=Paper=
{{Paper
|id=Vol-1179/CLEF2013wn-QALD3-HeEt2013
|storemode=property
|title=CASIA@QALD-3: A Question Answering System over Linked Data
|pdfUrl=https://ceur-ws.org/Vol-1179/CLEF2013wn-QALD3-HeEt2013.pdf
|volume=Vol-1179
|dblpUrl=https://dblp.org/rec/conf/clef/HeLCZL013
}}
==CASIA@QALD-3: A Question Answering System over Linked Data==
https://ceur-ws.org/Vol-1179/CLEF2013wn-QALD3-HeEt2013.pdf
CASIA@QALD-3: A Question Answering System over
Linked Data
Shizhu He, Shulin Liu, Yubo Chen,
Guangyou Zhou, Kang Liu, and Jun Zhao
National Laboratory of Pattern Recognition,
Institute of Automation, Chinese Academy of Sciences.
Zhongguancun East Road 95#, Beijing, 100084, China.
{shizhu.he,shulin.liu,yubo.chen,gyzhou,kliu,jzhao}@nlpr.ia.ac.
cn
Abstract. We present a question answering system (CASIA) over Linked Data
(DBpedia), which focuses on construct a bridge between the users and the Linked
Data. Based on the Linked Data consisting of subject-property-object (SPO) triples,
each natural language question firstly is transformed into a triple-based represen-
tation (Query Triple). Then, the corresponding resources in DBpedia, including
class, entity, property, are mapped for the phrases in the query triples. Finally,
the optimal SPARQL query is generated as the output result. Our system can not
only deal with the single-relation questions but the complex questions containing
multi-relations. We evaluate our approach on QALD-3 test dataset and achieve
an F-measure score of 0.36, an average precision of 0.35 and an average recall of
0.36 over 99 questions.
Keywords: Question Answering, Linked Data, DBpedia
1 Introduction
With the rapid development of the Web of Data, there are many RDF datasets pub-
lished as Linked Data[1]. Thus, developing user-friendly interface for accessing those
linked data become increasing important. However, there are the “gaps” between the
users and the Linked Data. On the one hand, the users, even expert programmers, need
a lot of practices to handle standard structured query languages like SPARQL. On the
other hand, due to the diversity and high heterogeneity of the Linked Data, it is difficult
for humans to select relevant resources and discover useful information.
Question answering over Linked Data is aimed at eliminating those “gaps”, which
attempts to allow the users to access those structured data with natural language. To
fulfill this aim, some systems have already been proposed during the last few years.
PARALEX[2] maps open-domain questions to subject-property-object (SPO) triples
2 Shizhu He, Shulin Liu, Yubo Chen, Guangyou Zhou, Kang Liu, Jun Zhao
database by leveraging an automatically constructed lexicon. The limitation of [2] is
that it can only answer the simple question which only contains one single relation,
but fail to deal with the questions with complex structures. DEANNA[3] believes the
translating natural language questions into the structured SPARQL queries consists with
several disambiguation tasks. They solve those disambiguation tasks jointly by using
an integer linear program. However, some questions still cannot be answered by both
systems, such as aggregation, filter, comparative and superlative questions. TBSL[4]
addresses those problems by utilizing some hand-written templates, and filling slot in
SPARQL templates with resources (entity, class or property).
In this paper, we demonstrate our system in QALD-3, which focuses on translating
natural language expression into standard RDF query language expression (SPARQL).
Our system implements a basic pipeline framework which consists three main com-
ponents, including question analysis, resource mapping and SPARQL generation. In
specific, we first employ shallow and deep linguistic analysis to transform NL-queries
into a set of Query Triples with format. Second, we map
each phrase in Query Triple to the corresponding resource (class, entity, or property)
in DBpedia. As a result, Ontology Triples are generated. Thirdly, the SPARQL queries
will be constructed based on Ontology Triple and question type. At last, the generated
SPARQL queries is used to search on the Linked Data, and the best answer can be
picked out through validating and ranking. The framework of our system is similar to
PowerAqua[5]. However, besides different technologies used for question analysis, the
big difference is that we use relational patterns when mapping relations rather than use
WordNet and the context to disambiguate the phrase in query triples. Furthermore, our
system is similar to QAKiS[6], which is to exploit a relation-based matching and con-
struct relational patterns automatically from Wikipedia. In contrast, our work focuses
not only the simple questions, but also on complex questions. Moreover, we use sev-
eral free resources including PATTY[7] and WordNet[8] instead of constructing relation
pattern repository by ourselves.
