ConNeKTion: A Tool for Exploiting Conceptual Graphs Automatically Learned from Text

Università degli studi di Bari “Aldo Moro”
Dipartimento di Informatica

ConNeKTion: A Tool for Exploiting Conceptual
Graphs Automatically Learned from Text
F. Leuzzi, S. Ferilli, F. Rotella
L.A.C.A.M. {fabio.leuzzi, stefano.ferilli, fulvio.rotella}@uniba.it
http://lacam.di.uniba.it

9th Italian Research Conference on Digital Libraries
Università la Sapienza - Rome, Italy
January 31 - February 1, 2013

Overview

● Introduction & Objectives
● Tool overview
● Knowledge Representation Formalism
● Relevant concepts
● Information Retrieval
● Reasoning by Association
● Exploiting Tool
● Conclusions & Future Works

ConNeKTion: A Tool for Exploiting Conceptual Graphs Automatically Learned from Text
F. Leuzzi, S. Ferilli, F. Rotella 2

Introduction
Some repositories leave the responsibility of quality to the authors.

+
Anybody can produce and distribute documents.

=
Possible low average quality of the repository contents.

The study, understanding and exploitation of the content of a digital library,
with the aim to easily explore the semantic content of huge amounts of text.


Introduction
Possible solution:
● Natural Language Processing systems
● Provide the grammatical structures contained in text
● Knowledge Representation formalisms
● Semantic networks
● Graph learning techniques
● To obtain a semantic network starting from the text
● In order to satisfy the information needs, the knowledge base
can be exploited:
● To make summarizations
● To reason with it
● ...


Objectives

Improving fruition of a DL

● Use of a tool providing advanced functionalities
● Mixed strategy for relevant concept recognition
● Semantic approach to information retrieval
● Automatic inference over the acquired knowledge
● Reasoning by association


Tool overview


Knowledge representation
formalism
Only subject, verb and complement have been considered.
● Subjects and complements → concepts
● Verbs → relations between them

subject, subject,
verb, complement
complement

The frequency of arcs between the concepts in positive and negative
sentences has been taken into account.
● Enrich the representation formalism
● Give robustness to our solution through a statistical approach

Relevant Concepts

● Relevant nodes are sought in the graph

● Mixed strategy
● Semantic network structure
● EM clustering provided by Weka
● Keyword Extraction
● Quantitative approach based on co-occurrences
● Qualitative approach exploiting WordNet
● Psychological approach based on principles of an effective
presentation
● Components empirically weighted


Information Retrieval
● Word Sense Disambiguation
● One Domain per Discourse assumption: many uses of a word in a
coherent portion of text tend to share the same domain
● Prevalent domain individuation
● Extraction of all synsets for each term
● Extraction of all domains for each synset
● Choice of prevalent domain synset
● Pairwise Complete Link Agglomerative Clustering
● Each synset generates a singleton cluster
● For each pair of clusters
● If the complete link property holds
● Merge the involved clusters


Information Retrieval
● Multi-strategy Similarity Measure on WordNet
● 3 components summed and normalized in ]0,1[
● depth (ancestors)
● breadth (direct neighbors)
● breadth (inverse neighbors)
● Document Partitioning
● For each document
● Each synset votes for a cluster
● User Query Processing
● Brute force WSD to find the best synsets combination
● Best combination used to return a ranked list of clusters
● Each cluster has a list of related documents obtained by the Document
Partitioning phase


Reasoning ‘by association’
Breadth-First Search
Given two nodes (concepts), a Breadth-First Search starts from
both nodes, the former searches the latter's frontier and vice
versa, until the two frontiers meet. Then the path is restored
going backward to the roots in both directions.

We also provide the number of positive/negative instances, and
the corresponding ratios over the total to help understanding
different gradations (permitted, prohibited, typical, rare, etc.) of
actions between two objects.


Breadth-First Search
The table below shows a sample of possible outcomes.
E.g., an interpretation of case 1 can be:

“the young looks television that talks about (and criticizes) facebook,
because it typically does not help (rather distracts) schoolwork”.


Probabilistic approach
Real world data are typically noisy
and uncertain → need for strategies
that soften the classical rigid logical
reasoning

Defined a formalism based on ProbLog language: pi :: fi
● fi : ground literal of the form link (subject, verb, complement)
● pi : ratio between the
sum of all examples for
which fi holds and the sum
of all possible links between
subject and complement


ConNeKTion


Conclusions

ConNeKTion allows to learn conceptual graphs from plain text and to
enrich them by finding concept generalizations.

The resulting graph can be used for several purposes:
● finding relationships between concepts (if any)
● filtering the concepts from a particular perspective
● relevant concepts recognition and information retrieval

A suitable control panel is provided for the user to comfortably carry out
these activities.


Future Works
We plan to improve the natural language text pre-processing using anaphora
resolution in order to replace, where possible, pronouns with the explicit concept
they express.

All functionalities have parameters set empirically. A criteria for automatical
setting of suitable parameters is needed.

The preseted functionalities are based on the exploitation of WordNet. A strategy
to make the operators WordNet free can be desirable.

We also wish to extend the reasoning operators by adding an argumentation
operator, that could exploit probabilistic weights, intended as a rate of reliability,
to provide support or attack to a given statement.


ConNeKTion: A Tool for Exploiting Conceptual Graphs Automatically Learned from Text

More Related Content

What's hot (11)

Viewers also liked (8)

Similar to ConNeKTion: A Tool for Exploiting Conceptual Graphs Automatically Learned from Text (20)

ConNeKTion: A Tool for Exploiting Conceptual Graphs Automatically Learned from Text