Multilingual Information Access for

1
Culture OnLine, 5-7.6
Multilingual Information Access for Cultural
Heritage Carol Peters ISTI-CNR, Italy
2
Outline

• What is MLIA/CLIR
• What is the State-of-the-Art?
• Where are the Problems?
• MLIA for Cultural Heritage
• The MultiMatch project

3
Europes Linguistic Diversity
4
There are 6,800 known languages spoken in 200
countries 2,261 have writing systems (the others
are only spoken)
Just 300 have some kind of language processing
tools
5
Multilinguality in the Information Society

• Acquisition and dissemination of information in
  digital form over language boundaries
• Web as platform for knowledge dissemination and
  acquisition
• Content available and needs to be accessed in
  many languages

• Information providers and seekers should have
  equal opportunities
• Preservation of national languages

6
Multilinguality in the Information Society

• Information providers and seekers should have
  equal opportunities
• Preservation of national languages

7
Trend Toward Multilingual Web
Source http//global-reach.biz/globstats/evol.htm
l
8
i2010 Digital Libraries(Europeana)

• Multilinguality is a key issue and impacts not
  only on
• Online accessibility of Europes cultural
  heritage (a common multilingual access point)
• but also
• Digitisation
• Preservation and storage

9
Online Access

• Increasing pressure for access to information
  without language or cultural barriers
• Find information in foreign languages
• Read and interpret that information
• Merge with information in other languages
• Need for Multilingual Information Access

10
What is MLIA?

• MLIA related research regards the storage,
  access, retrieval and presentation of information
  in any of the world's languages.
• Two main areas of interest
• multiple language access, browsing, display
• cross-language information discovery and
  retrieval

11
Multi-Language Access, Browsing, Display

• The enabling technology
• character encoding
• specific requirements of particular languages and
  scripts
• internationalization localization

12
Cross-Language Information Retrieval

• Crossing the language barrier
• querying of multilingual collection in one
  language against documents in many other
  languages
• filtering, selecting, ranking retrieved documents
• presenting retrieved information in an
  interpretable and exploitable fashion

13
The Problem
14
(No Transcript)
15
(No Transcript)
16
CLIR methods

• How is it done?
• Pre-process index both documents and queries
  generally using language dependent techniques
  (tokenisation, stopwords, stemming, morphological
  analysis, decompounding, etc.)
• Translate queries or documents (or both)
• Translation resources
• Machine Translation (MT)
• Parallel/comparable corpora
• Bilingual Dictionaries
• Multilingual Thesauri
• Conceptual Interlingua
• Find relevant documents in target collection(s)
  present results

17
CLIR for Multimedia

• Retrieval from a mixed media collection is non-
  trivial problem
• Different media processed in different ways and
  suffer from different kinds of indexing errors
• spoken documents indexed using speech recognition
• handwritten documents indexed using OCR
• images indexed using significant features
• Need for complex integration of multiple
  technologies

18
Main CLIR Difficulties (I)

• Language identification
• Morphology inflection, derivation, compounding,
  
• OOV terms, e.g. proper names, terminology
• Multi-word concepts, e.g. phrases and idioms
• Ambiguity, e.g. polysemy
• Handling many languages L1 -gt Ln
• Merging results from different sources / media
• Presenting the results in useful fashion

19
Main CLIR Difficulties (II)

• MLIA systems need clever pre-processing of target
  collections (e.g. semantic analysis,
  classification, information extraction)
• MLIA systems need intelligent post-processing of
  results merging/ summarization / translation
• MLIA systems need well-developed resources
• Language Processing Tools
• Language Resources
• Resources are expensive to acquire, maintain,
  update

20
Involving the user

• Interactive CLIR systems can help users locate
  and identify relevant foreign-language documents
• Formulate and translate the query (e.g. entering
  diacritics, selecting translation alternatives)
• Query re-formulation (e.g. selecting query
  expansion terms)
• Browsing/navigating results (e.g. translating
  metadata)
• Identifying relevant documents (e.g. summarising
  and translating results)

21
Implementing MLIA is Complex

• Multilingual Portals
• How many languages / how many levels should be
  multilingual / how to handle updates /linguistic
  and cultural dependent issues
• Monolingual Search for Multiple Languages
• encoding and representation issues / language
  identification / indexing issues (stop words,
  stemmers, morphological analysers, named entity
  recognition, ..)
• Cross-Language Search
• translation resources (dictionaries, corpora, MT
  systems)
• Presentation of Results
• in form interpretable and exploitable by user

