The exploration of fuzzy linguistic research: A scientometric review based on CiteSpace

2.1 Co-citation analysis

(1) Data collection for using CiteSpace

Firstly, WoS is selected as the data source for bibliometric analysis because it is a huge platform that provides an access for readers to know specific information about the articles published in roughly 12,000 leading journals worldwide, which includes the Science Citation Index Expanded (SCI-EXPANDED), the Social Sciences Citation Index (SSCI), and the Arts & Humanities Citation Index (AHCI) databases [61]. Then, by searching the titles “fuzzy” and “linguistic”, 1114 publications related to fuzzy linguistic research are found. In order to eliminate “noise” that may affect the accuracy of research results, the papers published in some databases such as Conference Proceedings Citation Index-Science, Emerging Sources Citation Index and Book Citation Index– Science are filtered out. Ultimately, there are 648 papers about fuzzy linguistic research which were published between 1975 and 2018. The period (1975–2018) is chosen for three major reasons: (1) based on the records on WoS, the literatures published before 1975 were limited. (2) the papers published in the recent 43 years which are related to fuzzy linguistic research reflect the development of this field. (3) the data were retrieved on October 1st, 2018. In this study, the latest papers on WoS were also taken into consideration.

(2) Cluster network and the most cited articles

Co-citation can be explained as the frequency of two earlier literatures being cited together by the later literature [54]. In order to better understand the mechanism of specialty development, co-citation analysis is introduced as a useful tool which identifies prominent journals, articles and authors. The co-citation analysis of fuzzy linguistic research can be obtained by utilizing CiteSpace, which can be seen from Fig. 2. The articles with co-citations are divided into several clusters to make it easier for readers to know the main research areas. Detailed results about the main clusters and the most cited articles with co-citation frequency are shown in Tables 1 and 2, respectively.

Figure 2 shows the co-citation network based on the 648 references between 1975 and 2018. The corresponding data can be seen from Table 1 which lists top 6 clusters of this research area. In Table 1, the size means the number of publications in the cluster. The LLR (Log-Likelihood Ratio) is an algorithm to calculate each label and it is also the core concept that summarizes each cluster. Moreover, the silhouette score is used to test the homogeneity of clusters. Basically, if a silhouette is greater than 0.5, then the corresponding result is reliable. The largest value of a silhouette is 1, so if the silhouette of a cluster is near to 1, the result is considered to be more reliable. According to Table 1, all of the silhouette scores are above 0.8, which suggests a reliable quality of each cluster. The largest and the latest cluster is #0 cycle tour with 49 member references. Besides, most of the clusters’ mean years are before 2010, #6 is relatively older than other ones.

By knowing the top 6 clusters in fuzzy linguistic research area, the top 10 cited articles with co-citation frequency of over 40 times and their clusters are demonstrated in Table 2. According to Chen et al. [13], the most cited papers have significant contributions to certain research area. Therefore, it can be seen from Table 2 that the most cited article is from #0 with both 137 citations. This article is “Hesitant fuzzy linguistic term sets for decision making” [51] which is mentioned in section 2.1 as the most cited paper in Phase Three. Hesitant fuzzy linguistic information can be flexibly utilized to deal with experts’ uncertain evaluation and rich expressions in decision making.

Also, it is clear to see that the most cited articles are almost from #0, except for the fifth paper which is from #1. Therefore, multi-hesitant fuzzy linguistic information dose made great contributions in fuzzy linguistic research area. When talking about hesitant fuzzy sets, the second paper in Table 2 is of great necessity to be discussed. Torra [60] proposed hesitant fuzzy sets and introduced some basic operations. Numerous decision-making and evaluation methods therefore can be further developed. Moreover, Rodriguez et al. [52] and Zhu and Xu [79] proposed new models by utilizing comparative linguistic expressions and considering about preference relations. Liao et al. [42] and Liu and Rodríguez [44] investigated hesitant fuzzy linguistic terms sets and their applications to multi-criteria decision making. Xia and Xu [67], Wei et al. [65] and Beg and Rashid. (2013) analyzed and compared different information aggregations, then proposed aggregation methods which can be introduced in real-life decision making. Being different from the scholars above, Martinez and Herrera (2012) focused on the 2-tuple linguistic model and its extended forms, applications and challenges in decision making, which has also significantly impacted fuzzy linguistic research area.

The number of publications about fuzzy linguistic directly reflects the development process of this research field. Besides, it is also the obvious evidence to demonstrate whether this research topic is gaining increasing attention of scholars around the world. In this study, the articles on WoS that investigate fuzzy linguistic research between 1975 and 2018 are analyzed in detail, including the number of articles and the most cited articles in different development stages.

