Browsing Documents on a Dense Embedding that Captures ...

Browsing Documents on a Dense Embedding that Captures

Theme Evolution

Alessandro Perina

Microsoft Research

One Microsoft Way, Redmond, WA

alperina@

ABSTRACT

We describe a new interaction strategy for browsing documents consisting of text and images. The browser represents

a collection of documents as a grid of key words with varying font sizes that indicate the words¡¯ weights. The grid is

computed using the counting grid model [7], so that each

document approximately matches in its word usage the word

weight distribution in some window (6 ¡Á 6 in our experiments) in the grid. In comparison to other document embedding approaches, this strategy leads to denser packing of documents and higher relatedness of nearby documents: The two

documents that map to overlapping windows literally share

the words found in the overlap. This leads to smooth thematic

shifts that can provide connections among distant topics on

the grid. The images are embedded into the appropriate locations in the grid, so that a mouse over any location can invoke

a pop-up of the images mapped nearby. Once the user locks

on an interesting spot in the grid, the summaries of the actual

documents that mapped in the vicinity are listed for selection.

In this document browser the arrangement of related words

and themes on the grid naturally guides the user¡¯s attention to

topics of interest. For an illustration we describe and demonstrate (in video submission) a browser of four months of CNN

news.

INTRODUCTION

Summarizing, visualizing and browsing text corpora are

important problems in computer-human interaction. As the

data becomes more massive, ambiguous, or conflicting, it

may become hard for people to glean insights from it. To

help the users, researchers have developed several visual

analytics tools facilitating the analysis of such corpora.

These tools are used to interactively make sense of complex

datasets, a process referred to as sensemaking [10].

We describe a new approach to browsing documents consisting of text and images, e.g. news stories on the web, social

media, special interest web sites, etc. The browsing through

documents is based on the exploration of the hidden variable

Permission to make digital or hard copies of all or part of this work for

personal or classroom use is granted without fee provided that copies are

not made or distributed for profit or commercial advantage and that copies

bear this notice and the full citation on the first page. To copy otherwise, or

republish, to post on servers or to redistribute to lists, requires prior specific

permission and/or a fee.

CHI¡¯13, April 27¨CMay 2, 2013, Paris, France.

Copyright 2013 ACM 978-1-XXXX-XXXX-X/XX/XX...$10.00.

Nebojsa Jojic

Microsoft Research

One Microsoft Way, Redmond, WA

jojic@

of the on the counting grid (CG) generative model [7], which

has recently been used for a variety of tasks related to regression and classification. The counting grid model represents

the space of possible documents as a grid of word counts.

Each individual document is mapped to a window into this

grid so that the tally of these counts approximately matches

the word counts in the document. The grid can vary in size,

and so can the window. As the documents are allowed to be

mapped with overlap, in order to maximize the likelihood of

the data, the learning algorithm has to map similar documents

to nearby locations in the grid, so that the words that the two

documents share appear in the grid positions in the overlap of

the corresponding windows. This leads to a compact representation where the theme of the documents smoothly varies

across the grid, achieving a higher density of packing than

previous embedding approaches (e.g. Egypt unrest news are

placed close to other stories about Arab Spring, with Libya

taking another distinct location in that area of the CG; nearby

are stories about oil prices, and near these are more stories

about the markets and economy, near which are stories referring to Fed¡¯s Bernanke, near which are stories about congress

and the President, which, in a counting grid defined on a torus

may loop back to Libya through military themes.) To provide

natural means of summarization and browsing of the documents, we render a CG representation based only on the most

frequent words in each position. We further embed the images from each document into the appropriate locations in the

counting grid, so that they can pop up when the user focuses

on a particular area of the grid (e.g. by mouse over). This provides the user with both a global and local perspective on the

underlying set of documents and their relationships, without

observing directly the underlying documents, but rather the

CG model¡¯s representation of the document space. Once the

user locks on an interesting spot in the grid, the summaries of

the actual documents that mapped in the vicinity are listed for

selection. This idea leads to an intuitive document browser

that is especially well suited to touch devices, where moving

a cursor is the most natural interaction modality, while typing

is particularly difficult. Additionally, the interface assists the

user in discovering documents of interest without having to

define a particular target and associated keywords first: The

arrangement of related words and themes on the grid naturally guides the user¡¯s attention to topics of interest. For an

illustration we describe and demonstrate (video submission) a

browser of four months of CNN news from winter and spring

2011, a period particularly rich in news-worthy events.

