0
0
0
0
0
0
8
0
1
0
0
0
0
0
7
0
0
0
0
1
0
0
6
0
0
0
0
0
0
0
5
0
1
0
0
0
0
0
4
0
0
1
0
1
0
0
3
0
0
0
0
0
0
0
2
0
1
0
0
0
0
0
1
0
0
1
0
1
0
0
1
0
1
0
2
0
1
0
1
2
0
1
0
0
1
0
0
1
df (Sum over tes) ...
df2 ...
df1 ...
df0 ...
0
3
(SetPred=sum>0) ...
9
0
Pte,doc=1
Term Pos
A pTree organization
for a corpus of documents
1st occurence pos
0
9
3
0
6
0
0
term existence
1
1
0
1
0
0
0
1
0
0
1
0
0
0
1
1
0
0
0
0
0
1
1
0
1
0
0
term frequency
Atf,doc=1
0
7
6
0
3
0
a
or
on
no
in
it
to
3
2
3
1
2
2
3
0
0
0
0
0
3
1
0
1
0
0
0
0
Best pTree organization?
level-1 gives te, tf (term level)
level-2 gives df.(document level)
level-3 ? (corpus level)
0
TermArray (= ContentWord/Stem?)
DocFreqArray (df)
Gives the Comprehensive Text ?
rooted by all digital media,
then libraries,
then corpi,
than documents,
then texts,
then terms...
Level-2 pTrees
1
1
0
Level-1 pTrees
1
Ptf,doc=1,0
Ptf,doc=1,1
Ptf,doc=1,2
0
0
tf ...
Pte,doc=1
0
tf2 ...
tf1 ...
tf0 ...
Afop,doc=1
book1
I Heard a
Little
Baa
22 words
this
lively
rhyming
book
gives
readers
a
chance
to
practice
animal
sounds
identify
animals
and
guess
what
animal
is
coming
up
next
book2
Mr. Grumpy's
Outing
32 words
book3
The Very Hungry
Caterpillar
28 words
book4
Pat The Bunny
book5
Goodnight Moon
32 words
14 words
on
a
perfect
summer
day
mr
grump
decides
to
take
a
ride
on
the
river
in
his
small
boat
a
goats
kick
causes
a
cumulative
effect
and
everyone
lands
in
the
water
a
very
hungry
caterpillar
works
his
way
through
holes
cut
into
pages
as
he
eats
a
variety
of
foods
spins
his
cocoon
and
emerges
as
a
beautiful
butterfly
the
classic
touch
and
feel
book
in
which
a
boy
and
girl
experience
the
world
around
them
by
patting
a
bunny
looking
in
a
mirror
smelling
flowers
and
touching
daddys
scrathy
face
a
young
rabbit
says
goodnight
to
everything
in
his
room
before
bedtime
book6
Moo Baa La La La
24 words
book7
Clap hands
18 words
the
animals
in
this
rhyming
board
book
demonstrate
the
very
different
and
amusing
sounds
that
they
make
the
humor
is
just
right
for
toddlers
clap
hands
dance
and
spin
open
wide
and
pop
it
in
sweet
babies
go
through
various
rhymed
activities
book8
Spots, Feathers
and Curly
Tails
31 words
who
does
that
curly
tail
belong
to
what
animal
has
black
and
white
spots
just
turn
the
page
to
reveal
the
familiar
farm
animals
that
match
the
verbal
and
visual
clues
book9
Baby Bathtime
book10
Higher, Higher
13 words
29 words
this
lift
the
flap
board
book
helps
babies
explore
the
world
around
them
when
her
dad
pushes
a
young
girl
in
a
swing
she
goes
higher
higher
until
she
reaches
fantastical
heights
and
meets
an
airplane
a
rocket
and
even
an
alien
book11
Rhymes Round the
World
19 words
book12
Shades of
People
18 words
rhymes
and
songs
from
across
the
globe
will
delight
young
children
full
page
illustrations
dramatize
the
sense
of
play
beautiful
full
color
photographs
of
children
along
with
a
simple
text
highlight
the
many
shades
of
skin
color
book13
The Sleepy
Little
Alphabet
32 words
parents
try
to
round
up
the
letters
of
the
alphabet
as
they
skitter
scatter
helter
skelter
rhyming
text
and
humorous
illustrations
highlight
each
letter
until
they
are
all
finally
asleep
in
bed
book14
Ten Tiny Babies
14 words
babies
enjoy
a
bouncy
noisy
day
until
they
are
finally
fast
asleep
at
night
Latent semantic indexing (LSI) is an indexing and retrieval method that uses Singular value decomposition to identify patterns in the
relationships between the terms and concepts contained in an unstructured collection of text.
LSI is based on the principle that words that are used in the same contexts tend to have similar meanings.
