French Language Interface for Xml Databases
Hanane Bais and Mustapha Machkour
Team of Engineering of Information Systems, Information Systems and Vision Laboratory Faculty of Sciences, Ibn Zohr
University Agadir, Morocco
Keywords: xml database, french language, natural language processing, context-free grammar, machine learning
Abstract: To extract information from a database, it is necessary to formulate queries in terms of the language used by
the database, such as Structured Query Language (SQL). However, non-expert users who are not computer
skills are unable to write such requests. To address this problem, several solutions are proposed. One of
these solutions is to use the natural language that represents the ordinary means of obtaining information.
Yet, without any help, computers cannot understand this language. This is why it is very important to
develop an intelligent interface capable of translating users queries from natural language to queries in
database query language. In this paper, we present a generic French language interface for the XML
database. In this paper, we present a generic French language interface for the XML database. The proposed
interface translates French language query potentially after reformulation, into an XPath expression that can
be evaluated against an XML database. The advantage of this system is that it functions independently of the
database domain and automatically improve its knowledge base through experience.
1 INTRODUCTION
The Natural Language Interface for Databases
(NLIDB) is an intelligent and flexible system
capable of translating a natural language query into a
query in database query language [Androutsopoulos
et al., 1995]. NLIDB can be considered as one of the
fundamental subjects of artificial intelligence and
databases. Traditionally, people are used to using
forms to access information stored in databases.
Nevertheless, their expectations depend strongly on
the capabilities of these forms. However, NLIDB
provides powerful improvements in the use of this
information. Because it offers a large number of
users a simple, uniform and unlimited access to the
data. Success in this area of research due partly to
the benefits that may come from NLIDB [Bais et al.,
2016].
Access to information stored in a database by
natural language is a very important and motivating
method. Numerous research prototypes have been
proposed to arrive to process the user's queries in
natural language. Most of these works have been
designed to interface Prolog and relational
databases. For XML database, there is one
contribution proposed to extract data from XML
databases using the English language.
Our proposed system functions with XML
databases. It interprets the user’s queries in the
French Language Query (FLQ) to an XPath
command, and it retrieves the suitable answer from
the XML database. The system has the ability to
process a very important number of FLQ.
The remainder of the paper is organized as
follows Section 2 gives an overview of existing
work, showing their advantages and limitations.
Section 3 presents a brief description of the
proposed system and the architecture of the system.
Section 4 reports the experimental results. Finally,
Section 5 presents the conclusions and some
possible extensions of this work.
2 RELATED WORKS
Over the past fifty years, many attempts have been
made to create an intelligent interface in natural
language. There has been a great deal of research
presenting the theories for the implementation of
NLIDB. The first attempts are as old as any other
search area in NLP. In fact, NLIDB has been one of
the most important successes of NLP since its
Bais, H. and Machkour, M.
French Language Interface for Xml Databases.
DOI: 10.5220/0009772200730081
In Proceedings of the 1st International Conference of Computer Science and Renewable Energies (ICCSRE 2018), pages 73-81
ISBN: 978-989-758-431-2
Copyright
c
2020 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
73
inception. The first NLIDB appeared in the late
sixties and early seventies [Waltz., 1978]. Some of
this system like LUNAR [Akerkar et al., 2008] and
BASEBALL [Gauri et al., 2010] were designed for a
particular database domain and thus could not be
easily adapted to interface other different databases.
By the end of the seventies, several other NLIDB
had appeared. Some of these systems like LIFFER /
LADDER (Hendrix, 1978) are used semantic
grammars inherent to the database domain. The
problem of these systems was that the semantic
grammar used was adjusted to a certain domain and
that reduce the portability of these systems since the
rewriting of the grammar was necessary whenever
the domain of application is modified. In the mid-
1980s, several NLIDBs appeared. This research area
had become very popular and numerous research
prototypes were put in place with different
approaches to handle user queries in natural
language. The CHAT-80 system [Warren et al.,
1982] is one of the most referenced systems in the
1980s. The system has been implemented in Prolog.
The major problem in this system is that it can be
used only for a specific database domain. The
CHAT-80 code has been widely distributed, and it is
the basis for several other systems such as
MASQUE (Modular Answering System for Queries
in English) [Auxerre et al., 1986]. MASQUE was
also intended for Prolog databases. It answers
questions written in English that are related to
certain areas such as geography and airplanes.