COUNTING GRIDS (CGS)

¦Ðz

...

k-1

k

k+1

k+2

...

...

i-1

music

afghan

alcohol

religi

spirit

religi

mean

influenc

spirit

blame

homeland

reason

oil

crazi

recent

rid

realiz

side

lose

hear

failur

conflict

gunfir

unanim

eastern

diplomat

gadhafi

libyan

pope

mean

radic

speech

radic

youth

syria

undermin

islam

repeat

revolut

cabl

cabl

ra

local

fight

sanction

altern

kimoon

secretarygener

resolut

justic

digniti

necessari

aircraft

necessari

offer

unit

cuba

court

guantanamo

trial

attack

oldest

brought

rock

threw

holm

corrupt

ralli

toppl

zine

abidin

tunisia

tunisia

upris

tunisian

flee

capit

upris

african

ambassador

asset

closer

adopt

refer

ban

loss

ban

i

council

council

libya

i+1

attack

particip

islamist

scene

angri

inspir

protest

peac

protest

mubarak

hosni

peac

protest

gather

demonstr

chant

prodemocraci

tuni

ali

capit

regim

clash

libya

tripoli

crackdown

demonstr

protest

protest

africa

opposit

daraa

opposit

libya

ben

libya

civilian

nato

militari

civilian

militari

minist

group

egypt

sanaa

ouster

ouattara

ivori

resolut

rebel

gbagbo

muslim

robertson

gadhafi

moammar

libyan

libya

moammar

nato

gbagbo

rebel

libyan

...

unrest

gadhafi

nato

libyan

libya

libyan

gadhafi

misrata

coalit

nofli

fighter

bomb

nofli

strike

19 Jan 2011 : Protesters march in Tunis amid

Arab League fears

27 Jan 2011 : Social media's role in North

Africa's unrest

11 Apr 2011 : None of us was Gadhafi's

lover, one of his nurses says

28 Feb 2011 : N. Africa, Mideast protests:

Diplomat says Libya long thought Gadhafi

crazy

21 Feb 2011 : N. Africa, Mideast protests Gadhafi: I'm still here

23 Feb 2011 : N. Africa, Mideast protests:

Libya's interior minister kidnapped, media

say

27 Feb 2011 : N. Africa, Mideast protests:

Egyptian stock market set to reopen Tuesday

Figure 1. A part of the counting grid trained on the news stories. Three windows are highlighted along with seven stories that mapped there. Color

indicates the mapping. The movement through the grid captures the spread of the Arab Spring in North Africa, and the subsequent UN reaction.

The counting grid consists of a set of discrete locations indexed by ` in a map of arbitrary dimensions (30¡Á30 to 40¡Á40

2D torus grids in examples here). A part of a counting grid

is illustrated in Fig 1. Each location contains a different set

of weights for the Z words in the vocabulary (Z = 10000

here). The weight of the z-th word atPlocation ` is denoted by

¦Ðz,` and the weights add up to one, ` ¦Ðz,` = 1. Thus ¦Ð is a

probability distribution over words and defines the local word

usage proportions. (These weights are partially illustrated in

Fig. 1 using font size variation, but showing only the top 3

words at each location.) A document has its own word usage

counts cz and the assumption of the counting grid model is

that this word usage pattern is well represented at some location k in the grid in the following way: When a window of

a certain size is placed at location k in the CG, and the CG

weights are averaged across

P N CG locations in the window

Wk to obtain hz = N1 `¡ÊWk ¦Ðz,` , then this distribution is

approximately proportional to the observed document counts

hz ¡Ø cz . In other words, approximately the same words in the

same proportions are used in the document and in its corresponding counting grid window Wk . The window size 6 ¡Á 6,

and thus N = 36 was used in our experiments, but due to

space limitations 3 ¡Á 3 windows were used in Fig. 1.