LSI feature: ability to extract conceptual content of a body of text by establishing assoc between those terms that occur in similar contexts.[1]
Called Latent Semantic Indexing because of its ability to correlate semantically related terms that are latent in a collection of text. It uncovers
the underlying latent semantic structure in the usage of words in a body of text and how it can be used to extract the meaning of the text in
response to user queries, commonly referred to as concept searches. Queries against a set of docs that have undergone LSI will return
results that are conceptually similar in meaning to search criteria even if results don’t share specific word or words with the search criteria.
Benefits of LSI: LSI overcomes two of the most problematic constraints of Boolean keyword queries:
multiple words that have similar meanings (synonymy)
and words that have more than one meaning (polysemy).
Synonymy and polysemy are often the cause of mismatches in the vocabulary used by authors of docs and users of info retrieval systems.[3]
As a result, Boolean keyword queries often return irrelevant results and miss information that is relevant.
LSI is also used to perform automated document categorization.
Doc categorization is assignment of docs to one or more predefined categories based on their similarity to conceptual content of the categories.[5]
LSI uses example documents to establish the conceptual basis for each category.
During categorization, the concepts contained in the docs being categorized are compared to the concepts contained in the example items, and a
category (or categories) is assigned to the docs based on similarities between concepts they contain and the concepts contained in ex docs
Dynamic clustering based on the conceptual content of docs uses LSI. Clustering is a way to group docs based on their conceptual similarity to
each other without using example docs to establish the conceptual basis for each cluster - useful with an unknown collection of
unstructured text.
Because it uses a strictly mathematical approach, LSI is inherently independent of language. This enables LSI to elicit the semantic content of
information written in any language without requiring the use of auxiliary structures, such as dictionaries and thesauri.
LSI is not restricted to working only with words. It can also process arbitrary character strings. Any object that can be expressed as text can be
represented in an LSI vector space.[6] e.g., tests with MEDLINE abstracts have shown that LSI is able to effectively classify genes based
on conceptual modeling of the biological information contained in the titles and abstracts of the MEDLINE citations.[7]
LSI automatically adapts to new and changing terminology, and has been shown to be very tolerant of noise (i.e., misspelled words, typographical
errors, unreadable characters, etc.).[8] This is especially important for applications using text derived from Optical Character Recognition
(OCR) and speech-to-text conversion. LSI also deals effectively with sparse, ambiguous, and contradictory data.
Text does not need to be in sentences. It can work with lists, free-form notes, email, Web-based content, etc. As long as a collection of text
contains multiple terms, LSI can be used to identify patterns in relationships between important terms and concepts contained in the text.
LSI has proven to be a useful solution to a number of conceptual matching problems.[9][10]
LSI used to capture key relationship information, including causal, goal-oriented, and taxonomic information.[11] LSI Timeline
Mid-1960s – Factor analysis technique first described and tested (H. Borko and M. Bernick)
1988 – Seminal paper on LSI technique published (Deerwester et al.)
1989 – Original patent granted (Deerwester et al.)
1992 – First use of LSI to assign articles to reviewers[12] (Dumais and Nielsen)
1994 – Patent granted for the cross-lingual application of LSI (Landauer et al.)
1995 – First use of LSI for grading essays (Foltz, et al., Landauer et al.)
1999 – First implementation of LSI technology for intelligence community for analyzing unstructured text (SAIC).
2002 – LSI-based product offering to intelligence-based government agencies (SAIC)
2005 – First vertical-specific application – publishing – EDB (EBSCO, Content Analyst Company)
Mathematics of LSI (linear algebra techniques to learn the conceptual correlations in a collection of text).
Construct a weighted term-document matrix, do Singular Value Decomposition on it. Use that to identify the concepts contained in the text.
Term Document Matrix, A: Each (of m) term is represented by a row, and each (of n) doc is represented by a column, with each matrix cell, aij,
initially representing number of times the associated term appears in the indicated document, tfij. This matrix is usually large and very
sparse.
Once a term-document matrix is constructed, local and global weighting functions can be applied to it to condition the data. The weighting
functions transform each cell, of , to be the product of a local term weight, , which describes the relative frequency of a term in a
document, and a global weight, , which describes the relative frequency of the term within the entire collection of documents.
Some common local weighting functions [13] are defined in the following table.
Binary if the term exists in the document, or else TermFrequency , the number of occurrences of term in document Log Augnorm
Some common global weighting functions are defined in the following table.
Binary Normal GfIdf , where is the total number of times term occurs in the whole collection, and is the number of documents in which
term occurs.Idf Entropy , where
Empirical studies with LSI report that the Log Entropy weighting functions work well, in practice, with many data sets.[14] In other words, each
entry of is computed as: