Emerging Content Requirements for News Products

1
Emerging Content Requirements for News Products

• Howard Williams
• February 28, 2004
• IA Summit

2
Presentation Topics

• Current trends in online news
• Design issues for presentation of news content
• Reading learning the news
• Personalization summarization
• Explanation framing the news
• Emerging content requirements
• Enabling technologies
• Content management

3
Information Architecture
Reading Learning
?
Online News
Text Processing w/ News Corpus
Internet Capabilities
Reading the News
IR / AI
4
Current Trends in Online News

• Media convergence w/ multiple delivery channels
• News portals, content aggregation syndication
  services
• Web-based capabilities w/ multimedia
• News monitoring alert services
• Personalization based on reader profiles
• Interactive features and reader services
• Enriched text metadata for indexing retrieval
• XML-based specifications for content
  representation
• Sophisticated categorization of news content
• Advanced search retrieval capabilities
• Extensive use of news briefs (summaries)
• Domain specific applications (e.g., financial
  news)

5
Information Architecture for News Content

• Poynter Eyetrack (early 90s) - print
• Stanford-Poynter Eyetrack (late 90s) - online
• Redesigning Print for the Web (Mario Garcia)
• Reading online news
• Web information design
• Story structure
• Typography
• Color

6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Reading the News

• Scanning skimming is common
• Eyes gravitate toward visuals (print)
• Eyes gravitate toward text (online)
• Summaries often hold readers attention
• Headline level knowledge is typical
• Most of whats written isnt read!

12
Observations

• User-centric design
• Exploiting potential of unread text
• Extracting economic value from content
• Variables of usability that impact reading
  learning

13
Reading Learning

• Text can be enhanced for readability recall
• Summaries are good surrogates
• What readers learn is influenced by prior
  knowledge, goals strategies
• Readers benefit from clear, explicit connections
  between events in a story
• Readers look for why how information
  causes effects motives, goals beliefs of
  people
• In their absence, readers make inferences based
  on their own knowledge (mental models)

14
News Reading as Inference
Main Parties Boycott Ivory Coast Vote Coup
Leader's Candidacy Condemned by U.S., OAU
ABIDJAN, Ivory Coast, Oct. 22 Residents of
this once-stable West African country voted
today in a disputed presidential election that
pitted military leader Gen. Robert Guei against
mostly unknown civilians. The ballot was
boycotted by the two main parties and condemned
by the international community as not being free
and fair. Voter turnout appeared to be extremely
light throughout this city of 2 million and
across most of Ivory Coast, and there were no
reports of violence. Results are expected
Monday. The stability of Ivory Coast,
sub-Saharan Africa's third-largest economy, is
of growing concern to West Africa, already
racked by a festering civil war in nearby Sierra
Leone, new fighting in Liberia and growing
tension in Guinea.
15
10 Paragraphs Later
Ivory Coast was stable and relatively prosperous
under Houphouet-Boigny, its first president and
national icon, who ruled from independence from
France in 1960 until his death in 1993. Under
Guei, the military mutinied twice, demanding pay
raises that have left the country unable to
service its foreign debt. He has also tried to
capitalize on anti-foreigner sentiments fueled by
the slumping economy. Immigrants from
surrounding countries, who were welcomed by
Houphouet-Boigny, now make up about 40 percent
of Ivory Coast's estimated 16 million
inhabitants.
16
Enhancing Presentation Value of News Content

• Illustrated News (photos headlines)
• Highlighting main ideas
• Information chunking
• Easy to locate explanatory information
• Headers that signal function
• Summarization of essential points
• Explicit story connections

Writing stories vs. Re-presentation of stories
vs. Generation of news products derived from
large text databases multiple sources.
17
Presentation
Text
Content
Domain Knowledge
18
(No Transcript)
19
(No Transcript)
20
Bush to pose changes to immigration law for
Mexicans WASHINGTON (AP) President Bush is
proposing to let foreign workers who have U.S.
jobs waiting for them enter America a move
designed to help repair relations with Mexico and
capture Latino voters in this year's election.
Virtually silent on the immigration issue for
two years, Bush was working Tuesday on remarks he
will make about immigration on Wednesday
afternoon at the White House to about 150 people
active in immigration issues. White House press
secretary Scott McClellan would not say whether
the president would announce a detailed
proposal, or general outlines of a proposal to
match willing foreign workers, mostly from
Mexico, with receptive U.S. employers. There are
an estimated 10 million undocumented workers in
the United States, as many as half of them from
Mexico. "It's the issue of what he's talked
about before about matching willing workers
with willing employers where there are jobs that
Americans are not interested in filling,"
McClellan said. "That's an economic need that
exists. It's important that we have a fair
immigration policy and an immigration policy that
addresses those economic needs."
21
Domain Knowledge Reader Questions

• Who is George Bush?
• What are his motivations to change immigration
  law?
• What is the problem?
• Why do foreign workers want U.S. jobs?
• Why do U.S. employers hire foreign workers?
• Why is this change being proposed now?
• Who will benefit from the changes?
• What is the background to the issue?
• How does this impact U.S. workers?
• What are the underlying economic issues?