The CG estimation algorithm [7] starts with a random initialization which gives all words roughly equal weights everywhere. The subsequent iterations (re)map the documents

to the windows in the grid and rearrange words to match the

weights currently seen in the grid. In each iteration, after the

mapping, the grid weights at each location are re-estimated to

match the counts of the mapped document words. We found

that the algorithm converged in 70-80 iterations, which sums

up to minutes for summarizing months of news on a single

standard PC. As this EM algorithm is prone to local minima,

the final grid will depend on the random initialization, and

the neighborhood relationships for mapped documents may

change from one run of the EM to the next. However, as

shown in the supp. material, the grids qualitatively always

appeared very similar, and some of the more salient similarity relationships were captured by all the runs (e.g. the

Arab Spring news that referred to multiple different countries

with very different unfolding of events are always grouped

nearby). More importantly, a majority of the neighborhood

relationships make sense from a human perspective and thus

the mapping gels the documents together into logical, slowly

evolving themes. As discussed below, this helps guide our visual attention to the subject of interest. As the algorithm optimizes the likelihood of the data, all resources (grid locations)

must be used, and the packing is much denser than in the

previous embedding approaches, thus occasionally squishing

themes together even though no documents map to their interface. In our opinion, it is a small price to pay for high real

estate utilization and, for the most part, intuitive arrangement

of themes

MULTIMODAL CG DISPLAY AND BROWSING

To browse a collection of multimodal documents consisting

of both text and images, we first fit a CG model to the corpus, and then embed the images into appropriate locations of

the grid, so that each image is placed in the grid position in

the center of the window to which the source document was

mapped (Fig. 2). This results in a grid of images of the same

size as the word counting grid with a rough semantic alignment: In each image¡¯s vicinity the grid locations have high

weights on the words related to the image. Obviously, there

is now a multitude of possible approaches to visualizing this

embedding in a way that explores the two modalities in concert. To show the image embedding, we can simply show a

tiling of images (e.g. based on the 30 ¡Á 30 CG). In locations where multiple images are mapped, we can pick one at

random (as in our experiments), or the one that was used in

multiple documents, or the one selected by a computer vision

algorithm. In addition, the images mapped to the same location can slowly cycle. To visualize the CG word weights ¦Ðz,`

in each grid location, we show the top k words (k = 3 in our

experiments) using the font size to indicate the word weight.

In our browser, we can switch between the two representations, or show them one on top of the other with a certain

level of transparency (Fig. 2). In addition, a pointer (a mouse

cursor, fingertip on touch devices, etc.) can be used to force

the switch between images and words locally in a window of

a certain size (5 ¡Á 5 in our experiments). In this way the user

can base their exploration primarily on one modality, bringing the other modality to the fore by hovering over the grid

parts of interest. In particular, we find the word representation particularly useful in drawing the user¡¯s attention across

related themes to the point of interest. As the user naturally

moves the pointer toward their eyes¡¯ focal point the pointer

uncovers images underneath to further refine the user¡¯s understanding of the grid content. At any point, the user can stop

and indicate (e.g. by a click) their desire to see the source

documents that mapped in this region. We implemented two

ways of uncovering the images in the region where the user

hovers. In the first approach, the words in the grid locations

around the cursor are highlighted and the images from these

locations are shown next to the highlighted area. In the second approach, we simply replace the area around the cursor

with images. As the embedding is based on overlapping windows, in both cases it is possible that some of the images that

pop up this way are related to the themes slightly outside the

highlighted area. Once the user is used to this it becomes imperceptible as the matching words (or images) are never far

and slight movements of the pointer help lock onto the topic

of interest. To further indicate the smooth nature of the mapping, we experimented with varying sizes and intensities of

images that pop up. For example, in Fig. 2 the central image

of the highlight is of larger size and it slightly overlaps the

6 images around it, which themselves are larger and overlap

even more the images around them, creating an impression

of the underlying images popping out from the words, with

relationship being approximate but smooth, inviting the user

to move the cursor around.