Figure 1 illustrates the number of publications from 1975 to 2018. It is obvious that the literatures about fuzzy linguistic research are increasing in the recent 43 years. In order to analyze the change of publication amount, this paper divides the whole period into four phases.

Phase One (1975–1994). As shown in Fig. 1, there was a small number of publications related to fuzzy linguistic in the early stage. According to the records on WoS, the average number of papers published in each year between 1975 and 1994 was just 2.3. However, many papers published in this phase have significantly impacted the further research, especially the paper entitled “Experiment in linguistic synthesis with a fuzzy logic controller”, which was written by Mamdani and Assilian [46]. This was an attempt to explore the possibility of human interaction with a learning controller. Besides, other papers related to fuzzy linguistic research mainly focus on investigating decision making (Tong and Bonissone, 1980) [36] and fuzzy linguistic quantifiers [33, 36]. In this phase, Mamdani and Kacprzyk made great contributions for the development of fuzzy linguistic research.

OPEN IN VIEWER

Fig.1

The number of publications about fuzzy linguistic research from1990–2017.

OPEN IN VIEWER

Fig.2

Cluster network of fuzzy linguistic research.

3.1 Productive countries and institutions

The most productive countries are identified by using the records of WoS, which makes the results more accurate and objective. Additionally, the most cited authors and collaboration network among institutions are analyzed by CiteSpace. Based on the results given by WoS and CiteSpace, further discussion can be seen below.

The top 10 countries with most of literatures about fuzzy linguistic research are listed in Table 3. It is obvious that China is the largest contributor with 310 publications, followed by Spain, the USA and India. The publications of these countries account for more than 70% of all the publications related to fuzzy linguistic research in the world. Although the rest of countries’ published papers are relatively less, they also made big contributions in this research area.

Figure 3 shows the country collaboration network in fuzzy linguistic research. Combining with Table 3, the node size in Fig. 3 indicates the publication number of each country. Besides, the links among these countries construct a geographic collaboration network. It can be found that the prolific countries in fuzzy linguistic research field have more collaboration with other countries. However, the thinness of the links implies a low level of cooperation.

OPEN IN VIEWER

Fig.3

A visualization of the country collaboration network.

Moreover, the number of publications about fuzzy linguistic research in each country/region is also related to the amount of institutions and research funding. The top 10 institutions which made great contributions to fuzzy linguistic research are listed in Table 4. There are 507 institutions published literatures about fuzzy linguistic research. Among these institutions, The University of Granada occupies a prominent position of this research field with 83 publications. Universidad de Jaén also has significant contributions to fuzzy linguistic research with 33 publications, and both of these two institutions are from Spain. The rest of the institutions are mainly from China. As aforementioned, China is the most productive country in fuzzy linguistic research. However, there are over 17% of articles related to fuzzy linguistic research are published by the first two institutions in Table 4, and both of them are in Spain. While among the 10 prolific institutions, 5 of them are from China which published about 16% of the total amount of papers in this research area. The rest of the institutions are from Poland, Saudi Arabia and the USA, respectively. As a result, the most prolific institutions are from the most prolific countries. There are lots of institutions in China investigating on fuzzy linguistic research, so that China can be the most productive country in this research area. Nevertheless, although the institutions in Spain are less than which in China, The University of Granada and Universidad de Jaén have a large amount of research outputs, leading Spain to become one of the greatest contributors in this area.

Besides, Fig. 4 shows the significant collaboration network among institutions in the world between 1975 and 2018. The structure of this network indicates the maturity of the institution collaborations. If the structure is relatively tight and close, then the collaborations among the corresponding institutions are more mature. According to Fig. 4, there are more collaborations among the institutions in the same countries. Thus, it is necessary to improve the cross-national interactions, especially for those countries which significantly impact the development of fuzzy linguistic research. Moreover, the development of some top universities in Fig. 4 can be seen from Fig. 5 which illustrates the publication numbers of these institutions between 2009 and 2018. It can be found that the development of The University of Granada is stable in the recent 10 years. Obviously, even though there was no record for Sichuan University before 2014, its research outputs sharply increased after that year. The article number even reached to the highest point compared with which of other universities in 2018. Overall, The University of Granada and Universidad de Jaén started fuzzy linguistic research earlier than other universities. After 2014, there are more institutions focus on this research area, and many of them are from China.

OPEN IN VIEWER

Fig.4

The development of the prolific institutions.

OPEN IN VIEWER

Fig.5

A visualization of the institution collaboration network.

