A rhetorical segment in traditional abstract displaying a sign of particular function is frequently referred to as a move. One of the most common moves is the Background, Aim, Method, Results, and Conclusion (BAMRC). The objective of this paper is to investigate the move patterns of research article abstracts in the field of social sciences based on BAMRC moves. Using the Scopus bibliographic database, a total of 467 abstracts from 298 research journals in the field of social sciences were analyzed. The result showed a wide range of move patterns. The implication of the result of this study suggests the existing traditional abstracts in social sciences might not be sufficiently "informative" due to missing moves and due to various move orders. To this end, automatically mapping moves in traditional abstracts to sub-headings in structured abstracts can be a more challenging task, requiring additional procedures to resolve these types of compatibility issues. Future studies can compare this study's result to other fields or disciplines within social sciences in order to find a more precise nature of abstracts in the field of social sciences.
문장이나 문단 내에서 특정 기능을 하는 표현을 이동마디라고 하며, 여러 이동마디는 논문의 초록에서 쉽게 발견된다. 대부분의 이동마디는 보편적으로 배경-목표-방법-결과-결론(이하 BAMRC)의 순서로 많이 사용한다. 이 논문은 이러한 BAMRC 이동마디에 기초하여 사회과학분야의 논문 초록 중에서 나타나는 일정한 패턴을 분석 하여 연구한 것이다. 기존 연구와 달리 이 연구에 사용된 데이터베이스는 Scopus로 사회과학분야의 학술지 298개에서 467편의 초록을 샘플로 삼아 막대한 데이터를 사용하였다. 분석결과 이동마디가 넓은 범위에서 패턴을 나타내므로 결국 논문 초록에서 논문에 대한 정보를 얻는데 한계가 있는 것을 알 수 있다. 이 연구 결과, 기존에 있는 전통적인 사회과학 논문초록은 결국 이동마디가 없거나 다양한 이동마디의 패턴이 있어, 결국 논문초록 내에서 충분한 정보가 제공되지 않는다는 것을 보여준다. 따라서 기존 서술형식의 초록에 있는 이동마디들을 일정 양식을 따르는 초록으로 자동적으로 옮기는 것은 적합하지 않는 방법이며, 이러한 작업은 확인 절차를 밟을 것이 요구되므로 어려운 작업일 수 밖에 없다. 이 연구 결과를 토대로 추후에 진행되는 연구에서는 사회과학분야 논문 초록의 본질을 보다 구체적이고 정확하게 알아내기 위하여 사회과학 외의 다른 분야와도 비교하는 것이 도움이 될 것이다.
An analysis of abstracts in a given genre can be useful in understanding their nature and the way authors prefer to include information in them. Finding structural patterns in collected abstracts can indicate a particular functional type of information that is expected in a typical abstract in a given domain or field. A commonly used method to analyze abstracts is the move analysis. Here, a ‘move’ refers to segments of text that display a particular function.
In general, two distinctive forms of abstracts are used in academia. One is the traditional abstract and the other is the structured abstract. The major difference between the traditional abstract and the structured abstract is their typographical layout.
Compared to the traditional abstract, the overall argument for having a structured abstract is that it is more ‘informative’. The particular benefits of structured abstracts were investigated by Sharma and Harrison
For the field of social sciences, an informative abstract, whether structured or traditional, should contain information related to BAMRC categories. The BAMRC categories were supported by a number of previous analyses including those of Hartley
Moreover, Hartley and Betts
In general, traditional abstracts in the social sciences generally contain a mixture of informative or indicative abstracts. Each type of abstract has been widely discussed in literature.
A straight forward approach to generate an abstract in another form is by mapping the moves in traditional abstracts and turning them in sub-headings in a structured abstract. In this approach, an informative abstract could be more easily mapped into structured abstracts either by some type of machine processing or by a manual processing. Although many have overlooked the benefits of reproducing an abstract in another form, transforming a traditional form to a structured form can be desirable since it would allow a user to view selected moves for even more concise information at a glance. In particular, in an online environment, viewing only selected components of abstracts allows a quicker and a more flexible scanning through retrieved reference materials. For example, a user might want to view sentences related to the results of the research, instead of viewing entire abstracts.
With this ultimate goal in mind, this study is going to examine move patterns of traditional abstract based in the field of social sciences. In particular, the order of moves and distribution of moves will be examined extensively. Furthermore, the objective of this study is to examine whether move analysis can uncover the capability issues between traditional abstracts and structural abstracts.
