Query Reformulation Approach Using Domain Specific Ontology

1. Systematic Literature Review

These publications conduct a systematic review of the research literature. First, a search methodology was created, then systematic searches were carried out in accordance with it. These searchers were guided by strings that were created in accordance with the determined study question. After then, a search method was developed to group all of the searches into categories based on the search journals. Also, research publications were screened based on their title, abstract, and objectives as well as included based on their inclusion criteria.

Initial, a search methodology was created, then systematic searches were carried out in accordance with it. These searchers were guided by strings that were created in accordance with the determined study question. Afterwards, all of the searches were categorized using a search method. Also, research publications were screened based on their title, abstract, and objectives as well as included based on their inclusion criteria.

1.1. String Development

The strings were developed by using three synonyms of each keyword.

Research Question: Query Reformulation Approach using Domain Specific Ontology

• Searching protocol:

In accordance with a search protocol, papers published throughout a four-year period (2019, 2020, 2021, 2022, and 2023) were chosen for search. Moreover, three databases and three synonyms for each term were searched. There were only 10 papers chosen for each string. In Figure 4, the search tactics are shown.

1.2. Filtering:

In filtering the first stage is title-based filtering, second is abstract-based filtering, and third is objective-based filtering as shown in Figure 5.

2.2.1. Title-based Filtering:

Title-based filtering was the initial stage of the filtering process, as depicted in Figure. Any publications that did not address the issue at hand were removed from the databases that were chosen.

Figure 6. Title based Filtering.

Figure 6. Title based Filtering.

• Abstract-based Filtering:

Abstract-based filtering was done in the second section. All of the chosen databases did not include any papers whose abstracts did not address the issue. Figure 5 illustrates the third stage of filtering, which involved the use of objective-based filtering. A table was created showing papers organized by their objectives after all of the papers were filtered according to their objectives.

Figure 7. Abstract based Filtering.

Figure 7. Abstract based Filtering.

• Objective-based Filtering:

The goals of the research papers were determined and grouped into clusters following the title and abstract-based clustering. This objective-based screening is displayed in Table 2.

Figure 8. Objective based Filtering.

Figure 8. Objective based Filtering.

Figure 9. Filtering.

Figure 9. Filtering.

Table 3 briefly shows the notations of the terms that were used in other tables to increase the readability of the readers.

3. Detailed Literature:

3.1. Query reformulation approaches using domain-specific ontologies

The research of [6] provides ways for query reformulation that are helped by ontologies, making it simpler for users to access the millions of medical records in the UK and Europe from any location. By this research, users and their applications will be helped to create queries without having to fully understand the information structure of the underlying data sources. Ontology-based search criteria and associated domain knowledge have been interpreted by query reformulation techniques and algorithms, which can then reformulate a relational query.

The research of [7] provides a method for automatically reformulating keyword queries that takes advantage of the structural semantics of the underlying structured data sources. The two stages of the reformulation solution are offline term relation extraction and online query creation. In order to extract pertinent terms from the context of the graph, we first create a heterogeneous graph to model the words and items in structured data. We present an effective probabilistic generation module to suggest replaceable reformulated queries during the online query reformulation stage. On a real-world data set, numerous tests are carried out, and our method produces encouraging results.

In the research of [1], the researcher used Protégé 5.0 to build a string ontology from scratch in a new domain (the music domain) for semantic information retrieval for a search engine query. The suggested ontology-based semantic information retrieval method (OBSIRM), which has been developed to improve web search in the music domain, also makes use of this taxonomy. The researcher has suggested a revolutionary method for improving web searches in which the original question is first substituted with acronyms and then the multilingual concept is used to search a query. Google has been used to compare the proposed method's outcomes. Precision and recall have been used to gauge the suggested method's effectiveness and accuracy.

In the research of [8] we take the most likely interpretations as the query concepts and choose the most related fundamental concepts. We talk about the problem of creating the basic concepts from the retrieved collection of documents or from the entire corpus. We use the TREC8 collection for our experiments. The experimental analysis demonstrates that our method is superior to existing query reformulation techniques and is especially useful for subpar queries. Also, we discover that the strategy using the fundamental concepts derived from the collection of retrieved papers produces the best results.

In the research of [9], The usage of ontology-based information retrieval methods and procedures is examined in this study, taking into account the modelling, processing, and conversion of ontological knowledge into database search terms. An effective optimized ontology model with query execution for content extraction from documents (OOM-QE-CE) is suggested in this study work. In terms of data and semantic loss, structural mapping, and domain knowledge applicability, the current ontology-to-database transformation and mapping approaches are also thoroughly analyzed.

In the research of [6] propose a new QE semantic approach based on the modified Concept2vec model using linked data. The novelty of our work is the use of query-dependent linked data from DBpedia as training data for the Concept2vec skip-gram model. This research considered only the top feedback documents, and we did not use them directly to generate embedding’s; we used their interlinked data instead. Also, this research used the linked data attributes that have a long value, e.g., “dbo: abstract”, as training data for neural network models, and, we extracted from them the valuable concepts for QE. In our exploration of a query reformulation approach using a domain-specific ontology, our research draws upon foundational insights presented in [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34].

• Technique based filtering:

In Technique-based filtering the researcher has read out all the techniques of the paper and identify which method is used by researcher to overcome the research problem.

Table 4. Technique based Filtering.

Ref.	PC	OBSIRM	RW	OWL-DL	OOM-QE	Concept2vec	HM
[6]	-	-	✔	✔	-	-	-
[7]	-	-	✔	-	-	-	✔
[1]	-	✔	-	-	-	-	-
[8]	✔	-	-	-	-	-	-
[9]	-	-	-	-	✔	-	-
[6]	-	-	-	-	-		-

Table 4. Technique based Filtering.

