Upload
jonas-rice
View
213
Download
0
Tags:
Embed Size (px)
Citation preview
SYSTEMATIC THOUGHT LEADERSHIP FOR INNOVATIVE BUSINESS
Marek Kowalkiewicz, Konrad JuenemannSAP Research
Improving information quality inemail exchanges by identifyingentities and related objects
External reports and user studies
AIIM on Ford Motor Company: knowledge workers spend 15–25% of their time on non-productive information–related activities.
Time Wasted Searching (IDC*) A general average of time spent searching: 2.5 hours per day, or roughly 30% of the
workday. Only 50 % of searchers led to successful results.
Assumptions: salary = USD80000 annual plus benefits, 1000 knowledge workers, 2.5 hours searching on average, 50% of the information not indexed properly.
Result: An enterprise employing 1,000 knowledge workers wastes $48,000 per week, or nearly $2.5 million per year, due to an inability to locate and retrieve information.
In the aggregate, the Fortune 1000 stands to waste at least $2.5 billion per year due to an inability to locate and retrieve information.
http://www.viapoint.com/doc/IDC%20on%20The%20High%20Cost%20Of%20Not%20Finding%20Information.pdf
Trends
Mash-ups are becoming more and more important
Context-enriching technologies predominantly used in Web advertising
Operating systems and standard applications collect related information (Apple Mail)
Current situation
Email still a killer application, especially in businesses
Emails often suffer from low contextual information quality -> automatic provision of related information can improve that
Entity recognition technologies have matured -> can we use them to improve contextual information quality?
Object identification
NER – a subject of NLP
The best systems achieve over 93% of f-measure (humans: around 97%) in NER
Popular NER systems Open source: ANNIE Dual licence: Calais Commercial: ThingFinder
Yowie – Situation & Proposed Solution
Situation
Proposed Solution
Facilitating Technology / Selected Industries
Image
• Data and information volumes reaching epidemic levels
• Employees spending up to 25% of their time searching (but often not finding)
• Problem: „I don’t know what I need to know”
• Text processing technologies have matured
• MS Office has become fully extensible
• Create one-stop-shop for information gathering
• Provide actionable information (+ links to actions)
• Seamlessly integrate with standard office tools
• Become fully extensible and transparent to users
• Text analytics technology from Business Objects
• Business productivity software (i.e. MS Office)
• Cross-industry solution
• Defense as a first proof-of-concept industry
1. Analyse text1. Analyse text
2. Collect and showinformation
2. Collect and showinformation
3. Link to actions
3. Link to actions
4. Trigger externalapplications
4. Trigger externalapplications
Yowie in MS Outlook
Yowie4Outlook• Fully integrated with MS Outlook 2007
• Ability to access internal information (contact folders, email content)
• Ability to create new documents in Outlook (emails, appointments)
• Pre-fetching related information as emails arrive
• Caching (pre-) fetched information for future use
• Customisable
Yowie in MS Word
Yowie4Word• Fully integrated with MS Word 2007
• Processing documents of any size
• Highlighting recognized entities
• Customisable roles
Yowie – Architecture
Integration with Business ProductivitySoftware
Access to other Applications (e.g. ERP)
Make it easy to extend Yowie
to support new entities /provide new actions
Recognize Entities
“Host Plugins”
“Guest Plugins”
Inxight ThingFinder™
“Modules”
Yowie – Architecture
© SAP 2008 / Page 10
Yowie Local (Application Domain)
DLL
Entity Recognition Engine(ThingFinder 4.2)
Web-Server
Host Plugin (Application Domain)
Guest Plugin (Application Domain)
Entity Recognition Engine(ThingFinder 4.2)
Yowie Host Plugin
Yowie GuestPlugin
Wrapper N
Yowie Local
R
Postprocessor N
Yowie Connector
Yowie / ThingFinder Mapper
R
Postprocessor 1
Wrapper 1
R
R
YowieRegistry
R
.NET Remoting
.NET Remoting
R
RWebService
© SAP 2008 / Page 10
“Host Plugins”
“Guest Plugins”
Inxight Thingfinder
“Modules”
…Sammy…
{Sammy Jankis,email=…}
{Sammy Jankis,email=…,revenue=…}
{Sammy Jankis,email=…,revenue=…}
Prototype summary
Directly access data and functionality in business productivity software
Able to access ERP, Web Services, applications, …
Extensible, modular architecture
Supports “hot-plugging” of modules
Automatic module composition
Offers background processing, caching, uniform configuration
Supports definition of user roles
THANK YOU!Contact: [email protected]