Contributions in the area of the NLIDB are
continuous in the nineties. The majority of these
contributions have focused on the querying of
relational databases by the use of natural language
instead of SQL. However, these systems are always
designed for a specific of domain application.
Androutsopoulos, et al developed an extended
version of the MASQUE system, called MASQUE/
SQ [Androutsopoulos et al., 1993]. This system can
be interfaced with all commercial databases that
support the SQL language. MASQUE/ SQL answers
the questions of the user in English by generating an
SQL code that is executed by the relational DBMS.
After the 1990s, there have been interesting
approaches operate to design NLIDB independently
of the database domain. These systems can
efficiently handle requests for different domains
without any reconfiguration. One of the best
examples of this approach is PRECISE [Popescu et
al., 2004]. PRECISE is developed at the University
of Washington by Ana-Maria Popescu et al in 2004
and it targets relational databases. The majority of
these works interface Prolog and relational
databases. For the XML database, the only
proposition is NALIX [Li et al., 2005]. It is a natural
language interactive natural interface developed at
the University of Michigan by Yunyao Li et al in
2006. The database used for this system is XML
database (Extensible Markup Language) with
'Schema-Free XQuery' as databases query language.
Schema-Free XQuery is a language designed
primarily to extract XML database information. The
idea of this language is to use keyword search for
databases. The transformation processes take place
in three steps, generation of the parse tree, validation
and then translation of this parse tree into an XQuery
query.
Generally, most of the NLIDBs are translating
English language queries. Other systems translate
queries written in Spanish [Rangel et al., 2002],
Urdu [Ahmad et al., 2009], Chinese [Li et al., 2015]
and Hindi [Kataria et al., 2015] languages. For the
French language, the first system process this
language is Edite [Reis et al., 1997]. Edite is a multi-
langue natural language interface to relational
databases. It answered user’s queries expressed in
Portuguese, French, English, and Spanish. It
functions independently of database domain. The
second system designed for the object-oriented
database [Hemerelain et al., 2010]. It presents the
semantic analysis of queries written in the French
language.
In this study, we focused on the development of
another generic interface for querying XML
databases by using the French language. This
interface operates independently of the domain of
application. In addition, the use of automatic
learning allows our system to have the ability to
automatically improve its knowledge base via
experience.
3 PROPOSED ARCHITECTURE
In Figure 1, we present the proposed architecture.
The proposed architecture is based on the
intermediate representation languages approach. The
use of this approach due to the difficulties of directly
translating the FLQ into an XPath. The idea of this
method is to map firstly the FLQ into a logical
intermediate query, expressed in an in XML form.
Then, this logical query translated into an XPath
query and evaluated against the XML database [1].
This architecture can be divided into two major
components.
The first part starts from the natural language
query to the logical query. The second part starts
ICCSRE 2018 - International Conference of Computer Science and Renewable Energies
74
from the logical query until the XPath query is
generated. The idea to express the logical query in
XML form, adds reasoning capabilities to the
system. Moreover, since the logical request is
independent of the database, it can be ported to
different database query language as well as to other
domains.
3.1 Linguistic Component
As shown in Figure 1, the linguistic component
composes of three analyses: morphological,
syntactic and semantic.
Figure 1: System architecture
3.1.1 Morphological Analysis
The term morphology comes from classical Greek
(morpheme) and means the study of form.
Morphology is concerned with the study of the
external structure of an object. In linguistics, it is the
study of how words combine to formulate the
sentence. The morphological analysis is responsible
for reading the FLQ and dividing it into primitive
elements called token. Then, it returns information
about each token. This process is performed using
the next functions. We note that in this function, we
give the example by the FLQ: "Affiche les noms et
les addresses des étudiants ayant un age < 19" which
means: give me names and addresses of student
whose age < 19. The FLQ used in this example
contains some spelling errors for shows the using of
the function spell checking:
Tokenization: this function used to divide the
FLQ into a primitive element, which is
considered as a single logical unit.
Example: < affiche> < les> <noms> < et> <les>
< addresses> <des> < étudionts > < ayant > <l> <
un > <âge> < ‘<’ > <19>
Spelling checker: used to make sure that each
word is in the dictionary. If this is not the case,
then spell checking is performed or a new word
is added to the system vocabulary.
Example: < affiche> < les> <noms> < et> <les>
< adresses > <des> < étudiants > < ayant > < un >
<âge> < ‘<’ > <19>
POS Tagger: Is the process used to associating
the grammatical function to each word of the
FLQ.