2 System Description
The architecture of our system will be presented in this section. We will give a
detailed description of each component and give a step by step explanation with the
following example:
Who are the parents of the wife of Juan Carlos I?(id=67, test set).
2.1 System Architecture
The current version of our QA system targets the questions which do not contain
comparative and superlative words. We also do not consider the questions which contain
aggregation and filter condition. Generally, our system is composed of three main steps
(as shown in Figure 1):
1) Question Analysis: Given a natural language question, we first detect the type
of this question (yes/no, number, etc.) and extract the name entities from it. Then we
CASIA@QALD-3: A Question Answering System over Linked Data 3
Fig. 1. The architecture of CASIA
modify some phrases and their POS tags by combining name entities and keywords.
After that, we take deep linguistic analysis to generate the typed dependency parsing
tree for this question. At last, a set of linguistic triples is
constructed from the dependency parsing tree. The upper example will be transformed
to two query triples, where the second one is a sub-triple of the first one:
2) Resource Mapping: For every phrase in query triple, the most important thing is
to identity the corresponding resource in Linked Data, such as DBpedia. For different
types of resource, we use different techniques and resources. The output of this step is a
list of ontology triples. One query triple will generate several possible ontology triples.
Possible triples of the aforementioned example are:
3) SPARQL Generation: Based on identified question type (bool, number, etc.) and the
results of resource mapping, we generate SPARQL query candidates. Then DBpedia
SPARQL endpoint will be used to validate those query candidates. The legal queries
with the highest score will be selected and returned as the right answer. If there are
several different results with the same highest score, we combine them using SPARQL
language keyword “UNION”. A SPARQL query candidate of the example question is
as following:
4 Shizhu He, Shulin Liu, Yubo Chen, Guangyou Zhou, Kang Liu, Jun Zhao
PREFIX rdf:
PREFIX dbo:
PREFIX dbp:
SELECT DISTINCT ?URL
WHERE {
?URL rdf:type dbo:Person.
?URL dbo:parent ?wife.
?wife dbo:spouse dbp:Juan Carlos I of Spain.
}
2.2 Question Analysis
We design some simple heuristics to identify the question type, based on which
we will use different SPARQL templates to generate queries. For example, the yes/no
question will use “ask where” template; the “sum” question will use “select count(?x)”
template. In general, we use the following regular expressions:
— “ˆ (are|did|is|was|does|were|do).*”, corresponding to the yes/no question;
— “ˆ how(many|tall|long).*”, corresponding to the ”sum” question;
—“ˆ (when|how often|since when|how long|for how long).*”, corresponding to
the Date question;
— “ˆ (what|which|where|who).*” and “ˆ (show me|give me|list).*”, correspond-
ing to normal “select” question.
For a yes/no question, we transform it to a declarative sentence based on phrase
structure parsing tree. For example, “Is Michelle Obama the wife of Barack Obama?”
(id=67, test set) will be transformed to “Michelle Obama is the wife of Barack Obama.”.
Considering the Linked Data consisting subject-property-object (SPO) triples, we
transform questions into query triples based on dependency parsing tree. Given a natural
language question and corresponding keywords, we use Stanford NER tool1 to identify
name entities in question, and then generate a typed dependency parsing tree combining
keywords, name entities and POS tags. However, the name entities in question cannot
be occasionally identified by Stanford NER tool. Through our coarse statistics, there are
74 entities identity correctly in all 95 entities from 99 questions. However, we cannot
extract any NEs from such question “Give me the Apollo 14 astronauts”; and we just
extract “Juan Carlos” from question “Who are the parents of the wife of Juan Carlos
I?” (the right result is “Juan Carlos I”). Therefore, we use keywords of question pro-
vide by QALD-3 organizers; we extract keywords with capitalization as proper nouns.
