Developing question answering (QA) systems that process natural language is a popular research topic. Conventionally, when QA systems receive a natural language question, they choose useful words or phrases based on their parts-of-speech (POS) tags. In general, words tagged as nouns are mapped to class entities, words tagged as verbs are mapped to property entities, and words tagged as proper nouns are mapped to named entities, although the accuracy of entity type identification remains low. Afterward, the relationship between entity types as RDF types determines the first element to be a pivot word to generate the SPARQL (acronym for SPARQL protocol and RDF query language) query on the basis of the sequences by a specific graph or tree structure, such as dependence tree or directed acyclic graph (DAG). However, the generated SPARQL query is difficult to adapt to the given query request in that the sequences are decided by a fixed structure. Unlike in previous research, SPARQL generation occurs automatically according to the entity type identification and RDF type identification results. This study attempts to design a method that leverages machine learning to learn human experiences in entity type identification as well as RDF-type identification. We approach the problem as a multiclass classification problem and propose a two-stage maximum-entropy Markov model (MEMM). The first stage identifies the entity type and the second identifies the RDF type for the purpose of generating appropriate SPARQL queries to meet the query request. Along with the templates designed for the two-stage MEMM model, we develop an automatic question answering prototype system called QAWizard. The experimental results show that QAWizard outperforms all other systems in question answering when evaluated on Linked Data version 8 (QALD-8) metrics.

Much like traditional database querying, the question answering process in a Question Answering (QA) system involves converting a user’s question input into query grammar, querying the knowledge base through the query grammar, and finally returning the query result (i.e., the answer) to the user. The accuracy of query grammar generation is therefore important in determining whether a Question Answering system can produce a correct answer. Generally speaking, incorrect query grammar will never find the right answer. SPARQL is the most frequently used query language in question answering systems. In the past, SPARQL was generated based on graph structures, such as dependency trees, syntax trees and so on. However, the query cost of generating SPARQL is high, which creates long processing times to answer questions. To reduce the query cost, this work proposes a low-cost SPARQL generator named Light-QAWizard, which integrates multi-label classification into a recurrent neural network (RNN), builds a template classifier, and generates corresponding query grammars based on the results of template classifier. Light-QAWizard reduces query frequency to DBpedia by aggregating multiple outputs into a single output using multi-label classification. In the experimental results, Light-QAWizard’s performance on Precision, Recall and F-measure metrics were evaluated on the QALD-7, QALD8 and QALD-9 datasets. Not only did Light-QAWizard outperform all other models, but it also had a lower query cost that was nearly half that of QAWizard.

SPARQL is a powerful query language which has been widely used in various natural language question answering (QA) systems. As the advances of deep neural networks, Neural Machine Translation (NMT) models are employed to directly translate natural language questions to SPARQL queries in recent years. In this paper, we propose an NMT-based approach with Transformer model to generate SPARQL queries. Transformer model is chosen due to its relatively high efficiency and effectiveness. We design a format to encode a SPARQL query into a simple sequence with only RDF triples reserved. The main purpose of this step is to shorten the sequences and reduce the complexity of the target language. Moreover, we employ entity type tags to further resolve mistranslated problems. The proposed approach is evaluated against three open-domain question answering datasets (QALD-7, QALD-8, and LC-QuAD) on BLEU score and accuracy, and obtains outstanding results (83.49%, 90.13%, and 76.32% on BLEU score, respectively) which considerably outperform all known studies.

DBpedia is one of the most resourceful link databases today, and to access information in DBpedia databases, we need to use query syntax (e.g., SPARQL). However, not all users know SPARQL, so we must use a natural language query system to translate the user’s query into the corresponding query syntax. Generating query syntax through the query system is both time-consuming and expensive. To improve the efficiency of query syntax generation from user questions, the multi-label template approach, specifically Light-QAwizard, is utilized. Light-QAwizard transforms the problem into one or more single-label classifications using multi-label learning template approaches. By implementing Light-QAwizard, query costs can be reduced by 50%, but it introduces a new label during the transformation process leading to sample imbalance, compromised accuracy, and limited scalability. To overcome these limitations, this paper employs employ two multi-label learning methods, namely Binary Relevance (BR) and Classifier Chains (CC), for question transformation. By employing Recurrent Neural Networks (RNNs) as a multi-label classifier for generating RDF triples, we predict all the labels that align with the query intentions. To better account for the relationship between RDF triples, we integrate BR into an ensemble learning approach, resulting in the Ensemble BR. Experimental results demonstrate that our proposed method outperforms previous research in terms of improving query accuracy.