Example: <affiche , V> <les , DET> <noms ,
NC> <et , CC> <les , DET> < adresses , NC>
<des , P> < étudiants, NC> < ayant, PROREL>
French Language Interface for Xml Databases
75
<un , DET> <âge , NC> < <, PUNC > <19 ,
ADJ>
Morpheme: this function determines the
morpheme of each token which is the minimal
unit bearing meaning .
Example: < noms : nom> < adresses : adresse>
< étudiants: étudiant>
3.1.2 Syntactic Analysis
The syntactic analysis used to show how the tokens
in the FLQ are related to each other [Tari et al.,
2010]. This function allows our system to generate
the parse tree or the derivation tree.
For generating the parse tree, we use a context-
free grammar. The context-free grammar used the by
our system is presented in Figure 2. The elements of
CFG are defined by: CFG=[N,T,R,S] where:
N a set of non-terminal symbols. T is a set of
terminal symbols.
R is a set of context-free productions.
S is the start symbol used to represent the
ANLQ.
R1: SENTENCE OBJECTS CONDITION ORDER
CONJUNCTION SENTENCE
R2: SENTENCE QUE AUX_VERB OBJECTS
CONDITION ORDER CONJUNCTION SENTENCE
R3 :SENTENCE VERB_PHRASE OBS
CONDITION ORDER CONJUNCTION SENTENCE
R4 :VERB_PHRASE VERB
PERSONAL_PRONOUN
R5 :AUX_VERB EST | SONT
R6: QUEQUI | QUEL | QUELLE | QUELLES
R7: OBJECTS OBJECT CONJUNCTION
OBJECTS
R8 :OBJECT NOUN_PHRASE
R9 :NOUN_PHRASE QUANT
POSSESSIVE_PRONOUN DET NOUN
CONJUNCTION NOUN_PHRASE
R10 :NOUN_PHRASE NOUN_PHRASE
PREPOSITION_PHRASE CONJUNCTION
NOUN_PHRASE
R11 :NOUN_PHRASE DETERMINER
ADJECTIVE NOUN CONJk9UNCTION
NOUN_PHRASE
R12 : QUANT TOUS | TOUTE | TOUT| CHAQUE
R13 :PREPOSITION_PHRASE PREPOSITION
NOUN_PHRASE
R14 :PREPOSITION DE | DES | DU |
R15 :PERSONAL_PRONOUN MOI | NOUS |
VOUS | LUI
R16 :POSSESSIVE_PRONOUN MON| VOUS|
NOUS |TON| LEUR | LEURS
R17 :CONDITIONCOND OP CONJUNCTION
CONDITION
R18 :COND DONT | AVEC | OU
R19 : VERB DONNER| AFFICHER| MONTRER
R20 :CONJUNCTION ET | OU
R21 :OP OBJECT SYMBOL VALUE
R22 :SYMBOL IS | = | > | >= | < | <= | <> | COMME
R23 :ORDER ORD NOUN_PHRASE
CONJUNCTION ORDER
R24 :ORD ORDONNER PAR | TRIER PAR
R25: NOUN CLIENT| FACTURE | PROJET…
Figure 2: The context-free grammar used by the system
This grammar is composed of a set of syntactic
rules, which corresponds to the possible syntactic
structure of the FLQs. We can devise these rules in
two categories:
Domain-independent rules: are the syntactic
rules in which the right part content only non-
terminal symbols (i.e. rules 25 in Figure 3).
Domain-dependent rules: are the syntactic rules
in which right part content terminal symbols
(i.e. rules 9 in Figure 3) [Albert et al., 2011]
In order to be independent of the domain of
database, we can’t interpret all possible instance of
domain-dependent rules. For that we add to our
system a generator of syntactic rules. This generator
checks whether all the syntactic rules necessary to
parse the FLQ Knowledge Base (KB). If not, it
automatically detects the necessary syntactic rules,
creates and adds them to the KB. The addition of
syntactic rules generator to our system, help to adapt
its KB with the FLQ. So, the system automatically
improves its knowledge base through experience.
The function of the generator of syntactic rules is
described in the next algorithm.