This study intends to make a contribution with regards to understanding the structural characteristics of abstracts. Compared to previous studies on abstracts, this study conducted a move analysis based on a relatively large sample size. The majority of previous studies relied upon relatively small samples consisting less than 100 abstracts. Unlike other studies, in this study, a wide range of journals from the field of social sciences was selected while other similar studies focused on only on small quantity of journals. Generally, a larger sample size of data is more effective for an in-depth analysis of move patterns. This study intends to serve as a milestone for future studies that intend to conduct a similar research. Moreover, future studies could use the present study as a reference point within the field of social sciences or other fields.
One of the criteria of measuring well-written abstracts is the ability of researchers to speedily determine the articles’ relevancy for the information they need. From this point of view, various efforts have been made to improve the overall qualities of abstracts.
A number of previous studies have used MEDLINE (http://www.ncbi.nlm.nih.gov/pubmed), a bibliographic database for the medical domain, as a dataset for investigating the applicability of machine learning techniques. Agarwal and Yu
Attempts to promote the use of structured abstracts have been made by a number of researchers. For example, Zhang
Other studies examined the basic differences between moves in traditional abstracts and sub-headings of structured abstracts. In the field of library sciences, Khasseh and Biranvand
1. Problems/purpose/objective/research question/focus of study 2. Sample/population size/characteristics 3. Method (e.g., data collection procedures, intervention, research design) 4. Findings 5. Conclusions/implications/recommendations
In traditional abstracts, Khasseh and Biranvand
There has been a number of studies that analyzed moves in fields other than social sciences. Hyland
Pertaining to the field of social sciences, Kafes
However, there has been a lack of research that focuses on examining research article abstracts in the field of social sciences. An in-depth investigation of move patterns based on a variety of journal abstracts in the field could perhaps point to fundamental differences between the traditional abstract and the structured abstract. Such a finding could further support the weaknesses of the traditional form of abstracts as Hartley and Betts
Also, as previous research suggests, analyzing move patterns can be an effective method for understanding the underlying characteristics of abstracts in a particular field. For instance, Tseng
To conduct the present study, bibliographic citation database from the Scopus website (http://www.sciencedirect.com) was used. The search was placed with the keyword “attitude” to constraint the retrievable pool. As a result, a total of 467 research article abstracts from 298 journals in the field of social sciences were downloaded in a plain text file. These journals were published from 2012 until March, 2014.
A preprocessing step was undertaken in order to remove irrelevant tags from the datasets. Then, each sentence was manually tagged with a type of move. The details of labeling scheme are being discussed in the following section. To calculate the distribution of sentences in each move, all abstracts had to be segmented into sentences. A number of customized UNIX shell scripts were written to segment the abstract.
Using BAMRC moves, this study aimed to investigate the move patterns by observing the following:
· the degree of having all five BAMRC moves, · the move orders, and · the distributional characteristics of sentences within each move
To reduce ambiguity, descriptions of each of BAMRC move are provided.
During the process of tagging sentences, linguistic indicators were noted as they aided in accurately defining each move. They were also useful from the standpoint of classifying the sentences of traditional abstracts. Table 1 shows an example of common linguistic indicators identified for each move. In spite of common noticeable linguistic indicators, for this study, a comprehensive list of linguistic indicators for each move was not pursued as it was not a feasible option. Various inflected terms of linguistic indicators could be considered although not shown in the table. Sentences that do not belong to the above categories were grouped as the
A wide array of traditional journal abstracts was considered for the purpose of this study altogether. A partial listing of journals that have been used is shown in the Appendix. Table 2 shows the basic statistics of the dataset that were used for this study. Out of 467 journal abstracts that were examined in total, 415 were traditional abstracts and 52 were structured abstracts. Since this study critically examined only traditional abstracts, the 52 structured abstracts were discarded from the dataset. The traditional abstracts were from 298 different types of journals. The largest amount of abstracts from a single journal were from a journal called
For the purpose of critically analyzing moves, the qualitative experience in labeling sentences needs to be mentioned first. Initially, the following labeling scheme was used to examine the overall distribution of categories. Each sentence was tagged with move by placing numeric values in front of each sentence. For example, a move number 1 was tagged as <1>. Thus, for each move the following tagging scheme is being used:
· <1> for Background · <2> for Aim, · <3> for Method, · <4> for Results,· <5> for Conclusion, and· <4> for Undefined.