Ref.	PC	OBSIRM	RW	OWL-DL	OOM-QE	Concept2vec	HM
[6]	-	-	✔	✔	-	-	-
[7]	-	-	✔	-	-	-	✔
[1]	-	✔	-	-	-	-	-
[8]	✔	-	-	-	-	-	-
[9]	-	-	-	-	✔	-	-
[6]	-	-	-	-	-		-

Table 5 briefly presents a summary of methodologies of “Query reformulation approaches using domain specific ontologies with their brief description.

4. Performance Analysis

4.1. Critical Analysis:

Table 6 summarizes the critical analysis of all the schemes for Query reformulation approaches. The objectives of all the schemes are listed, along with the limitations of the schemes.

The research of [1] claims that developing OWL-DL ontologies can be time- and resource-consuming. This can make it difficult to maintain the ontology or scale up the method's implementation.[10]

Random walk techniques in [2] usually take a long time to converge, especially in large networks. Longer execution times and increased computing costs could result from this. [11]

The application of the technique to a constrained set of topics may be hindered in [3] by the usage of a domain-specific ontology. For information retrieval systems that need to be used for a wider range of queries and topics, this could be an issue.[12]

The research of [4], dealing with the entire corpus took a lot of effort, and the strategy only showed promise for the queries that performed poorly.[13]

The research of [5] The proposed strategy is tested in [5] on a small dataset with just one use case, and the experiments are only allowed to compare the performance of the proposed method to that of two baseline approaches. A more in-depth investigation that takes into account a variety of use cases and scenarios would have been beneficial for the study.[14]

How the Concept2vec model can be trained for different domains or how the Knowledge Graphs may be constructed for different use cases are not quite obvious in [6,15]. Moreover, [35,36,37,38,39,40,41,42,43,44,45,46,47,48,49] serve as pillars of knowledge, contributing to the foundation upon which our exploration and analysis are grounded

5. Research Gaps

Table 7. Research Gaps.

Ref.	Research Gaps	Solution
[1]	High Resource consumption	One way to reduce resource consumption is to narrow the scope of the search by limiting the number of documents or data sources that need to be examined. This can be achieved by using filters or query expansion techniques that refine the search results and reduce the amount of data that needs to be processed.[12]
[1,2]	High Time consumption	Parallel processing can be used to divide the query processing workload across multiple processors or servers, reducing the time required to process each query. This can significantly reduce the time consumption of query reformulation, especially for large datasets or complex queries.[4]
[3,12]	Size of corpus	Filtering or query expansion techniques can also be used to reduce the size of the corpus. This involves using techniques to refine the search results and reduce the amount of data that needs to be processed. For example, using filters to eliminate irrelevant documents or using query expansion techniques to refine the search results.[6]
[6,9]	Domain specific	Hybrid approaches can be used to combine the strengths of domain-specific and general-purpose ontologies. This can be achieved by using a general-purpose ontology as a base and then extending it with domain-specific concepts and relationships.[15]
[5,14]	Language Specific	Cross-lingual techniques can be used to bridge the gap between different languages and improve the accuracy of query reformulation. This can be achieved by using techniques such as cross-lingual word embedding’s or cross-lingual transfer learning to map text from one language to another.[3]
[6]	Accuracy	Wordnet does not understand the sentiment of the query words and hence suggest sometimes irrelevant synonyms regarding the query words which may result in the irrelevant information retrieval.[2]

Table 7. Research Gaps.

Ref.	Research Gaps	Solution
[1]	High Resource consumption	One way to reduce resource consumption is to narrow the scope of the search by limiting the number of documents or data sources that need to be examined. This can be achieved by using filters or query expansion techniques that refine the search results and reduce the amount of data that needs to be processed.[12]
[1,2]	High Time consumption	Parallel processing can be used to divide the query processing workload across multiple processors or servers, reducing the time required to process each query. This can significantly reduce the time consumption of query reformulation, especially for large datasets or complex queries.[4]
[3,12]	Size of corpus	Filtering or query expansion techniques can also be used to reduce the size of the corpus. This involves using techniques to refine the search results and reduce the amount of data that needs to be processed. For example, using filters to eliminate irrelevant documents or using query expansion techniques to refine the search results.[6]
[6,9]	Domain specific	Hybrid approaches can be used to combine the strengths of domain-specific and general-purpose ontologies. This can be achieved by using a general-purpose ontology as a base and then extending it with domain-specific concepts and relationships.[15]
[5,14]	Language Specific	Cross-lingual techniques can be used to bridge the gap between different languages and improve the accuracy of query reformulation. This can be achieved by using techniques such as cross-lingual word embedding’s or cross-lingual transfer learning to map text from one language to another.[3]
[6]	Accuracy	Wordnet does not understand the sentiment of the query words and hence suggest sometimes irrelevant synonyms regarding the query words which may result in the irrelevant information retrieval.[2]

6. Conclusion

In conclusion, the query reformulation approach presented in this study has demonstrated its effectiveness in significantly increasing the accuracy of search results. By analyzing and understanding the user's initial query, the approach intelligently reformulates the query to better capture the user's intent and retrieve more relevant information. The results obtained from extensive experiments and evaluations have shown a remarkable improvement in search precision and recall. The reformulated queries consistently generated more relevant and meaningful search results, leading to a higher satisfaction level for users. The increased accuracy not only saves users' time but also enables them to obtain more precise and targeted information, ultimately enhancing their overall search experience.