Algorithm Syntactic_Rules _Generator
Input
FLQ a French Language Query
Output
a set of syntactic rules R = {(SRk), 1 k m }
Begin Divide the FLQ into a set of words W = {(wi : GFi)
, 1 i m} where GFi is the Grammatical Function of
each word wi ;
For each word (wi : GFi) W loop
Generate the syntactic rule SRk correspond to wi;
If SRk KB
add SRk to KB;
ICCSRE 2018 - International Conference of Computer Science and Renewable Energies
76
End if
End loop
Return R;
End Syntactic_Rules _Generator
Figure 3 displays the parse tree corresponded to
FLQ:
affiche les noms et les adresses des étudiants ayant
un âge < 19
3.1.3 Semantic Analysis
The semantic analysis produces an XML logical
query which used to assign a logical interpretation to
the parse tree created by the syntactic analysis. This
is done by using a set of semantic rules. Each
syntactic rule defined in the CFG has an equivalent
semantic rule. For that this process called rule-by-
rule style [Reis et al., 1997]. Table 1 displays some
instances of semantic rules with their equivalent
syntactic rules:
Table 1: Semantic rules with their equivalent syntax rules.
We have already mentioned in the previous
paragraph that the application of semantic rules on
the parse tree produces an XML logical query. In
XML, we can define the structure of XML logical
query by the following Document Type Definition
[DTD]:
<?xml version="1.0" encoding="UTF-8"
standalone="yes"?>
<!DOCTYPE logical query [
<!ELEMENT REQUEST [SELECT, COND*,
ORDER* ]>
<!ELEMENT SELECT [OBJECT+]>
<!ELEMENT CND [OBJECT, SYMBOL, VALUE+
]> <!ELEMENT ORDER [OBJECT]>
<!ELEMENT OBJECT [NAME, ATTRIBUTE+]>
<!ELEMENT ATTRIBUTE [NAME, AGGREGA +]>
<!ELEMENT NOM [PCDATA]>
<!ELEMENT VALUE [PCDATA]> <!ELEMENT
SYMBOL [PCDATA]> <!ELEMENT AGGREGA
[PCDATA]>
]>
The following XML logical query displays the
logical query associated to the parse tree of the FLQ:
Figure 3: Example of parse tree
<QUERY>
<SELECT>
<OBJECT>
<NAME>étudiants</NAME>
<ATTRIBUT>
<NAME>noms</NAME>
</ATTRIBUT>
<ATTRIBUT>
<NAME>adresses</NAME>
</ATTRIBUT>
</OBJECT>
</SELECT>
<CONDITION>
<OBJECT>
<ATTRIBUT>
<NAME>âge</NAME>
</ATTRIBUT>
</OBJECT>
<SYMBOL> < </SYMBOL>
<VALUE>19</VALUE>
</CONDITION>
</QUERY>
Syntactic rule Semantic rule
NOUN_PHRASE
PREPOSITION_PHRASE
<attribute >
<object>
NOUN_PHRASE
NOUN_PHRASE
PREPOSITION_PHRASE
< attribute 1>
< attribute 2 >
<object>
NOUN_PHRASE
PREPOSITION_PHRASE
PREPOSITION_PHRASE
< attribute >
<object1> <
object2 >
NOUN_PHRASE SYMBOL
VALUE
<attribute>
<symbol> <
value>
French Language Interface for Xml Databases
77
3.2. Database Knowledge Component
The function of the database knowledge is to
translate the XML logical query into XPath query.
This is done by mapping the elements of the XML
logical query into its clause in XPath query. This
process done in many steps. Each one manipulates
particular part in the logical query.
Step1: extract the names of the attribute from
the logical query.
Step1: select the portion of the logical query
that is contented the table name or a group of
table names
Step1: form the conditions of selection from
the logical query.
Each step is followed by an examination, to
validate if the name of tables and attributes exist in
the XML database. If it’s not the case, a mapping
table will use. This later stores the synonyms of
table and attribute names. After the generation of the
XPath queries, the system executes it and then,
shows the answers in XML form.
4 SYSTEM RESULTS
The interface shown in Figure 4 displays the result
of translation of the FLQ:
affiche les noms et les adresses des étudiants
ayant un âge < 19.
Figure 4: System interface
The following tables show a list of a variety of
FLQ that are successfully translated and executed by
our system.
