=Paper=
{{Paper
|id=None
|storemode=property
|title=Proposal for Using NLP Interchange Format for Question Answering in Organizations
|pdfUrl=https://ceur-ws.org/Vol-1004/paper1.pdf
|volume=Vol-1004
|dblpUrl=https://dblp.org/rec/conf/ruleml/Latifi13
}}
==Proposal for Using NLP Interchange Format for Question Answering in Organizations==
https://ceur-ws.org/Vol-1004/paper1.pdf
Proposal for Using NLP Interchange Format for
Question Answering in Organizations
Majid Latifi
Department of Software, Universitat Politècnica de Catalunya – BarcelonaTech(UPC), Barce-
lona, Spain
mlatifi@lsi.upc.edu
Abstract. The growth of technology and sciences has greatly influenced the ar-
ea of management and decision-making procedures, and has dramatically
changed the decision-making processes in different levels, both quantitatively
and qualitatively. Knowledge management plays a vital role in supporting en-
terprise learning, since it facilitates the effective collective intellect of the en-
terprise. Different methods for user-friendly knowledge access have been de-
veloped previously. The most sophisticated ones provide a simple text box for a
query which takes Natural Language (NL) queries as input. Question Answer-
ing (QA) system is playing an important role in current search engine optimiza-
tion. Natural language processing technique is mostly implemented in QA sys-
tem for asking user‟s question and several steps are also followed for conver-
sion of questions to query form for getting an exact answer. Query languages
have complex syntax, requiring a good understanding of the representation
schema, including knowledge of details like namespaces, class and property
names. In this research we proposed an model to implement Conceptual Ques-
tion Answering and Automatic Information Inferences for the enterprise's oper-
ational knowledge management in ontology-based learning organization.
Keywords: Enterprise Ontology, Learning Organization, Question Answer-
ing(QA), Information Inference, NLP.
1 Introduction
Retrieval and extraction processes - for enterprise management and decision-making -
have gained an excessive importance as the mass of data and information stored in
various resources increases. Knowledge is considered a key factor for enterprise pros-
perity at present and future. Knowledge management is an integrated, systematic pro-
cess that applies a suitable combination of information technologies and human coop-
eration in order to identify, manage and share the information capitals. In addition, it
both includes the explicit and implicit knowledge of the staff and it applies various
and extensive methods to retrieve, store and share knowledge in a certain enterprise.
The application of “Semantic Web” technologies to learning processes is receiving
an increasing attention from the perspective of facilitating the selection, delivery and
tailoring of learning experiences. But most of the current approaches are centered on
the final interaction of the learner with the “learning objects” provided for him/her,
neglecting the organizational perspective. From the viewpoint of an organization, the
application of Semantic Web technologies should be motivated by the improvement
of learning-oriented mechanisms, including both cultural and structural aspects, and
considering the ideal of achieving a state of continuous improvement in learning be-
havior. Such an approach to achieving a “semantic learning organization” gives a
complementary perspective to existing “educational Semantic Web” propositions [2].
A main need for the semantic enterprise model is one which extracts and displays the
enterprise semantics.
Most knowledge bases provide facilities for querying through the use of some
formal language such as SPARQL or SeRQL. However, these have a fairly complex
syntax, requiring a good understanding of the data schema and being prone to errors
due to the need to type long and complicated URIs. These languages are homologous
to the use of SQL for interrogating traditional relational databases and should not be
seen as an end user tool[13].
The obvious solution to these problems is to create some additional abstraction
level that provides a user friendly way of generating formal queries. It may be possi-
ble to infer from this information for the machine so that we can carry out the deci-
sion-making and planning procedures in enterprise processes through automatic infer-
ence.
2 Statement of the Problem and Related Work
A basic method to transform an organization into a learning organization is to apply
knowledge management within the organization. By facilitating the process of creat-
ing and sharing knowledge, and through providing positive working environments
and effective rewarding systems, knowledge management accelerates enterprise learn-
ing and helps the enterprise adjust itself to today's rapid changes and hence survive in
pace with these changes[9]. By using ontology, we can identify the meanings related
to a domain, an enterprise or a society or even determine these meanings within dif-
ferent societies in details as desired [3]. In Ontology-based QA system, the
knowledge based data, where the answers are sought, has a structured organization.
The question-answer retrieval of ontology knowledge base provides a convenient way
to obtain knowledge for use, but the natural language need to be mapped to the query
statement of ontology. Accessing structured data such as that encoded in ontologies
and knowledge bases can be done using either syntactically complex formal query
languages or complicated form interfaces that require expensive customization to
each particular application domain.