Although the CG model glues the documents together based

on the vocabulary overlap that can contain a large number of

different words, to a human observer, just the top words for

each location seem to provide enough insight into the thematic shifts in the grid. The grid in Fig. 2 gels the disaster

stories together due to their common vocabulary (e.g. disaster, response, emergency, etc.), but in the browser most of

that shared vocabulary is overtaken by the words that get high

weight in individual locations (earthquake, tornado, airplane,

crash, snow, storm, etc.). The human mind easily detects connections among these and need not observe all of the ¡°glue¡±

that linked these topics together. In our experience, the CG

visualization seems to stimulate the user¡¯s own associations

and memory and guides the user to the target even if they did

not start with a particular target in mind: A look at a salient

Japan and earthquake keywords creates an association with

local weather disasters, reminding the user that they were following an airplane crash story. This association process is

guided by CG¡¯s own ¡¯associations¡¯ so that the spot in the grid

is found quickly. Further interaction with the grid to invoke

visual stimulus increases the pace of news discovery.

To accommodate for variable display sizes and corpora diversities, we can train a hierarchy of CG models of various sizes,

where model of one size is initialized by an upsampled version of the model of the smaller size. In this multi-granular

approach, the user can zoom in and out of any part of the grid.

Window size choice provides the tradeoff between finer document overlaps and the computational complexity of the CG

estimation, but for the CNN news stories at least, the latter

was not a limiting factor.

DISCUSSION

Our approach provides some important advantages over

the existing visualization/browsing/search approaches. The

10x10 grid website 1 also arranges images into a grid. But,

the placement of images is not optimized so that the nearby

locations capture related stories. Previous methods for spatially embedding documents [5, 2] produce sparse representations (e.g. ¡°The Galaxy of News¡± [8] ), which are only

locally browsable, whereas the counting grids use the screen

real estate much more efficiently. In addition, our approach

allows embedding of multiple modalities. Various galaxy

approaches required that the user interact with the embedding through the statistical model, manipulating its parameters and/or weights, which may be impenetrable to the user,

thus requiring a laborious guess and check strategy [1, 6].

This issue is still a subject of research in HCI [3]. In contrast, the CG parameters (grid size and the scope of overlap,

i.e. the window size), are more intuitive, and multi-granular

approaches may remove a need for parameter selection altogether.

The CG visualization reminds one of tag clouds, visual representations that indicate frequency of word usage within textual content. Google News Cloud 2 sorts words alphabetically, varying the font based on the relevance. If a word is

selected other similar words are highlighted. But the links

among the complex documents that combine a variety of

words are not evident. Other tools (e.g., Toronto Sun, Washington Post websites) cluster words based on co-occurrence

or proximity and then position the words belonging to the

same clusters near each other and use color to emphasize the

structure. Still, the words are not spatially embedded within

a cluster, and so only cluster hopping can be performed, in

contrast with smooth thematic drifts found in CGs. For the

most part, the tag clouds are designed to provide a useful and

visually pleasing summary of the news [9, 4], rather than a

1

2





two-dimensional densely organized multimodal browsing index which CG provides. In terms of providing a means for

traversing an organization of news, our method shares some

similarities with Newsmaps3 which use a hierarchical representation, a tree. But the traversal paths descend along the

branches of the tree while CGs often capture many different

directions of thematic drifts which can loop back.

REFERENCES

1. Alsakran, J., Chen, Y., Zhao, Y., Yang, J., and Luo, D. STREAMIT:

Dynamic visualization and interactive exploration of text streams.

131¨C138.

2. Chen, Y., Wang, L., Dong, M., and Hua, J. Exemplar-based

visualization of large document corpus (infovis2009-1115). IEEE

Transactions on Visualization and Computer Graphics 15 (2009),

1161¨C1168.

3. Endert, A., Fiaux, P., and North, C. Semantic interaction for visual text

analytics. In ACM CHI (2012), 473¨C482.

4. Helic, D., Trattner, C., Strohmaier, M., and Andrews, K. Are tag clouds

useful for navigation? a network-theoretic analysis. Journal of Social

Computing and CyberPhysical Systems 1 (2011), 33¨C55.

5. Iwata, T., Yamada, T., and Ueda, N. Probabilistic latent semantic

visualization: topic model for visualizing documents. In ACM KDD

(2008), 363¨C371.