22
MLIA for Cultural Heritage

• The main problems are the same
• BUT
• Need for fine tuning with respect to
• Specific terminology
• Specific media involved
• Specific user profile

Case Study The MultiMatch project
23
MultiMatch The Initial Idea

• Problem
• The Web contains a wealth of fragmented CH
  information but users are left to discover,
  interpret and aggregate it.
• Objectives
• Develop a search engine that provides targeted,
  enriched access to heterogeneous CH objects
• across all media types and language boundaries
• supporting various user classes
• with aggregate views on complex task scenarios
• Assist CH institutions to raise visibility and
  disseminate content

24
acquisition
crawling
25
Main Research Challenges

• From document to complex objects retrieval
• Focused crawling for acquisition of CH-related
  information from heterogeneous MM resources
• CH concept and relation extraction using
  information extraction and text mining techniques
• Multimedia search and mixed media search
• Multilingual management with support for query
  formulation, cross-language retrieval and
  summarization
• Integration and representation of related objects
• Presentation of aggregated search results
• User support (e.g. search history, annotation
  facilities, personalized presentation of results,
  etc.)

26
Multilingual Support

• Provide system with monolingual and multilingual
  search functionalities (initially four languages,
  extended to six)
• Provide effective translation strategies e.g.
  multilingual dictionaries, machine translation,
  thesaurus term expansion for domain specific
  content
• Multilingual query expansion
• Dynamic summarization of results over languages

27
Multilingual Indexing

• Six separate monolingual index files
• Dutch, English, German, Italian, Polish, Spanish
• One multilingual (English) index file
• Translate all incoming docs to English
• Store all in a single index
• Translate incoming queries to English
• Submit to English index

28
Query Translation Methods

• Machine translation
• Commerical MT system
• Dictionary-based translation
• General Machine-readable dictionaries
• CH Domain-specific lexicons

29
Machine Translation

• WorldLingo commercial machine translation system
  used under licence
• Supports all language pairs for the six selected
  languages
• Easy to use and integrate into prototype
• Good well documented API

30
Dictionary-based Query Translation

• Word-by-word (phrase-by-phrase) translation via
  bilingual dictionary look-up
• Translation resources
• General translation lexicon
• FREELANG dictionary www.freelang.net
• Universal dictionary www.dicts.info
• Domain-specific translation lexicon
• Wikipedia www.wikipedia.org

31
Domain-specific Lexicon Construction

• Multilingual Wikipedia
• A Wikipedia page written in one language often
  contains hyperlinks to counterparts in other
  languages
• For example

32
Domain-specific Translation Example
33
(No Transcript)
34
Query expansion

• Thesaurus expansion
• Terms from Eurowordnet
• Relevance Feedback
• Terms from user-selected relevant documents
• Blind feedback
• Terms added from the system top-tanked and
  assumed-relevant documents

35
Cross Media Data Aggregation

• Enhance retrieval techniques by better
  integration of modalities
• Combined visual plus text search (free text
  metadata) machine-learning based fusion of
  mono-modla search results

36
User Interaction and Interface Design

• User-centred design process
• Evaluate and refine the interface based on
  empirical evaluation and usability testing
• Interface supporting multilingual and multimedia
  search
• Default search on all types of content
• Specialized search on metadata fields and on
  different media
• Use of semantic structures for search and browse

37
(No Transcript)
38
Cross-Language Search
39
The Consortium

• Academia
• Istituto di Scienza e Tecnologie
  dellInformazione
• University of Sheffield
• Dublin City University
• University of Amsterdam
• University of Geneva
• Universidad Nacional de Educación a Distancia
• Industry
• OCLC
• WIND Telecomunicazioni S.p.A.
• Cultural Heritage
• Fratelli Alinari Istituto Edizioni Artistiche SpA
  
• Netherlands Institute for Sound and Vision
• University of Alicante Biblioteca Virtual
  Miguel de Cervantes
• Coordinator Pasquale Savino savino_at_isti.cnr.it

40
More Information

• Website with Online Demo
• www.multimatch.eu
• MultiMatch Final User Workshop, 25 October,
  Florence, Italy
• Everyone is welcome
• Contact Sam Minelli, User Groups Coordinator
  sam_at_alinari.it