Denmark in Europe: Benchmarking HLT Research

1
Denmark in Europe Benchmarking HLT Research
Development

• EUROMAP National Seminar - Copenhagen
• 30 April 2003
• Rose Lockwood
• Director of Research, Bowne Global Solutions

2
The EUROMAP Final Report Benchmarking HLT
Progress

• COVERING
• State-of-the-Art in HLT
• The Shape and Evolution of the Market
• Driving Next-Generation HLT Policy Imperatives
• The HLT Scorecard
• HLT Benchmark
• HLT Opportunity Index
• Conclusions Recommendations

3
State-of-the-Art in HLT

• What is HLT and who are the companies making it
  happen?
• Segmentation of the technologies
• Introduction to applications and opportunities
  for technology transfer
• Review of types of HLT players in Europe
• Snap-shot profiles illustrative of the rich HLT
  scene in Europe
• Comprehensive directory of European researchers
  and suppliers (to be published separately)
• Showing focus and types of applications
• Languages covered by tools and products

4
From components to complex solutions the focus
of HLT applications
5
Grounded in the basics suppliers of
Componentware and Resources

• Neurosoft (Greece) Greek language components
  for text mining
• Polderland Language Speech Technology
  (Netherlands) - Core components for multiple
  languages
• Connexor (Finland) Embedded multilingual
  language analysers
• Daedalus (Spain) - Document processing tools for
  Castilian

6
Innovators - Interface and Interaction

• Telisma (France) Speech recognition for
  telecommunications voice services
• Auralog (France) Speech technology for the
  language learning industry
• Sympalog (Germany) State-of-the-art voice
  dialog systems
• Loquendo (Italy) A global speech technology
  powerhouse
• Rhetorical Systems (UK) - High quality speech
  synthesis in multiple languages

7
Innovators in Cross-Lingual Applications

• ESTeam (Sweden) - Resource-driven translation
  automation
• Sail Labs Technology (Austria) - An advanced
  language understanding agenda
• Synthema (Italy) Tools for multilingual
  knowledge management
• Systran (France) Industrial-strength machine
  translation
• Aixplain (Germany) Cross-lingual solutions for
  speech and text

8
Innovators in HLT-based Knowledge Processing

• Ankiro (Denmark) User-centric dialogue and
  knowledge robots
• Language Computing (Belgium) - Semantics for
  medical knowledge
• Xtramind Technologies (Germany) Intelligent
  enterprise information processing
• Knowledge Concepts (Netherlands) - Boosting
  cross-lingual access to corporate content
• Wordmap (UK) Enterprise Taxonomy Management
  Systems

9
Ankiro - Products based on (English) language
tools and databases

• Advanced products that promote effective
  communication, information searching and
  knowledge management.
• Dialog robots used for
• Guidance (e.g. as guides on web sites)
• Support (e.g. as Call centre robots and FAQ
  robots)
• Entertainment (e.g. chat robots)
• Search engines and crawlers include
• Web search engines (available both as index based
  search and as full-text search)
• Site search engines (searches on a company's web
  page)
• Intranet search engines (searches through all of
  a company's databases)
• Lotus search engines (plug-ins for effective
  searches in Lotus applications)

10
Crossing the Chasm with Language Technology
embedded solutions
niche products
components
Market Take-up
Original Chart Source Geoffrey Moore, The Chasm
Group
11
Market Forecast Speech and Language Applications
TOTAL HLT product market circa 2B in 2004
CAGR 21
1600
50
1400
40
Total revenues
1200
language
Year on year growth
30
1000
speech
Revenues (m)
Year on year growth ()
20
800
600
10
400
0
200
0
-10

2000
2001
2002
2003
2004
Source Datamonitor, Voice Business Quarterly
Update (Q3 2002)
Source Steve McClure, IDC (2002)
12
Policy - Background and Evolution

• How important is HLT from a policy perspective?
• How has public investment supported the domain,
  and
• Who are the Showcase Labs?
• How has public-sector investment in HLT evolved
  over the years?
• National and EU programme timelines

13
Programme Support and HLT Transfer to Market
14
Are we ready for the next generation?Benchmarking
HLT in Europe
Demand-side Factors
Technology Development
Supply-side Factors
15
Benchmarking Method

• Measuring HLT maturity
• Factors include research and tech-transfer
  record, rates-of-investment (at national level,
  both public and private), breadth of language
  coverage
• Measuring HLT opportunity
• Factors include infrastructure and business
  environment
• Combining data from varied sources to establish a
  useful benchmarking scale
• HCID/World Economic Forum survey - uses a 1-to-7
  rating scale (1poor, 7excellent) for a wide
  range of questions
• Other studies normalised to this scale
  (Innovation Scorecard, EITO ICT infrastructure
  data, EuroStat, OECD, EUROMAP HLT maturity, etc.)