6. Jeong, D. H., Ziemkiewicz, C., Fisher, B. D., Ribarsky, W., and Chang,

R. ipca: An interactive system for pca-based visual analytics. Comput.

Graph. Forum 28 (2009), 767¨C774.

7. Jojic, N., and Perina, A. Multidimensional counting grids: Inferring

word order from disordered bags of words. In UAI (2011), 547¨C556.

8. Rennison, E. Galaxy of news: An approach to visualizing and

understanding expansive news landscapes. In ACM Symposium on User

Interface Software and Technology (1994).

9. Sinclair, J., and Cardew-Hall, M. The folksonomy tag cloud: when is it

useful? J. Inf. Sci. 34 (2008), 15¨C29.

10. Thomas, J., and Cook, K. Illuminating the Path: The Research and

Development Agenda for Visual Analytics. IEEE Press, 2005.

3



A

media

spy

corp

push

york

ticket

iran

pack

american

held

steeler

embassi

open

relat

staff

seat

song

barricad

leav

depart

ferri

bullet

land

staterun

charter

malta

aircraft

statement

super

bowl

biggest

weak

close

solid

record

consecut

focu

emploi

tourism

road

tour

eat

wors

observatori

half

fan stadium

crash taiwan

field

australiaflooddodger

plane

airport

vice

construct

hood

negoti

citizen

hezbollah

travel

frustrat

immedi

drill

remov

reportedli

flight

runwai

structur

zealand

seek

michigan

blizzard

ici

mph

total

chicago

ic

snow

oklahoma

rain

metro

fall

omar

pt

mobil

termin

march

food

contact

toyota spokesperson

prime

naoto

subsequ

helicopt

blast

disast

fuel

reactor

weight

devast

affect

record

japan

wave coast

earthquake

boat mile

tsunami

ukushimanuclear

water

explos

evacu

coastal

chile

pose

kilomet

interact

ocean

tower

regul

monitor

accid

fish

leap

studi

asia

lion

scientist

vietnam

columbia

wrote

documentari

side

radio

buckl

habitat

atmospher

particular

veteranforest

learn

hit

wind

heavi

central

atlanta

lotteri

explet

grim

overal

author

washington

visit

check

ticket

detroit

gov

walker

contribut

law

islam

worth

transit

jakarta

controversi

muslim

iraq

foreclosur

rai

imag

coron

judg

testifi

plea

violat

lo

murrai

oprah

post

secret

teacher

perri

tone

critic

connect

listen

radio

terribl

childhood

chandler

span

link

restaur

fisherman

object

border

insid

rio

param

valu

rout

slowli

gen

improv

growth

lohan

look

worri

cia

cent

chines

bed

lindsei

note

increas

price

deadli

far

saudi

case

grace

hln

nanci

heavili

gunshot

snatch

gunmen

hospit

haleigh

rosemari

car

novemb

ireland

san

eventu

pearson

bank

ipad

energi

jump

zoo

quarter

market

index

share

verizon

ad

stock

dow

jone

sell

corpor

averag

nasdaq

industri

injuri

doctor

wit

river

girl

s

cooki

low

compani manag

feed

allow

B

Japan

11-Mar-2011

cl Fuk

ea u

r E sh

m im

er a

ge

nc

y

D

specif

homeland

syria

zine

protest

societi

crazi

repeat

tunisia

ben

unrest

demonstr

capit

protest

bahrain

penguin realiz

cabl

flee

davi

fight

african

medic

awar

act

wild

refer

sanction

crane ban

flown

unknown

held

enforc

shot

campu

raid

cuba

mubara

progovern democraci

libya gadhafi

limit

crisi

ministri

missil

smoke

ukrain

coalit

gbagbo

count

violenc

immigr

perjuri

franc

wound

miami

multipl

alleg

terror

switzerland

airmen

custodi

appeal

misconduct

assang

bin

student

stab

weapon

necessari

nato

militari

bombzone

air

ignit

book

europ

restrict

brown

drug

journal

michael

chicken

sexual

carri

appl

dead

doc

migrant

sex

financ

itali

execut

popular

newspuls

duke

rape

david

page

passport

page

canada

five

cooper

facebook

abl

digit

slain

fear

commit

work

hand

googl

app

suicid

soldier

son

felt

jon

driver

reader

pictur

iraqi

fit

air

traffic

asleep

true

robot

lewi

regist

gaga babi

parent

teen

bodi

try

