Semantic Annotation and Search of Software Artefacts

1
Semantic Annotation andSearch of Software
Artefacts

• Valentin Tablan
• Kalina Bontcheva
• Danica Damljanovic

2
Some Terminology

• Ontology population given an ontology, populate
  it with instances derived automatically from a
  text.
• Annotation (of text) associating labels to text
  snippets from a larger document.
• Can be linguistic, semantic, etc...
• Semantic annotation labels used in annotation
  are associated with an ontology
• Can also include ontology population, as a side
  effect.

3
Annotation
4
Semantic Annotation
5
Case Study Software Artefacts

• The GATE project
• Open source text mining infrastructure tools.
• Structured information
• XML configuration files
• Unstructured Information
• Software documentation (JavaDocs)?
• User guide
• Project website
• Publications
• Mailing lists.

6
Case Study Software Artefacts

• The problem?
• Information overload 000's of pages.
• Parameters for ANNIE Tokeniser?
• Where to look for the answer?
• The solution?
• Use an ontology as a shared store.
• Populate ontology using semantic annotation.
• Provide search facilities backed by the ontology.

7
The GATE Ontology
8
Ontology Population Structured Data
9
GATE Ontology - populated
10
Ontology PopulationUnstructured Data

• Extract all lexicalisations of ontological
  resources
• Names, labels, string property values.
• Normalise the lexicalisations
• Extract morphological root.
• Break text at underscores.
• Segment CamelCaseNames.
• Build a gazetteer with all the lexicalisations.
• Use the gazetteer to recognise mentions in text.

11
Ontology PopulationUnstructured Data
12
Information Access Conceptual Retrieval

• Can make use of abstractions and generalisations
  powered by ontology back-end.
• Provides retrieval options not available to
  full-text search, e.g.
• Capitals of countries in Asia
• Query language very complex, somewhat similar to
  SQL ? not really suitable for end users.

13
Capitals of countries in Asia (simplified
SeRQL)?

• select c0, p1, c3, p4, i6
• from
• c0 rdftype puppCapitalgt,
• c3 p1 c0,
• c3 rdftype puppCountrygt,
• c3 p4 i6,
• i6 rdftype puppContinentgt
• where
• p1ltpupphasCapitalgt and
• p4ltptopsubRegionOfgt and
• i6ltkimContinent_T.2gt

14
QuestIOQuestion-based Interface to Ontologies

• Natural Language interface for querying knowledge
  bases.
• Easy to use require no training.
• Domain independent.
• Works with short, agrammatical queries
  (Google-like).

15
QuestIO Initialisation

• Vocabulary built automatically from the KB (hence
  domain independent).
• Extract all possible textual descriptions from
  the ontology.
• Normalise for morphology, lack of tokenisation,
  CamelCasing, etc.
• Represent all lexicalisations into a GATE
  gazetteer (long init time, fast run time).

16
Query Construction

• Identify known concepts in the NL query
• Normalise the query for morphology, etc.
• Find concepts by matching lexicalisations from
  the gazetteer.

Capital City
Country
Continent
Continent_T4
Asia
Capitals of countries located in Asia
17
Query Construction (II)?

• Build a SerQL query by finding appropriate
  properties to link the concepts found.
• Build a list of candidate properties based on
  ontology schema (using domain and range
  constraints).
• Rank the properties.

18
Ranking Properties

• We combine three types of scores
• similarity score compare query fragments with
  candidate property names using Levenshtein string
  similarity metric.
• specificity score is based on the subproperty
  relation in the ontology definition.

19
Ranking Properties (II)?

• distance score inferring an implicit specificity
  of a property based on the level of the classes
  that are used as its domain and range.

20
Query Execution

• Execute queries in ranking order until results
  are obtained.

21
Evaluation coverage and correctness

• 36 questions extracted from GATE list
• 22 out of 36 questions were answerable (the
  answer was in the knowledge base)
• 12 correctly answered (54.5)?
• 6 with partially corrected answer (27.3)?
• system failed to create a SeRQL query or created
  a wrong one for 4 questions (18.2)?
• Total score
• 68 correctly answered
• 32 did not answer at all or did not answer
  correctly

22
Evaluation on scalability and portability

• Sizes of the knowledge bases created based on
• GATE ontology http//gate.ac.uk/ns/gate-ontology
  
• Travel ontology http//goodoldai.org.yu/ns/tgprot
  on.owl
• Ontologies have not been customised or changed
  prior using with QuestIO!

23
Evaluation on scalability and portability
24
DEMO

• http//www.gate.ac.uk/questio-client-app/search.js
  p
• Geography ontology
• Continents, countries, cities (capitals only).
• Part of PROTON KIM KB
• Example questions
• Countries in Europe or North America
• Capitals in Asia
• Capitals of countries (located) in Africa
• ...