For example, the parsing result of the example question is showed as follows(Figure 2):
Then, we use some heuristics to construct query triple from parsing tree. For example,
we collect the phrases in the path “nsubj→prep→ pobj” and transform it into a triple.
If there is a dependency path in a phrase (“the wife of Juan Carlos I” in the “pobj” of
phrase “parents”), we will construct an additional query triple. Because we mainly ex-
tract subject-predicate-object and proposition phrases relations, we cannot extract query
triples for many questions, such as the question “Which films starring Clint Eastwood
did he direct himself?”.
1
http://nlp.stanford.edu/software/corenlp.shtml
CASIA@QALD-3: A Question Answering System over Linked Data 5
Fig. 2. The dependency parsing tree of the example question
2.3 Resource Mapping
Query triples are the intermediate representation form of user’s natural language
question. To generate SPARQL queries, the most important thing is mapping them to
the right “resource” in Linked Data. There are four type resources in Linked Data:
Class (concept), Instance (Entity), Property (Relation) and Literal. Currently we ignore
the Literal resource. Thus, the task of this step is to find appropriate item in those three
resources sets by given a phrase in query triple.
We assume that different positions of a query triple will be mapped to different
type resources. Thus, in , “Subject” and “Object” may be
mapped to Class and Instance, and “Relation” will be mapped to Property. To fulfilling
this mapping, we build three Lucene2 Indexes separately. However, finding different
types of resource needs different techniques. In specific, mapping to instance, index
lookup in the inverted index can yield good performance. Additionally, we can find the
right resources even given different spellings by applying Wikipedia redirects. For ex-
ample, any phrase of “MJ23”, “His Airness”, “Micheal Jordan” and “Michael Jeffery
Jordan” will be mapped to the same URI < http://dbpedia.org/resource/
Michael_Jordan>. Sometimes the phrase may contain a class, such as in the ques-
tion “Which books by Kerouac were published by Viking Press?”, “book” should be
mapped to the class URI < http://dbpedia.org/ontology/Book>. If the
complete match cannot find appropriate URIs for phrase, we will perform the following
steps to revise the results:
1) Search the similar resources and calculate their string similarities (Levenshtein
distance) with the given words;
2) expand the given words using WordNet;
3) using lexical word when encountering “prep”, such as “by” will invoke “dbo:author”.
In the current version of our system, we just invoke four relations using hand-crafted
rules, including “by” invokes “dbo:starring” and “dbo:author”; “in” and “of” invoke
“dbo:country”, “dbo:starring” and “dbo:locationCity”.
However, when mapping property, we may still fail to get right properties through
the above steps. For example in the question “Who is the husband of Amanda Palmer?”,
“husband” should be mapped to “dbo:spouse” with URI < http://dbpedia.
org/ontology/spouse>. Several recent studies focus on finding the relation pat-
terns for a property, such as BOA[9] and PATTY[6]. We use PATTY3 instead of collect
the Relational Patterns by ourselves.
2
http://lucene.apache.org
3
http://www.mpi-inf.mpg.de/yago-naga/patty/
6 Shizhu He, Shulin Liu, Yubo Chen, Guangyou Zhou, Kang Liu, Jun Zhao
2.4 SPARQL Generation
Each phrase in query triple may be mapped to more than one resources in the Linked
Data. To filter excessive resources, we set a similarity threshold to rank SPARQL queries.
For every query triple S,P,O, we generate S,P,?x and ?x,P,O in order to check whether
the subject or object actually occurs with the property. If one question has 3 conditions
(3 query triples), we will generate 23 = 8 SPARQL queries. Based on the identified
question type, we generate complete SPARQL query, and use DBpedia SPARQL end-
point4 to validate the legality of each candidate query. The process of validating and
constructing rules is presented as follows:
—bool question: For the “bool” type questions, we use “ask” template. The query
is legal only if it returns “true”, and we cannot deal with the question which answer is
“wrong”.