jersei

search

woman

iowa

mother prostitut

attent

daughter legal

baltimor

barn

handl

card

reach

guarante

agreement

spend

stewart

marijuana

love

knew

good

ring

abortparenthood

shutdown

tax

cut

run

shut

health

fund

billion

debt

job

loomi

want

reader

budget bush

usa

queen

job

sure

ly

soccer

internet jame

good

tonight

wood

career

pitcher

school

break

cleveland

playoff

victori

basebal

brandon

deal

game

cap

butler

sport quarterback player

leagu

ncaa

tournament

nfl

denver

pit

number

championship

hometown

tiger

thought

photo

martin

face

joi

fbi

pennsylvania

ramirez

fellow

ryan

perman

king

cup

england

royal

explain

hack

respect

nick

just

palac

wed

happi

social

facil

embassi

babi

past

wikileak

high

princip

leonard assist

patrick

5

6

7

wind heavi

wind

airport cancel

tulsa

tulsa

system

collaps

airlin temperaturovernight region

blizzard

ici

mph storm tornado tornado south

outag weather forecast weather alabama

file

dalla

chicago blizzard storm weather midwest servic southern

structur zealand seek

futur

cross

send

clear christchurch main

snow snow storm storm line southern

food contact toyota

spokesperson

chicago

blog infrastructur suppli hardest

total

dalla

ic

inch oklahoma southeast servic northern

phone marathon sendai

effect

declar

et

warn missouri plain

morri tennesse brace

et

western

naoto subsequ helicoptcoastal weight chicago

oklahoma

snow

donat impact ireport

et

ic

inch

snow snowstorm tree

brace

kan

expos

impact

fat

shore

feet

snow

ic

inch vermont georgia georgia

veget

test

disast japanes tsunami devast

metro

affect

record

fall northeastgeorgia lightn

cold

highwai blast

tree

japanes japan

inch

fallen

toll

trigger

warn advisori

custom emerg stress

reactor prefectur earthquak tsunami

hit

feet accumul northwest affili

fuel

reactor

japan

wave coast motorist arctic ground atlanta

tokyo tsunami tsunami

earthquak ireport

feet

mile

event atlanta

path

hail

fukushima tokyo

littl

escap

touch

radiat radioact plant damag magnitud harbor extent remot arkansa

sea arkansa highwai

dig

fukushimanuclear japan japan quak boat

mile

bu

daiichi radioact tokyo

stuck

mile

cover

rock

pull

firefight

tokyo magnitud quak

radiat reactor nuclear damag earthquak pacif

ocean

area

sgt

fell

hunt

bear

explos nuclear nuclear

water quak chile volcano globe commiss difficultineighbor

headquart radiat

yukio

pool

trace

struck

floor

hawaii

bird

ev

bear

pull

contain

level

crippl rescuer kilomet kilomet

usg

pattern trauma midnight neighbor firefight

pose

kilomet interact

evacu

secretari level

rise

core

metropolitan

chernobyl chernobyl initi

evacu

small

2

3

4

5

6

7

8

9

10

11

12

13

regul

updat

mississippi

winter

rain

Weather Disasters

japan

water

tri

4

malta

sat

sourc

hood termin airport

crack

cyclon

port

fort tomorrow herald

frustrat airport

expedit flight

fly

milwauke

malta aircraftplane

mobil

path

collabor

juri

hay

bullet

land

land

jack

republ charlott

hole

hole

happen

pilot

tribun

updat

updat

michigan

detroit

explos

maryland

transport

egg

cyru

arial

yemen

syrian

libyan

council

court

johnson

egypt

cairo

reported flight runwai

Snow Emergency MidWest

Hawaii Volcano Feb-2011

Chile

11-Feb-2011

27-Feb-2011

06-Mar-2011

system

peac

canadian

North Carolina

Tornado

Tennessee Tornado

Arkansas

Tornado

Pakistan/Iran

18-Jan-2011

inspir

artist

C

Mississipi Flood

Georgia Thunderstorm

Nu

islamist

corrupt

counti

8

9

10

11

12

13

14

Counting Grids coordinates

Tsunami

New Zeland

21-Feb-2011

rock

speech

america

ferri

sourc

Earthquake

pope

influenc

displai

food

tourist

robertson

spirit

alert

educ

sheriff suspect

brotherhood

squar