Table 2: FLQ without projection and selection
FLQ
Generated
XPATH
affiche les clients
affiche tous les clients
/*/client/*
affiche moi nos clients
affiche moi tous les clients
Affiche moi les clients
affiche nos clients
affiche moi tous nos clients
affiche tous nos clients
clients?
quels sont nos clients?
liste tous nos salaries /*/ salarié /*
liste tous nos employés
ICCSRE 2018 - International Conference of Computer Science and Renewable Energies
78
montre moi tous nos clients
et projets
/*/client/*
/*/projet/*
Table 3: FLQ with projection
FLQ
Generated
XPATH
Donne moi les noms des
étudiants
/*/ étudiants
/nom
noms, âges et adresses des
employées
/*/ employées
/nom| /*/
employées /âge
/*/clients/adres
se
Montre moi tous les clients
noms, âges et adresses des
employées
Quels sont tous les noms et
les adresses des clients
/*/client/nom|/
*/client/adresse
Cherche tous the noms des
employés et des clients
/*/client/nom
/*/employé/no
m
Trouve les noms des clients
et les montants des factures
/*/client/nom
/*/facture/mont
ant
Table 4: FLQ with projection and selection
FLQ Generated XPATH
Montre tous les étudiants
dont le nom est "Hanane"
/*/ étudiant [nom =
“hanane”] /*
Tous notre clients dont
nom est "Hanane" ou "
Mustapha"
/*/client[nom =
“Hanane”]/*|
/*/client[nom=”Musta
pha”]/*
Montre tous les
enseignants avec l’âge
est entre 28 et 40
/*/ enseignant
[âge > 28] [âge<40]/*
quels sont les noms des
employées dont l’adresse
est "Agadir dakhla " ?
/*/employée [adresse=
" Agadir dakhla "]
/nom|/*/employée
[adresse = "Agadir
dakhla "] /âge
Affiche moi tous les
adresses des client Qui
ont l’âge inférieur à 30
/*/client [âge < 30] /
adresse
cherche les adresses des
clients avec âge supérieur
ou égale à 26 et le nom
est "Hanane"
/*/client [âge <= 26]
[nom = "hanane"]
/adresse
Table 5: Query with aggregate function
FLQ Generated XPATH
Donne moi le nombre
des fournisseurs dont le
nom est "Hanane"
/*/count(fournisse
ur [nom =
"hanane"])
compter tous nos projets
/*/count(projet)
Montre moi la moyenne
des âges des clients
avg (/*/client/âge)
Donne moi le minimum
âge des étudiants
min (/*/étudiant
/âge)
Affiche le maximum âge
des clients dont l’âge est
inférieur ou égale à 40
max (/*/client [âge
< 1000] /âge)
Quels sont les employés
avec le maximum âge?
/*/ employée
[âge=max
(/*/employée
/âge)]/*
To experience the performance of our system, we
use a set of 2000 FLQ. The results obtained by this
test were tabulated in Table 6.
Table 6: The results obtained by first test
Answered Queries
Unanswere
d Queries
2000
(100% )
1944 56
97,2% 2,8 %
Table 6 presents that our system translates 97.2%
of 2000 FLQ are translated to XPATH.
Table 7: The results obtained by the second test
XPath Correctly
generated
XPath
Incorrectly
generated
NB (1944)
1930
14
%
99,27% 0,72%
From Table7 we show that 1930 FLQs are
generated correct XPath queries. We said that query
French Language Interface for Xml Databases
79
is correctly generated if the database queries
produced is syntactically correct.
For the queries that are correctly generated, not
all of the generated XPath match FLQs. Table 8
displays the number of XPath matches FLQ and the
numbers of XPath don’t match FLQ.
Table 8: XPATH matches FLQ
XPath matches
NLQ
XPath doesn't
match NLQ
NB
(1930)
1897 33
% 98.29 % 1.70%
As presented in Table 8, 98.29% of XPath
queries that are correctly generated matches FLQ.
5 CONCLUSION AND FUTURE
WORK
This research paper presents a model of an
architecture of intelligent interface for querying
XML database. It proposes a method for translating
FLQ into an XPath queries. The main objective of
this interface is to allow communication between the
XML database and its users using the French
language. One of the advantages of this interface is
that it functions independently of database domain
and it improves automatically its knowledge base
through experience. The results of experimentation
show that the methods employed in the system have
the capabilities to produce an XPath queries for a
very important number of FLQ.
As future work, we will continue to solve more
complex queries. Also, we intend to apply the
methods used in the paper to translating queries in
other languages, such as the Arabic language.
REFERENCES
Akerkar, R., Joshi, M.,2008. Natural Language Interface
Using Shallow Parsing. In International Journal of
Computer Science Applications, Vol. 5, pp. 70-90.