Probably due to the extraordinary popularity of search engines such as Google,
people have come to prefer search interfaces which offer a single text input field
where they describe their information need and the system does the required work to
find relevant results. While employing this kind of interface is straightforward for full
text search systems, using it for conceptual search requires an extra step that converts
the user's query into semantic restrictions like those expressed in formal search lan-
guages. Following are discussed some examples of such query interfaces.
CLOnE[9], presents a controlled language for ontology editing and a software im-
plementation, based partly on standard NLP tools, for processing that language and
manipulating an ontology. The input sentences are analyzed deterministically and
compositionally, which the software consults in order to interpret the input‟s seman-
tics; this allows the user to learn fewer syntactic structures since some of them can be
used to refer to either classes or instances, for example. A repeated-measures, task-
based evaluation has been carried out in comparison with a well-known ontology
editor.
The Controlled Language for Ontology Editing (CLOnE) allows users to design,
create, and manage information spaces without knowledge of complicated standards
(such as XML1, RDF2 and OWL3) or ontology engineering tools. It was implemented
as a simplified natural language processor that allows the specification of logical data
for semantic knowledge technology purposes in normal language. CLOnE is designed
either to accept input as valid or to reject it and warn the user of his errors; because
the parsing process is deterministic, the usual IE performance measures (precision and
recall) are not relevant.
QACID [10] is based on collection of queries from a given domain which are ana-
lyzed and grouped as clusters and those are manually annotated using SPARQL que-
ries. Each query is considered as bag of words, mapping between words in NL queries
into KB by using string distance metrics. SPARQL generator replaces the ontology
with instances mapped for original NL query. It is domain specific and the perfor-
mance depends on the types of questions collected in domain.
ONLI (Ontology Natural Language Interaction) [11] is a natural language question
answering system used as front-end to the RACER reasoner and to nRQL, RACER's
query language. ONLI assumes that the user is familiar with the ontology domain and
works by transforming the user's natural language queries into nRQL. No details are
provided regarding the effort required for re-purposing the system.
QAAL [12] surveys different types of question answering system based on ontol-
ogy and semantic web model with different query format. For comparison, the types
of input, query processing method, input and output format of each system and the
performance metrics with its limitations was analyzed and discussed. There are basi-
cally three types of question classification methods available. Those are machine
learning approaches, knowledge based approach and template based approach. In
QAAL system is used template based approach for fast retrieval of answer. If the
question is already asked in that system, the retrieval takes place within question tem-
plate table, otherwise matching is performed using Graph Matching Algorithm and
uses Spread Activation Algorithm for query matching with the ontology.
1
eXtensible Markup Language
2
Resource Description Framework
3
Web Ontology Language
QuestIO [13] system has a natural language interface for accessing structured in-
formation, that is domain independent and easy to use without training. It brings the
simplicity of Google's search interface to conceptual retrieval by automatically con-
verting short conceptual queries into formal ones, which can then be executed against
any semantic repository. The QuestIO application is open-domain (or customizable to
new domains with very little cost), with the vocabulary not being predefined but ra-
ther automatically derived from the data existing in the knowledge base. The system
works by converting NL queries into formal queries in SeRQL. It was developed
especially to be robust with regard to language ambiguities, incomplete or syntactical-
ly ill-formed queries, by harnessing the structure of ontologies, fuzzy string matching,
and ontology-motivated similarity metrics. It works by leveraging the lexical infor-
mation already present in the existing ontologies in the form of labels, comment and
property values.
PANTO [14] model a Portable nAtural laNguage inTerface to Ontologies which
accepts input as natural language form and the output is in SPARQL query. It is based
on triple model in which parse tree is constructed for the data model using the off-the-
shelf Standford parser. Logic rules are applied for natural language queries as nega-
tion, comparative and superlative form. For mapping WordNet and String metric al-
gorithms are used. The parse tree forms the intermediate representation as Query Tri-
ples Form. Then PANTO converts Query Triples form into OntoTriples form which
are represented as entities in ontology.
OntoTriples are finally interpreted as SPARQL form. The performance of PANTO
is analyzed by using FMeasure type. At the maximum 88.05% Precision is achieved
for Geography domain with tested queries. So this system helps bridge the gap be-
tween the real world users with the semantic web based on logic model.
AquaLog [15] is capable of learning the user's jargon in order to improve his expe-
rience by the time. Their learning mechanism is good in a way that it uses ontology
reasoning to learn more generic patterns, which could then be reused for the questions
with similar context. In this system two major models are used as Linguistic Compo-
nent which is used to convert the NL questions into Query-triple format and Relation
Similarity Service (RSS) which takes Query Triple form into Onto-Triple form. The
data model is triple like {Subject, Predicate, Object} type. The Performance is based
on Precision, Recall and also failure types are referred separately. At average 63.5 %
of successive answers are retrieved from ontology with closed domain environment.
