns a set number, Si,

• These can be combined in Boolean expressions, e.g. 's S1 or S2 and S3'

• Remember that Boolean operations are set-theoretic!

Boolean operators on sets

• When using Booleans, be aware that "and" has higher precedence than "or".

• Thus:a or b and cis not the same as(a or b) and cbut it is a or (b and c)

• Use parenthesis when in doubt

DS (display sets)

• This command can be executed any time to review the sets that have been formed since the last B (begin) command.

• This can be useful to review your search history.

the target command

• "target set" where set is a search result set creates a subset of the "statistically most relevant results" in the original set.

• I have not seen details about how this subset is computed.

• A new result set is being formed.

display: the type command

type set/format/range

• set is a result set

• format is a format

• range can be – start – end

• start is a record number to start• end is a record number to end

– all

standard delivery formats

• 2 -- full record except abstract• 3 or medium – citation• 5 or long – full except full text• 6 or free – title and dialog number• 8 or short – title plus indexing terms

– useful to find other indexing terms

• 9 or full – everything• KWIC or K – keywords in context

options for delivery

• I once tried to email results to me, to no avail

• You can save the html of the search results in the browser.

• You can print the results within the browser.

http://openlib.org/home/krichel

Thank you for your attention!

• to do: set up consistent notation