22
Presentation

• Personalization
• Summarization
• Explanation
• Framing

Content
23
Personalization

• Response to increasing volumes of info (filter
  out) and consumer preferences (filter in)
• Profiles predicated on good user modeling
• Breadth of variables (subject, source, location,
  display, editorial views, format, readability,
  learning style, detail of explanation, level of
  summarization, etc.)
• Depth of variables (coarse-grain to fine-grain)
• Fine-grain profiles are difficult to implement
  community coarse-grain preferred to personalized
  fine-grain

24
(No Transcript)
25
(No Transcript)
26
Summarization

• Response to info overload consumer needs
• Commonly used, wide reader appeal
• Includes headlines (single, multi-deck), leads
• Result of good writing copy editing
• High volume, automatic generation at multiple
  levels, from multiple sources, for multiple
  purposes
• Basic model extracts and ranks sentences
  according to feature-based criteria
• Creating purposeful summaries is a non-trivial
  task NLP required to smooth results

27
(No Transcript)
28
(No Transcript)
29
Explanation

• Answers why and how
• Often lacking, inadequate or buried in text
• Good explanations answer reader questions, and
  some are predictable
• Background (history, timeline)
• Causes functional mechanisms (how things
  work)
• Beliefs, motives goals of actors
• Good explanations reduce the readers inferential
  burden (explicit, not implicit easy to read)
• Relevant content needs to be surfaced

30
Foreign trade deficit falls slightly The
Associated Press Updated 1135 a.m. ET Dec. 16,
2003 The deficit in the nation's broadest
measure of foreign trade fell slightly to
135.04 billion this summer after setting an
all-time high in the spring, the Commerce
Department said Tuesday. The drop in the
current account trade deficit in the
July-September quarter reflected a 3.1 percent
improvement from a record quarterly deficit of
139.39 billion set in the April-June period.
Even with the small improvement in the third
quarter, the country is well on its way to
setting an all-time record deficit for the year
of more than 500 billion in the current
account. The current account is the broadest
measure of trade because it covers not only
trade in goods and services but also investment
earnings and foreign aid. The country's soaring
trade deficit has proven to be a major political
headache for President Bush with his Democratic
presidential opponents charging that the record
deficits reflect the administration's failed
economic policies. The improvement in the
current account balance in the July-September
period reflected a better showing in all of the
major categories.
31
Foreign trade deficit falls slightly The
Associated Press Updated 1135 a.m. ET Dec. 16,
2003 The deficit in the nation's broadest
measure of foreign trade fell slightly to
135.04 billion this summer after setting an
all-time high in the spring, the Commerce
Department said Tuesday. The drop in the
current account trade deficit in the
July-September quarter reflected a 3.1 percent
improvement from a record quarterly deficit of
139.39 billion set in the April-June period.
Even with the small improvement in the third
quarter, the country is well on its way to
setting an all-time record deficit for the year
of more than 500 billion in the current
account. The current account is the broadest
measure of trade because it covers not only
trade in goods and services but also investment
earnings and foreign aid. The country's soaring
trade deficit has proven to be a major political
headache for President Bush with his Democratic
presidential opponents charging that the record
deficits reflect the administration's failed
economic policies. The improvement in the
current account balance in the July-September
period reflected a better showing in all of the
major categories.
32
Foreign trade deficit falls slightly The
Associated Press Updated 1135 a.m. ET Dec. 16,
2003 The deficit in the nation's broadest
measure of foreign trade fell slightly to
135.04 billion this summer after setting an
all-time high in the spring, the Commerce
Department said Tuesday. The drop in the
current account trade deficit in the
July-September quarter reflected a 3.1 percent
improvement from a record quarterly deficit of
139.39 billion set in the April-June period.
Even with the small improvement in the third
quarter, the country is well on its way to
setting an all-time record deficit for the year
of more than 500 billion in the current
account. The current account is the broadest
measure of trade because it covers not only
trade in goods and services but also investment
earnings and foreign aid. The country's soaring
trade deficit has proven to be a major political
headache for President Bush with his Democratic
presidential opponents charging that the record
deficits reflect the administration's failed
economic policies. The improvement in the
current account balance in the July-September
period reflected a better showing in all of the
major categories.
A trade deficit occurs when imports are greater
than exports.
The balance of payments for a country is
separated into two main accounts the current
account and the capital account. The current
account records sales and purchases of goods,
services and interest payments. The
entire merchandise trade balance is contained in
the current account.
33
Framing the News

• Frames (themes, news plots) simplify complex
  events
• Conflict
• Horse Race
• Personality
• Institutional forces at work
• Underlying messages and sources of bias (e.g.,
  crime)

Mental Model
Frame
Filtering without design!
34
Game Theory Model as Frame

• Goal is to preserve event complexity
• Game models applied to real world events
• Structures of the model provide basis of frame
• Players
• Rules
• Strategic options
• Payoffs
• Preferences
• Goals motivations
• Relevant content needs to be surfaced
• Variable for consumer preference (as is liberal
  or conservative)

Again, we could write the story this way, or
re-present it, but were mostly interested in
how we could generate products from large text
databases multiple sources.
35
Emerging Content Requirements

• Content is multi-purpose
• Identify or derive features structures apply
  use for various purposes (e.g. categorization)
• Features structures might include
• Main ideas
• Cause-and-effect statements
• Arguments
• Historical, biographical, geographical content
• Timelines
• Analogies or similar events
• Novel information on events or topics

and more.
36
Enabling Technologies

• ASIST SIGIR MUC, TIPSTER, TREC
• KR, NLP IR approaches
• Information extraction
• Text mining
• Text categorization
• Topic/event/novelty detection
• Profiling filtering
• Question answering
• Summarization

37
XML

• NITF (News Industry Text Format)
• Enriched text elements
• NewsML
• Content syndication (RSS, ICE)
• PRISM

Standards help to reduce costs through a content
value-chain and may help to develop new products.
38
Content Management
Not a workflow!
Content processing
News Sources
Value-added processing
KB
Books, Etc.
39
Business Model
Supporting Capabilities
Content Requirements
40
Thank You!
Howard Williams Colorado Springs,
CO 719-651-8852 howard_williams_at_mindspring.com