16
How to measure HLT maturity

• Expert opinion/knowledge of the research scene
• Review of activities, actors, players
• suppliers
• projects
• researchers
• Review of quasi-quantitative measures including
• projects-per-country (based on hltcentral
  database and other sources for pre-FP4/5
  programmes)
• experts/individuals with HLT focus per country
  (based on CDB and ELSNET contacts list)
• citations per country - from proceedings of
  major HLT conferences (EACL, ACL, COLING, MT
  Summit, Applied ACL, TMI, CLAW, Eurospeech,
  ICSLP, LREC)

17
Consolidated Results HLT Benchmark Opportunity
HLT Scorecard
LEADERS
HLT Opportunity
EU-14
HLT Benchmark
18
Key Findings Denmark
19
Key Findings Denmark vs EU average
20
HLT Indexes Denmark
EU Average
HLT RTD Tech Transfer
HLT RTD Investment
HLT Language Breadth
21
Scorecard Comparing Denmark to EU Leaders
22
Denmark Key Findings

• Denmark scores to the EU average on measures of
  robustness in HLT research
• Strong tradition of text and NLP applications,
  and speech research is well represented.
• Active professional translation community,
  including a dedicated machine translation system
  for processing patent documents between English
  and Danish.
• Five major HLT research centres, one of which
  acts as a national centre of excellence for
  language technology.
• Denmark has about seven HLT suppliers, though not
  all of them are dedicated language-technology
  focused companies

23
Denmark Strengths

• Traditionally export-focused country with strong
  multilingual capabilities
• Excellent cross-border facilitator
• Well-trained RD base in language and speech
  technology
• Strong EU project participation, and a leading
  role in regional Nordic language technology
  activities
• Healthy business innovation environment and
  excellent levels of public infrastructure
  readiness
• Basic technology components for Danish now exist,
  and research is carried out on other languages.

24
Denmark Challenges

• Ensuring that the Danish language community can
  benefit from speech and language technologies
  appropriate to Denmarks high degree of readiness
• Industry involvement in HLT research is scarce,
  tech-transfer weak
• Lack of large-scale high technology channels that
  can facilitate the transfer of language and
  speech technologies to market
• Small local market opportunity
• While 95 of the population are Danish speaking,
  the language is highly localised, and not used
  widely elsewhere
• The transfer of first generation language
  technology to the marketplace is still in waiting
  mode

25
Guide to Action
Strong correlation between Opportunity and HLT
Benchmark
World-Class knowledge-based products and services
DE/NE/FI/UK
SV/DK
Role for stronger HLT RTD and tech-transfer to
match opportunities
Boosting HLT research will pay off due to market
potential
HLT Opportunity
IE/IT/AT
EU-14
HLT lagging due to constraints of the environment
and under-investment in RD
FR/BE/ES
Strong exploitation potential tech-transfer
support key
Infrastructure, investment and national policy
support urgently needed
GR/PT
HLT Benchmark
26
General Recommendations

• EU-level
• Collaboration at national level HLT in the ERA
• Structures for visioneering
• Digital Infrastructure for HLT
• Infrastructure Funds for language technology
• Danish national level
• Need for a continued focus on cross-border
  collaboration to ensure the future vitality of
  the Danish HLT community
• within the Nordic Region, and in the wider
  European context.
• Greater effort at transfer opportunities
• high Internet penetration and networked
  educational system
• e-government, education and training may provide
  new opportunities for exploiting language
  technology in an inherently small market.

27
Thanks!

• Rose Lockwood
• Director of Research, Bowne Global Solutions
• Cambridge
• 44 1223 350 340
• rose.lockwood_at_equipe.co.uk