attack

terror

hear

shoot

ivan

nurs

conserv

constitu

trip

revolut

great

jim

loughner

outlet

nomine

advantag

rabbit

resign

music afghan

tie

presenc

somebodi

berlusconi

oil

machin

mike

newton

mental

arizona

gunman

china

citi

taken

giffordtucson

supermarket

januari

cultur

teacherpresent

isra

signauction

palestinianjudg

wisconsinunion

document

defens

california

brother

supremnegoti

afghanistan

actress

station

elimin

ballot

date

invit

yale

chrysler

vote

submit

believ

bone

iphonbernank

candid

luck

audienc

understand code

hama

press

nobel

anniversari

white

fire

femal

million

winner

chetri

discov

honor

shourd

bu

neighbor

observ

conclus

medic

counti

dig

difficulti

battl

reserv

birther

rick

doctor

respond

korea

comment

ohio

rep

resid

gold

miller

castro

door

colombia

ted

statu

brief

skull

confer

laugh

kasich

robert

branson

hostag

apolog

comment

repeatedli

weekend

daili

seven smartphon

muslim

releas

northeastgeorgia lightn

mississippi

winter

william

rd

journalist

progress

husband

decemb

publicli

capitol

wife

franklin

logan

bill

complex

investig

marriag

funer

vessel

son

straight

care

match

polic

book

brewer

ball

death

zombi

sen

wear

murder

fashion

senat

line

independ

deadlin

bird

gift

dont

birth

option

anim

hold

statu

aid

ground

delai

crew

parti

highwai

former

putin

appeal

elect

arctic

afternoon

drill

rule

reform

taylor

command

agreement

injunct

vice

rehabilit

academi navi

owner

circuit

biden

readi

sailor

richard

buffett

gibb

virginia

assign

free

anderson

award

novel

appreci

elector

hero

male

claim

yard

duncan

film

olbermann

answer

certif

presid

jersei

climb

bail

toilet

walter

councilman

treat

mad

morgan

poll

best

eight

schiller

jonathan

blood

guilti

statu

heart

samantha

western

reed

prison

deliv

draft

thank

job

second

southern

arkansa

april

show

houston

storm

sea

spacecraft

tattoo

south

motorist

fish

differ

snake

result

file

homelessmessag

train

tornado

planetmercuri phasejackson

research

discoveriearth

space nasa

challeng shuttl

mel

volunt

park

choos

sheensudaninterview

violent

univers

experi

burn

round

kentucki

memori

photograph

wildfir

hurt

volcanoglobecommiss

shark

plastic

taco

fun

whale

buri armstrong farm

busi

aviat

airlin

hire

swept mcdonald

cairo

museum

kati

outreach

15

16

17

...

Figure 2. Browsable counting grid. A. The text and image representation of the grid are combined with emphasis on text. In two locations images are

brought into the foreground. The grid is defined on a torus (with left matching the right and the top continuing at the bottom). Various theme drifts

are visible, e.g. the japan-tsunami-water-whale-study- scientist-research-development-space-shuttle-nasa-command-navy semicircle on the left, or the

region emphasized in B) and C) which captures the various disasters from the period. More can be seen in figures in supplemental material and the

video submission. The preprocessing of the words reduced them to their roots and also made other standard alterations used in text analysis, but the

unaltered words can be shown instead. B. Images mapped in the highlighted area. C. More of the top words in the highlighted area, and an illustration

of how the images were embedded: As each document maps onto a window, the images from the document go to a location in the window (top left in

the illustration to avoid clutter, but the middle of the window in actual implementation to provide more natural alignment). D. Some of the news that

mapped to the highlighted area. The area of interest can be selected by cursor hover and the news can be recalled by a simple click.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download