QASYO [16] is a sentence level question-answering system that integrates natural
language processing, ontologies and information retrieval technologies in a unified
framework. It accepts queries expressed in natural language and YAGO [18] ontology
as inputs and provides answers drawn from the available semantic markup which
combining several powerful techniques in a novel way to make sense of NL queries
and to map them to semantic markup. Semantic analysis of questions is performed in
order to extract keywords used in the retrieval queries and to detect the expected an-
swer type. In the QASYO model there are 4 phases: question classifier, linguistic
component, query generator and query processor which characterizing it´s architec-
ture as a waterfall model. One NL query gets translated into a set of intermediate,
triple-based representations, query-triples, and then these are translated into ontology-
compatible triples.
The whole QA process is composed of two consecutive phases: question analysis and
answer retrieval. This model requires both an evaluation of its query answering abil-
ity. Another extension is to provide information about the nature and complexity of
the possible changes required for the ontology and the linguistic component.
Knowledge management system includes methods for obtaining or gathering in-
formation, organizing, distributing and sharing information among the staff in an
organization. In this research, the potential role of the Semantic Web Technology as a
driver for advanced learning organizations and Question Answering system is focused
on providing access to the information stored in a KB by means of natural language
queries.
3 Research Objectives
The current research is aimed to show that using standard NLP tools, ontology and
informal to formal semantic query model proposed in the current research can estab-
lish a relationship among various sectors including duties, activities, resources and
information structure of a certain enterprise so that managerial requirements can be
desirably met through semantic modeling. As a result, we may have a better chance of
using this information for the managers and the users through conceptual queries on
the information system of the enterprise. In attention to the actual state of semantic
web technology and NLP, the recommended path for organizations that are commit-
ted to the view of a learning organization is that of first addressing infrastructural
elements. Such infrastructures can be considered as the study and provision of the
ontologies for each aspect of the semantic learning organization. Therefore, how can
we improve knowledge management in enterprises through an appropriate selection
based on ontology?. Also, how can we respond to the managerial requirements of the
enterprises from simple decisions to strategic ones and how can we perform automatic
extraction of the information?. Consequently, the following objectives are followed in
parallel with works carried out previously:
1. Conceptual framework for the notion of a semantic learning organization with
using semantic search model instead of using normal keyword search model is
provided.
2. Designing and presenting a method to translate user´s semantic queries into well-
defined queries using the results of NLP Interchange Format (NIF) to answer the
semantic questions.
3. The necessity to be robust and ability to deal with all kinds of input including
ungrammatical text, sentence fragments, short queries, etc.
4 Scope of Activity
4.1 Learning Organization Ontology
The existing organizational architecture is faced with a semantic shortage between
humans and systems for having a precise and general understanding of them, which in
turn causes communication problems between humans and systems or vice versa.
These problems prohibit the materialization of the organizations in an assembled and
concordant form with other organizations [7]. Our goal is not only to design a „con-
ceptual‟ ontology model but also to implement it as an operational ontology. This
approach, mainly favored by the research community, may be beneficial for integrat-
ing the domain ontology model with an inference engine for the language. Trying to
match the users´ requests by providing appropriate formal commands is faced with
restrictions, and thus making such semantic query by programmers is demanding,
time consuming and inefficient.
4.2 Translating Natural Language Questions into Well-defined Queries
There is technically too complicated to represent and comprehend the domain for a
domain expert who has little knowledge in the well-defined queries. More important-
ly, from a practical point of view, there is no publicly known robust engine to manage
a large KB with practical performance. On the other hand, we should increase the
machines' capability in understanding the organizational structure (Intelligent-
making). To this end, having analyzed the existing concepts in the scope of
knowledge management of the learning organizations, we reckon the significance of
the information capitals of an enterprise through an ontology-based method. Answer-
ing to semantic questions will help increase the capability of learning organizations.
The growing interest in Semantic Web applications and need to translate natural
language question into a machine-readable format create many uses for such applica-
tions. It is implemented as a natural language processor that allows the specification
of logical data for semantic knowledge technology purposes in normal language, but
with high accuracy and reliability. The components are based on NLP Interchange
Format(NIF) with using statistical machine translation method.