The top 10 prolific journals and their impact factors released in 2017 are listed in Table 5. Journal of Intelligent Fuzzy Systems is the most prolific journal among the ten journals, following with IEEE Transactions on Fuzzy Systems, Fuzzy Sets and Systems and Information Sciences which have published over 30 papers related to fuzzy linguistic research. Particularly, the impact factor of IEEE Transactions on Fuzzy Systems is the highest. This journal also made great contributions in fuzzy linguistic research area due to the large number of research outputs and its impact factor. In addition, many papers mentioned in Section 2 are published by the above journals.

Besides, the articles published by other journals also made considerable contributions to this research area, such as “Group Decision Making with Incomplete Fuzzy Linguistic Preference Relations” [1] published by International Journal of Intelligent Systems and “Multi-criteria decision-making methods based on the Hausdorff distance of hesitant fuzzy linguistic numbers” [66] published by Soft Computing. These articles also have been cited frequently by improving the further investigations in terms of dealing with decision makers’ preferences and uncertain information. Overall, not only the journals listed in Table 5 are the most prolific, the articles published by them are also of great significance in improving the development of fuzzy linguistic research.

The top 15 productive authors are listed in Table 6. According to the records on WoS, Herrera-Viedma and Herrera are the most productive authors in fuzzy linguistic research with 40 and 34 papers, respectively. It should be noted that the number of publications here is not the actual research outputs for each author due to the reason that some papers were filtered out based on the searching method in this study. However, most of the papers with significant influence and written by these authors are taken into consideration. In terms of affiliations, both of Herrera-Viedma and Herrera are the professors of the Department of Computer Science and Artificial Intelligence, the University of Granada. The main author collaboration network of fuzzy linguistic research is shown in Fig. 6. It can be seen from Fig. 5 that not only Herrera-Viedma and Herrera are the co-authors in many papers, they also collaborate with other scholars such as Xu, Liao, Martinez and Alcala who are also the productive authors listed in Table 6.

OPEN IN VIEWER

Fig.6

A visualization of the author collaboration network.

In order to further investigate the authors who have greatly impacted fuzzy linguistic research, a visualization of the author co-citation network is given in Fig. 7 to show the frequently cited authors in the recent 43 years. It should be noted that only the first author of a certain paper is considered in this study. Based on the author co-citation analysis provided by CiteSpace, there are 387 nodes and 1546 links in the network.

OPEN IN VIEWER

Fig.7

A visualization of the author co-citation network.

The top cited authors are Zadeh, Herrera and Yager. Since fuzzy set theory was proposed by Zadeh, various methods that utilize fuzzy sets, fuzzy logic, fuzzy algorithms, fuzzy system and etc. have been developed. Due to the reason that Zadeh provided the crucial theoretical basis of fuzzy linguistic research, he has become a best-known scholar in fuzzy mathematics. Based on the data on Google Scholar, Zadeh’s research outputs have been cited about 180,000 times in scholarly works. The paper “Fuzzy sets” [77] has received about 90,000 citations until September 2017.

In addition, Herrera and Yager also greatly impact fuzzy linguistic research area. According to the information on WoS, the article that introduced the 2-tuple fuzzy linguistic representation model [21] which is mentioned before has been cited for 1339 times. Herrera’s other two papers [20, 51] are identified as the frequently cited papers with 52 and 141 times, respectively in the nearly 5 years. Yager’s research outputs are also inspiring in this area. He analyzed the construction of multi-criteria decision functions from linguistic specifications in nature language based on fuzzy set theory [71]. Yager and Kacprzyk (2001) proposed basic thoughts and perspectives about the application of fuzzy logic for the derivation of data sets’ linguistic summaries. Besides, in order to apply fuzzy multi-criteria techniques to mobile APPs, Yager [72] pointed out the necessity to obtain aggregation operations on ordinal-based intuitionistic fuzzy subsets. In recent years, Yager focuses on the approaches of linguistic summarization and their extensions based on fuzzy sets [4, 7]. Comparing the 15 productive authors in Table 6 and the frequently cited authors shown in Fig. 7, even though there are still some differences, a high consistent correlation can be seen between the productive authors and the frequently cited authors.

4.1 Reference burst detection

In order to know the emerging trends of fuzzy linguistic research, the reference burst detection is utilized in this study. As shown in Table 7, there are 20 references with strong citation bursts in a certain period of time between 1975 and 2018. It is obvious to see that all the citation bursts started after 2000. The earliest reference citation burst in this research field was in 2002. This paper introduced a new fuzzy linguistic representation model on the basis of “Symbolic Translation” to deal with the loss of information led by other computation approaches [21]. The red lines represent the citation burst duration of each listed paper, so this paper’s impact lasted for a long time from 2002 to 2007. Besides, almost half of these articles were written by Herrera-Viedma listed in Table 7.