—number question: At first we use “select” template. And, if the retrieval results
are more than one, or the only one result is not an integer value, we will use “select
count” template.
—other question: If the query could get results, it is a legal query.
Each legal query will be associated with a score. We select the queries with the
highest score, and combine them using “UNION” if there are more than one queries
sharing the same highest score.
3 Results on QALD-3 data
Table 1 gives the evaluation results with average precision, average recall and F-
measure. It shows the number of question our system can answer, the number of right
and partially right answers among them.
Table.1. Evaluate results on QALD-3 test dataset.
Total Right Partially right Avg.Precision Avg.Recall F1
99 29 8 0.35 0.36 0.36
Table 2 gives the time consumption of 10 questions selected randomly. Except the ques-
tion ID, and the questions shown in column 1 and column 2 respectively, column 3
shows the number of generated query triples generated, column 4 shows question anal-
ysis step time consumption (second), and the SPARQL generated step time consump-
tion (second) is shown in column 5. The average time consumption of all 99 QALD-3
test set questions is shown in the last row. The more query triples generated in the first
two steps, the more time cost in the final step. Yet, the time cost in question analysis for
different questions is almost the same. The “OUT OF SCOPE” question do not have
final step.
4
http://vtentacle.techfak.uni-bielefeld.de:443/sparql
CASIA@QALD-3: A Question Answering System over Linked Data 7
Table.2. Time consumption on answering QALD-3 test questions.
Number of Time in Time in
ID Question Query first two first last
Triples steps step
Who was the successor of John F.
2 18 0.07 28.223
Kennedy?
Give me all people that were born in
19 10 0.459 8.684
Vienna and died in Berlin.
20 How tall is Michael Jordan? 39 3.083 35.177
21 What is the capital of Canada? 5 0.141 4.249
What are the nicknames of San
54 11 0.122 9.324
Francisco?
Who are the parents of the wife of Juan
67 240 7.566 855.661
Carlos I?
Which books by Kerouac were published
81 336 8.479 410.263
by Viking Press?
86 What is the largest city in Australia? 22 6.886 29.646
Who painted The Storm on the Sea of
97 18 2.887 30.016
Galilee?
100 Who produces Orangina? NULL 13 NULL
Average on all 99 questions 60.54 3.28764 83.2855
3.1 Error Analysis and Discussion
As introduced before, our current version system cannot address questions which
contain comparative, superlatives and negation words. We cannot answer complex ques-
tions with relative sentences and conjunctions also. For example, we cannot address
question “Which films starring Clint Eastwood did he direct himself?” (id=88, test set)
In our observation, the mistakes in our system can be classified into four types. We
discuss each of the four type errors in detailes.
First, some errors occur in the question analysis step. Because we just extract subject-
predicate-object and proposition phrases relations, wrong query triples generated for
some questions. For example, we only extract in question
“Give me all people that were born in Vienna and died in Berlin.”(id = 19, test set),
and the proper query triples should have two coordination triples and . We can deal with the questions with cascade
relation. For example, in question “Which books by Kerouac were published by Viking
Press?”(id = 81, test set), “publish by” is modify the phrase “books by Kerouac”.
Second, there are some entities mapping errors. We do not resolve the entity disam-
biguation problem, some entities extracted from question are wrong using string match-
ing. For example, we extract entity “GMT” (dbp:GMT) in question “List all games by
GMT.”(id = 40, test set), but “GMT” was an ambiguities word and it corresponded
“GMT Games” (dbp:GMT Games) in this question.
Third, considering the ontology we rely on for concept extraction, some concepts
(class) CASIA cannot identify. For example, “surfers” cannot be identified as proper
concept in question “Which professional surfers were born on the Philippines?”(id =
46, test set).