Albert, J., Giammarresi, D., Wood, D., 2001. Normal form
algorithms for extended context-free grammars.
Theoretical Computer Scienc, Vol. 267, pp. 35-47.
Androutsopoulos, I., Ritchie, G.D., Thanisch, P., 1993.
MASQUE/SQL - An Efficient and Portable Natural
Language Query Interface for Relational Databases.
Proceedings of the 6th International Conference on
Industrial and Engineering Applications of Artificial
Intelligence and Expert Systems,1_4 June1993
Edinburgh, Scotland, pp.327-330.
Androutsopoulos, I., Ritchie, G., Thanisch, P., 1995.
Natural Language Interfaces to Databases – An
Introduction. In Journal of Natural Language
Engineering 1 Part 1, pp.29-81.
Auxerre, P., Inder, R., 1986. MASQUE Modular
Answering System for Queries in English, User’s
Manual, technical report AIAI/SR/10, Artificial
Intelligence Applications Institute, University of
Edinburgh.
Bais, H., Machkour, M., Lahcen, K., 2016. Querying
database using a universal natural language interface
based on machine learning. In International
Conference on Information Technology for
Organizations Development (IT4OD). IEEE.
Gauri, R., Agarwal, C., Chaudry, S. , Kulkarni, N. , Patel,
S.H., 2010.Natural language query processing using
semantic grammar’. In International Journal on
Computer Science and Engineering, Vol. 2, pp. 219-
223.
Hemerelain, B.Belbachir, H., 2010. Semantic Analysis of
Natural Language Queries for an Object Oriented
Database. In JSEA, Vol. 03, pp. 1047-1053.
Hendrix, G., Sacerdoti, E., Sagalowicz, D., Slocum, J.,
1978. Developing a natural language interface to
complex data. ACM Transactions on Database
Systems, Vol. 3, pp. 105-147.
Kataria, A.Nath, R., 2015. Natural Language Interface for
Databases in Hindi based on Karaka Theory. In
International Journal of Computer Applications, Vol.
122, pp. 39-43.
Li, Y., Yang, H., Jagadish, H.V., 2005. Nalix: an
interactive natural language interface for querying
xml. SIGMOD 05: Proceedings of the 2005 ACM
SIGMOD international conference on Management of
data, pp. 900-902.
Li, Z., Li, J., Ning, W., 2015. Research on Chinese Natural
Language Query Interface to Database Based on
Syntax and Semantic. In Applied Mechanics and
Materials, Vol. 731, pp. 237-241.
Papadakis, N., Kefalas, P., Stilianakakis, M., 2011. A tool
for access to relational databases in natural language.
In Expert Systems with Applications, Vol. 38, pp.
7894-7900.
Popescu, A-M., Armanasu, A., Etzioni, O., Ko, D., Yates,
A., 2004. Modern Natural Language Interfaces to
Databases: Composing Statistical Parsing with
Semantic Tractability. In Proc. COLING’04.
Rachide, A., Khan, M.A., Ali, R., 2009. Efficient
Transformation of a Natural Language Query to SQL
for Urdu. Proceedings of the Conference on Language
and Technology, pp. 53-60.
Rangel, R.A.P., Gelbukh, A.F., Barbosa, J.J.G., Ruiz,
E.A., Mej_a, A.M., S_anchez, A.P.D., 2002. Spanish
Natural Language Interface for a Relational Database
Querying System. 5th International Conference, TSD
2002 Brno, Czech Republic, pp. 123-130.
ICCSRE 2018 - International Conference of Computer Science and Renewable Energies
80
Reis, P., Matias, J., Mamede, N., 1997. Edite A Natural
Language Interface to Databases: A New Dimension
for an Old Approach. Proceedings of the International
Conference in Edinburgh, Scotland, pp. 317-326.
Tari, L., Tu, P.H.,Hakenberg, J., Chen, Y., Son, T.C.,
Gonzalez, G., Baral, C., 2010. Parse Tree Database for
Information Extraction. Proceedings of IEEE
transactions on knowledge and data.
Waltz, D., 1978. An English language question answering
system for a large relational database.
Communications of the ACM 21, pp. 526-539.
Warren, D., Pereira, F., 1982. An Efficient Easily
Adaptable System for Interpreting Natural Language
Queries. In Computational Linguistics, Vol. 8, pp.
110-122.
French Language Interface for Xml Databases
81