5 Modelling of Conceptual Question Answering Method in
Learning Organizations
We designed an initial model to implement Conceptual Question Answering and Au-
tomatic Information Inferences for the enterprise's operational knowledge manage-
ment in learning organization. To achieve this goal, we evaluate the SPARQL and
SeRQL languages for semantic search. In [5] is shown an application of SPARQL-DL
query language to natural language processing, more especially as a rule engine to use
within a semantic parser. As shown, the use of such formalism for this task has sever-
al advantages including the straightforward conversion of a typed dependency graph
in an ontology. In Fig. 1, the general model of our proposed system is represented. It
has the following modules.
Query Parsing and Analysis: In this phase, the analytical operation of the ques-
tion is found out. This Analysis is responsible for Natural Language Processing
(NLP). It is a technique to identify the type of a question, type of an answer, sub-
ject, verb, noun, phrases and adjectives from the question. Tokens are separated
from the question and the meaning is analyzed and the reformulation of question is
sent to the next stage. The input is concerted into Natural Language and is imple-
mented using word segmentation algorithm. In word segmentation algorithm the
input query from the user is divided as keywords which is further subdivided and
searched in knowledge base to get correct answers.
Fig. 1. Suggested Model for Semantic Question Answering
Integration between Semantic Web and NLP: The tools available nowadays for
Natural Language Processing can achieve very good results on many complex
tasks such as the parsing of a sentence. An NLP Interchange Format for integrating
NLP applications is presented by [6]. NIF addresses weaknesses of centralized in-
tegration approaches by defining an ontology-based and linked-data aware text an-
notation scheme. The NLP Interchange Format (NIF) is an RDF/OWL-based for-
mat that aims to achieve interoperability between NLP tools, language resources
and annotations. The core of NIF consists of a vocabulary, which allows to repre-
sent strings as RDF resources. By being directly based on RDF, Linked Data and
ontologies, NIF also comprises crucial features such as annotation type inheritance
and alternative annotations, which are cumbersome to implement or not available
in other NLP frameworks [17].
Regenerating of Semantic Query: According to the user‟s choice, the formula-
tion of query is generated with the help of YAGO[18] and WordNet [19] which are
implemented as semantic matching model.
Semantic Search: At next stage, the Search is carried out using Conceptual Graph
Matching algorithm which is the best technique. All the sentences in repository are
framed as conceptual graph and the given question is also framed as conceptual
graph. The matching of question CG with given CG are checked out using CG
matching algorithms and the result us displayed at front-end of the our system.
Graph patterns are important concepts in semantic search. RDF model is organized
and graph patterns are used to formulate and encode constraint queries for locating
sub graph in RDF network.
Graph Matching in Ontology: Conceptual Graph acts as an intermediate lan-
guage for mapping natural language questions and assertions to a relational data-
base. Conceptual Graph (CG) contains concept, concept relation and argument. It
is a graph which represents logic based on semantic model of artificial intelligence
and existential graphs. Resource Description Framework (RDF) is a framework
which contains triple syntax to express annotations as subject, predicate and object.
Information resources are commonly represented as uniform Resource Identifiers
(URIs). URIs are described by RDF. RDF triples are visualized as directed labeled
graph in which subject; objects are represented as nodes and predicates as arcs.
Searching Ontology Nodes: Semantic Search Algorithm is based on Conceptual
Graph form of user query and domain ontology. In [8] Spread Activation is a
method for searching the nodes in ontology as in semantic manner. It exploits rela-
tions between nodes in ontology. Nodes may be terms, class, object etc. Relations
are labeled directed or weighted manner. SA algorithm creates initial nodes that are
related to the content of the user‟s query and assign weights to them. After that,
nodes will activate with different nodes on ontology by some rules.
Template based Approach: There are basically three types of question classifica-
tion methods are available. Those are machine learning approaches, knowledge
based approach and template based approach. In this research we use template
based approach for fast retrieval of answer. If the question is already asked in that
system, the retrieval get from question template table form, otherwise matching is
performed using matching algorithm.
Answer Retrieval with Entailment Engine: This part of the system is based on
an entailment engine. This module uses entailment techniques to infer semantic
deductions between a users´query collections and the SPARQL query collections
included in the formulation of user semantic query previously obtained. This pro-
cess allows the system to associate new incoming queries with their corresponding
SPARQL expressions in order to retrieve the answer sought from the RDF data-
base.
6 Conclusions
The main undertaking of the current contribution is to present ongoing work in facili-
tating learning organizations and their use of ontology-based tools by striving to
translate natural language queries into well-defined queries and retrieving exact an-
swers, which in turn can be executed in the framework presented here . A model was
introduced to automatically convert semantic query to formal query in a bid to pro-
vide answers for conceptual question and to infer information from organizational
knowledge base.