Finally, most of our current version systems mistakes concern wrong relation map-
ping. For example, we cannot extract any relation in question “Which countries are
8 Shizhu He, Shulin Liu, Yubo Chen, Guangyou Zhou, Kang Liu, Jun Zhao
connected by the Rhine?”(id = 45, test set), the same situation occur in “What is the
ruling party in Lisbon?”(id=53, test set) and “How tall is Michael Jordan?”(id = 20,
test set). Even though PATTY was a good relation patterns repository, mapping its pat-
terns to relation properties in DBpedia was not perfect.
Except upper errors, our system also cannot detect the right order between subject
and object. For example, in question “Who are the parents of the wife of Juan Carlos
I?”(id = 67, test set), we generated SPARQL expression contain meanings of “Who
are the parents of the wife of Juan Carlos I?” and “Who are the child of the wife of
Juan Carlos I?”. The same problem occurs in the question “Did Socrates influence
Aristotle?”(id = 62, test set).
4 Conclusion and Future Work
In this paper, we present a question answering system over Linked Data, which
translate the natural language questions into the standard RDF data queries (SPARQL).
The pipeline of our system consists of three main components, including question anal-
ysis, resource mapping and SPARQL generation. In specific, we use shallow (POS tag-
ging, NER) and deep linguistic analysis (dependency parsing) to transform each natural
language question into triple-based representation. Then we use PATTY, WordNet and
other free external resources to find the corresponding resource of the phrases in the
query triples. At last, based on the identified question type, we generate the SPARQL
queries and use the DBpedia endpoint to validate them. However, we still cannot an-
swer the comparative and superlative questions. In addition, the linguistic analysis can-
not identify all relations in questions. In the future we will address those problems and
include more Linked Data (currently we only use DBpedia).
5 Acknowledgements
Thanks to Prof. Hang Li, Dr. Yibo Zhang and Dr. Jie Zhang at Huawei Noah’S
Ark Lab for their insightful advices. This work was supported by the National Natural
Science Foundation of China (No. 61070106, No. 61272332 and No. 61202329), the
National High Technology Development 863 Program of China (No. 2012AA011102),
the National Basic Research Program of China (No. 2012CB316300).
References
1. Bizer, C., Heath, T., Berners-Lee, T.: Linked data - the story so far. Journal on Semantic Web
and Information Systems (in press) (2009)
2. Fader, A., Zettlemoyer, L., Etzioni, O.: Paraphrase-Driven Learning for Open Question An-
swering. In: Proc. of ACL’13(2013)
3. Yahya, M., Berberich, K., Elbassuoni, S.,Ramanath, M.,Tresp, V., Weikum, G.: Natural Lan-
guage Questions for the Web of Data. In: Proc. of EMNLP’12(2012)
4. Unger, C., Bhmann, L., Lehmann, J., Ngomo, A., Gerber, D., Cimiano, P..: Template-based
Question Answering over RDF Data. In: Proc. of WWW’12(2012)
CASIA@QALD-3: A Question Answering System over Linked Data 9
5. Lopez, V., Fernndez, M., Stieler, N., Motta, E.: PowerAqua: supporting users in querying and
exploring the Semantic Web content. Semantic Web Journal, In Press (2011)
6. Cabrio, E., Aprosio, A., Cojan, J., Magnini, B., Gandon, F., Lavelli, A.: QAKiS @ QALD-2.
In: Proceedings of Interacting with Linked Data (ILD 2012) [37] (2012), 87-95. http://ceur-
ws.org/Vol-913/07-ILD2012.pdf
7. Nakashole, N., Weikum, G., Suchanek, F.: PATTY: A Taxonomy of Relational Patterns with
Semantic Types. In: Proc. of EMNLP’12(2012)
8. George A. Miller (1995). WordNet: A Lexical Database for English. Communications of the
ACM Vol. 38, No. 11: 39-41
9. Gerber, D., Ngomo, A.: Bootstrapping the Linked Data Web. 1st Workshop on Web Scale
Knowledge Extraction, ISWC’11(2011)