Combining with the analytic results in Table 6, it can be seen that Herrera-Viedma has significantly impacted fuzzy linguistic research area. His articles also received long citation bursts. For instance, the article entitled “An information retrieval model with ordinal linguistic weighted queries based on two weighting elements” [22] proposed an information retrieval (IR) model based on ordinal fuzzy linguistic method which made the user-IR system interaction more flexible. In the same year, the similar investigation [23] was published, which also received the citation burst from 2003 to 2009.

Moreover, the papers that focus on new aggregation and operational laws also gained much attention from the scholars in this research field. Herrera-Viedma et al. (2006) evaluated the information quality of websites by introducing 2 new majority guided linguistic aggregation operators. Xu [68] defined some operational laws of linguistic variables, and the new aggregation approaches were proposed. Also, he proposed some new aggregation operators and applied it to deal with the uncertain multiplicative linguistic preference relations in group decision-making process [70]. Moreover, the consensus model is one of the emerging research trends in recent years. 10 Herrera-Viedma et al. [29] proposed a consensus model to help experts in different backgrounds reach the consensus in group decision making. Cabrerizo et al. [9] also developed a consensus model on the basis of a fuzzy linguistic approach which can be utilized to deal with unbalanced linguistic term sets. With respect to dealing with unbalanced linguistic term sets, Herrera et al. [20] also developed a fuzzy linguistic method to address this issue. It is clear that the newest citation burst started from 2009 and ended in 2015, suggesting one of the emerging trends is to utilize fuzzy linguistic method as an efficient tool to deal with unbalanced linguistic information.

In order to know fuzzy linguistic research more directly, the keyword analysis provided by CiteSpace can be useful for readers to learn the specific hot topics. The time zone view of keywords is presented in Fig. 8. Many keywords and their publication or peak time on the bottom constitute the major topics of fuzzy linguistic research in the recent 24 years. As shown in Fig. 8, the big clusters are presented between 2000 and 2010. Combining with the aforementioned development phases of fuzzy linguistic research, these clusters are closely related to the articles published in Phase Two and Phase Three. Even though the number of publications in these phases are not as large as which in Phase Four, the theories and concepts have provided significant theoretical basis for the further research. As a result, there are many links between the early investigations and the late ones. Besides, from the labels in each cluster, it can be seen that fuzzy linguistic research has been transformed from general to specific.

OPEN IN VIEWER

Fig.8

Time zone view of keywords.

The hot topics before 2000 are “system”, “fuzzy set” and “design”. Sine fuzzy logic was proposed by Zadeh [77], it had been applied to linguistic synthesis. There are many research directions such as fuzzy systems [43, 63] and Boolean information retrieval systems [8, 37] were developed to make some theoretical basis more general for the further investigations. After that, various models based on fuzzy logic and linguistics were proposed to solve practical problems, leading “model” become the largest hot spot in fuzzy linguistic research. Additionally, “group decision making”, “information”, “aggregation operator”, “environment”, “word” and “preference relation” are the main hot topics between 2000 and 2010. According to the analyses in the last sections, it is obvious that the fuzzy linguistic models are utilized to solve the real-world problems, especially group decision making. IR methods and aggregation operations are proposed to further improve fuzzy linguistic models and make them more efficient to help decision makers with their qualitative decision making. In addition, not only do fuzzy linguistic models can be utilized in fuzzy environment, the fusion of preference relations also further optimizes the models to meet decision makers’ specific requirements.

The recent research hot topics are “term set”, “distance”, “variables”, “hesitant fuzzy linguistic term set”, “multicriteria decision making”, “weight”, “framework” and etc. Among these hot spots, hesitant fuzzy linguistic term set is broadly investigated. Liao et al. [42] analyzed and developed different kinds of distance and similarity measurements for hesitant fuzzy linguistic terms sets. Liu and Rodriguez [44] proposed a new representation of hesitant fuzzy linguistic term sets based on a fuzzy envelope which can be directly introduced in fuzzy multicriteria decision making models. Besides, Wei [66] developed the arithmetic and geometric aggregation operators on the basis of interval valued hesitant fuzzy uncertain linguistic information to deal with multi-attribute decision making problems. According to the results given by Fig. 7, one of the latest research topics is multi-criteria decision-making. Yu et al. [76] proposed two hesitant fuzzy linguistic harmonic averaging operators to deal with multi-criteria decision-making problems in fuzzy environment. Feng et al. [17] extended the applicability of hesitant fuzzy linguistic term sets and developed two novel methods based on the priority method and the idea of PROMETHEE. Therefore, from general theories to specific model improvements and applications, the hot topics of fuzzy linguistic research is dynamic in different time periods.