Answers are retrieved from ontology using semantic search approach interopera-
bility for NIF components, web services and question-to-query algorithm is evaluated
in our system for analyzing performance evaluation. Finally performance of question
answering system of getting exact result can be improved by using semantic search
methodology to retrieve optimum answers from organizational ontology model.
Acknowledgments
I would like to thank our software department (LSI) from KEMLG research group in
Polytechnic University of Catalonia (UPC). Especially, I would like to thank Dr. Mi-
quel Sànchez-Marrè for his helpful comments and guidance. I acknowledge the finan-
cial support of the Generalitat de Catalunya through the AGAUR agency for Consoli-
dated Research Groups. This support (2009SGR 1365) was granted to the Knowledge
Engineering & Machine Learning group (KEMLG).
References
1. Latifi, M., Khotanlou, H., Latifi, H.: An Efficient Approach Based On Ontology to Opti-
mize the Organizational Knowledge Base Management for Advanced Queries Service. In:
3rd IEEE International Conference on Communication Software Networks(ICCSN),
ISBN: 978-1-61284-485-5, pp. 269 – 273 (2011)
2. Sicilia, M., Lytras, M.: The Semantic Learning Organization. In: The Learning Organiza-
tion, Vol. 12 Iss: 5, pp.402 – 410 (2005)
3. Daconta, M., C., Smith, K., T., Obrst, L., J.: The Semantic Web: A Guide to the Future of
XML, Web Services, and Knowledge Management. John Wiley & Sons, USA (2003)
4. Aggestam, L.: Learning Organization Or Knowledge Management: Which Came First,
The Chicken Or The Egg?. In: Information Technology and Control, vol 35, No.3 (2006)
5. Vitucci, N., Arrigoni Neri, M., Tedesco, R., Gini, G.: Semanticizing syntactic patterns in
NLP processing using SPARQL-DL queries. Politecnico di Milano, Dipartimento di Elet-
tronica e Informazione Via Ponzio 34/5, 20133 Milano, Italy (2012)
6. Hellmann, S., Lehmann, J., Auer, S.: NIF: An ontology-based and linked-data-aware NLP
Interchange Format. http:// svn.aksw.org (2012)
7. Kang, D., Lee, J., Choi, S., Kim, K.: An ontology-based Enterprise Architecture, Expert
Systems with Applications, pp.1456-1464 (2010)
8. Suchal, J., Caching spreading activation search. Slovak University of Technology(2007)
9. Funk, Adam, et al.: CLOnE: Controlled language for ontology editing. The Semantic Web,
Springer Berlin Heidelberg, pp.142-155 (2007)
10. Fernandez, O., R. Izquierdo, S. Ferrandez and J.L. Vicedo, Addressing ontology-based
question answering with collections of user queries. Inform. Proces. Manage., 45: 175-188.
DOI: 10.1016/j.ipm. (2008)
11. Shamima Mithun, Leila Kosseim, V.H.: Resolving quantifier and number restriction to
question owl ontologies. In: Proceedings of The First International Workshop on Question
Answering (QA2007), Xian, China (2007)
12. Kalaivani, S., and K. Duraiswamy, Comparison of Question Answering Systems Based on
Ontology and Semantic Web in Different Environment. Journal of Computer Science 8.9,
pp: 1407-1413 (2012)
13. Tablan, V., Damljanovic, D, Bontcheva, K, A Natural Language Query Interface to Struc-
tured Information, Springer-Verlog Berlin Heidelburg, ESWC 2008, pp. 361-375 (2008)
14. Wang, C., M. Xiong, Q. Zhou and Y. Yu, PANTO: A portable natural language interface
to ontologies. Proceedings of the 4th European Semantic Web Conference, (ESWC‟ 07),
Publication post of DBLP, pp: 473-487(2007)
15. Lopez, V., Motta, E.: Ontology driven question answering in AquaLog. In: NLDB 2004
(9th International Conference on Applications of Natural Language to Information Sys-
tems), Manchester, UK (2004)
16. Moussa, Abdullah M., and Rehab F. Abdel-Kader,: QASYO: A Question Answering Sys-
tem for YAGO Ontology. International Journal of Database Theory and Application 4.2
(2011)
17. Schierle, M.: Language Engineering for Information Extraction. Phd thesis, University at
Leipzig, Leipzig (2011)
18. F. M. Suchanek, G. Kasneci and G.Weikum,: YAGO: A Core of Semantic Knowledge
Unifying WordNet and Wikipedia. In Proceedings of 16th International World Wide Web
Conference (IW3C2), pp. 697-706 (2007)
19. Miller, George A.: WordNet: a lexical database for English. Communications of the
ACM 38.11, pp. 39-41 (1995)