Page 1
Software Engineering for Business Information Systems (sebis)
Department of Informatics
Technische Universität München, Germany
wwwmatthes.in.tum.de
Applying lexical knowledge to improve search
quality for a German legal information
database Master thesis final presentationLaura Altamirano Sainz - May 4th, 2014
Page 2
Administration matters
• Time:
October 15th, 2014 to April 15th, 2015
• Supervisor:
Prof. Dr. Florian Matthes
• Advisor:
Bernhard Waltl
© sebisLaura Altamirano Sainz 2
Page 3
Agenda
1. Motivation
2. Research questions
3. Research method
4. Demonstration
5. Evaluation
6. Conclusions
© sebisLaura Altamirano Sainz 3
Page 4
Motivation
• Integration of lexical information in legal searches
Related work:
• Ontologies integrated in the foreground of the systems
• Interaction between the users and the lexical knowledge
• Other areas integrate lexical knowledge for searches: Biology
• Lexical knowledge integrated in the background
of the systems
• Query expansion
© sebisLaura Altamirano Sainz 4
Page 5
Research questions
How can search quality be improved by lexicalinformation for a legal database?
What mechanisms and methods are common in legaldatabases?
Which search mechanisms and methods can beenhanced by lexical information and how?
How can a implementation for a support searchmechanism integrated with lexical knowledge look like?
© sebisLaura Altamirano Sainz 5
Page 6
Research method
• Needs assessment
• Interview with 6 experts in the legal domain
• Results:
• Lexical relations
• Hyponyms
• Troponyms
• Siblings / related terms
• Derivationally related terms
• Search support mechanisms in legal databases analysis
• Comparison between 5 legal databases
• Search support mechanisms categories
• Query formulation / specification
• Query reformulation
• Integration of navigation of the results with search
© sebisLaura Altamirano Sainz 6
“Systems are built to help people work
better. They cannot be built well without
understanding how people work”
(Holtzblatt & Beyer, 1997)
Page 7
Research method
Search system
• Integration of the lexical knowledge as a search support mechanism in a
search system
• GermaNet
• Lexical database for German
• Query expansion / refinement suggestions
• Query expansion with hypernyms
• Query refinement with hyponyms
© sebisLaura Altamirano Sainz 7
Organism
Animal
Marine
creature
Larva
Plant
Page 8
Research method
System architecture
© sebisLaura Altamirano Sainz 8
Page 9
DEMONSTRATION
© sebisLaura Altamirano Sainz 9
Page 10
Evaluation
• Limitations
• Word sense disambiguation
• Search context
• Lexical database
• Evaluation results – Expert interview
© sebisLaura Altamirano Sainz 10
System advantages
• Integration of lexical information as a searchsupport mechanism
• More than one lexical relation implemented
• Clean and clear interface
Areas of improvement
• Search context
• Personalization
• Explanatory mechanismfor highlighted words
Page 11
Conclusions
© sebisLaura Altamirano Sainz 11
• Lexical information can improve searches
• Current search support mechanisms can be improved by lexical information
• Law practitioners show interest for this area
• Users are able to interact with the lexical information
Outlook
• Context
• Improving lexical database
• Personalization features
Page 12
Technische Universität München
Department of Informatics
Chair of Software Engineering for
Business Information Systems
Boltzmannstraße 3
85748 Garching bei München
Tel +49.89.289.
Fax +49.89.289.17136
wwwmatthes.in.tum.de
Laura Altamirano Sainz
17124
[email protected]
Thank you for your attention!
Page 13
References
[1] http://www.toddtransportation.com/countdown-to-a-successful-move-the-
moving-checklist/
[2] http://www.horsesforsources.com/cognizant-051411
[3] AngularJS. https://angularjs.org/
[4] Elasticsearch. https://www.joyent.com/public-cloud/benchmarks/elasticsearch
[5] Play framework. https://www.playframework.com/
[6] Germanet. http://www.sfs.uni-tuebingen.de/GermaNet/
[7] Wordnet. http://wordnet.princeton.edu/wordnet/
[8] Java logo. http://onsitepcsolution.com/wp-
content/uploads/2014/08/java_tech.jpg
[9] Bootstrap logo. http://logonoid.com/bootstrap-logo/
[10] http://www.many-roads.com/2014/12/13/mega-search-engine/
© sebisLaura Altamirano Sainz 13
Page 14
Backup slides
• Lexical knowledge
• All what we know about a word
• Relationships with other words
• Ontological categories (Relations, hyponyms, hypernyms, synonyms,…)
• Lexical databases
1. Wordnet
117 000 synsets
2. GermaNet
93 246 synsets
© sebisLaura Altamirano Sainz 14
Organism
Animal
Marine
creature
Larva
Plant
Page 15
Backup slides
© sebisLaura Altamirano Sainz 15
Search supportmechanisms
Query formulation/ specification
Boolean operators
Autocomplete
Queryreformulation
Spellingsuggestions
Query expansion
Integration of navigation of the
results with search
Facetednavigation
Table of contents
Page 16
Backup slides
German legal information
• Number of German laws is increasing
• Frequently revised information
• Relevant information
• To build up cases, for resolutions, etc…
• Style of writing legal documentation
• Standard format
© sebisLaura Altamirano Sainz 16
Page 17
Backup slides
• Main objectives
© sebisLaura Altamirano Sainz 17
• Finding out common search supportmechanisms in legal databases
• Assess the needs of the user
• Integrate the lexical information as a searchsupport mechanim in a system
• The support tool must be intuitive to the user
• Evaluate the system
Page 18
Backup slides
© sebisLaura Altamirano Sainz 18
Search support mechanismsanalysis in 5 legal databases
• JURION
• beck-online
• LexisNexis
• FindLaw
• LEXinform
Select which ones are common
Select which ones can be enhanced by lexical information
• For example:
1. Autocomplete
2. Faceted navigation
Needs assesment
Page 19
Backup slides
Process
© sebisLaura Altamirano Sainz 19
Lexical information analysis
12
3
Lexical information
(Synonyms, hypernyms,
hyponyms, etc…)
4
3, 5
2
Page 20
Backup slides
Lexical relations
© sebisLaura Altamirano Sainz 20
richtig
(right)
zutreffend
(applicable)
korrekt
(correct)
wahr
(true)
recht
(right)regulär (regular)
vorschriftsmäβig (properly)
ordnungsgemäβ (properly)
regelgerecht (regular)
regelgemäβ (regularly)
vorschriftsgemäβ (properly)
angemessen
(appropriate)
adäquat
(adequate)
bewertungsspezifisch
(assessment specific)
klassenübergreifend
(across classes)
GNROOT
hyponyms
hypernyms
falsch
(false)
antonym
richtig
(right)
richtig regelrecht
richtiggehend
(right proper
real)
synset
synset
Page 21
Backup slides
Sequence diagram
© sebisLaura Altamirano Sainz 21