There were instances where categorizing a sentence based on a given set of moves lacked any clear indication. For these types of instances, the
Occasionally, one sentence indicated two moves. In fact, a move can vary in length, ranging from a phrase to multiple sentences as Swales and Feak
The BAMRC move order played a role in determining the move type of sentences since it was selected as a move order that is most likely to be found in the dataset. Moreover, as Swales and Feak
Often linguistic indicators became a contributing factor in determining the move. The linguistic indicator such as ‘purpose of study’ provides a strong clue in identifying the move type. By classifying one sentence with certainty, it aids in classifying the subsequent sentences since most sentences are expected to be in the progressive order. The subsequence sentence which starts with
Often it was more difficult to determine the move when the order of moves was not in progressive order or when a sentence did not contain a clear linguistic indicator. The reason sentences were tagged as the
A distribution of moves can indicate the degree in which all five BAMRC moves are present in social sciences abstracts. Using this, characteristics of abstracts in the field of social science can be obtained. Based on moves, frequency counts could be performed according to the abstracts and according to the sentences. Figure 2 shows the overall move distribution for the dataset. The left side (a) shows the frequency count of abstracts based on moves while the right side (b) shows the frequency count of sentences according to moves. Collectively, they represent the move distribution in a dataset.
The frequency count of abstracts according to moves suggests that, as a whole, all moves were important structural components of traditional abstracts. Figure (a) shows that 335 (81%) of abstracts contained some type of results. Thus, the
Compared to a frequency count of abstracts based on moves, a different result was obtained when a frequency count of sentences was performed. The results indicate that the amount of sentences in some moves were considerably larger than others. For example, the figure shows that there is a substantial sentence difference between the
Comparing both (a) and (b) in Figure 2, a greater number of sentences showed in the
The results shown in Figure 2 (a) can be compared to other studies such as in Hartley and Betts,
There are similarities among the studies. In Table 3, the percentage of abstracts that included the
In general, a wide range of move patterns were observed. Figure 3 shows the frequency count of orders that was performed. Not all move patterns are shown on this graph and some details have been omitted due to the lengthy list of move patterns. As shown in the left side of this figure, a wide variation of moves at a sentential level was noticeable. A total of 333 sentence patterns could be observed. The index number indicates the sentential-level move pattern. Equivalent to left hand graph, the right-hand graph in Figure 3 shows that most sentential-level move patterns are unique, having only 1 occurrence. The frequency count of index #331, #332 and #333 is 5. These are the most frequently occurred pattern.
Table 4 is to some extent equivalent to Figure 3, apart from the fact that the move pattern is shown in a block level. Since the list shows a wide range of sentential level moves, it was more convenient to remove the duplicate sentence tags to view only move patterns at a block level. For example, the move pattern <2> <3> <3> <4> <4> <5> can be reduced to <2><3><4><5>. In an ascending order, a frequency count of entire move patterns in a block level was performed. As shown in Table 4, a total of 76 block-level move patterns is found. Since the frequency of moves is counted based on block-level instead of sentential-level, the variation of move patterns is substantially less than Figure 3.
An interesting observation can be made from Table 4. The most frequently used move pattern is <1> <2> <3> <4> <5>, which is BAMRC. It is a complete sequential order that does not contain any missing moves. The frequency count of this pattern is 40, which is approximately 10% of total traditional abstracts. Although this is a small percentage in comparison to the overall amount of pattern, it does confirm that some traditional abstracts contain all five BAMRC categories and the ordering of moves in a sequential order.
The next frequency pattern is <2> <3> <4> and <2> <3> <4> <5>. The frequency count of move pattern <2> <3> <4> is 39, while the frequency count of <2> <3> <4> <5> is 37. In other words, a substantial amount of the abstracts opened with the
There are numerous instances where one or more moves are omitted in the abstract. Often the move order that contained one or more missing move is not in sequential order. For instance, the move pattern <2> <1> <3> <4>, which is not in sequential order, does not contain #5 either. A more common case is where move ordering of the sentences were not in complete sequential order due to missing moves. For instance, the move pattern <1> <2> <3> <4> is a partial sequential order due to the missing move # 5. The move pattern <2> <3> <4> <5>, is also in a partial sequential order due to the missing move #1.
In order to provide a more in-depth analysis, opening and closing move patterns were examined. As shown in Table 5, 272 abstracts opened with move <1> which is the
Unlike the opening move pattern, a more wide range of move pattern associated with closing of abstracts was found in the dataset. As shown in Table 5, the most frequent closing move pattern is <5>, which is the
Up to this point, by examining move patterns in detail, the structural characteristics of traditional abstracts were discovered in the field of social sciences. In terms of move patterns, the findings of this study imply that traditional abstracts in the field of social sciences might not be homogeneous. Although there was a strong tendency toward BAMRC move orders, other forms of traditional abstracts such as indicative abstract might follow different move orders. In effect, in the field of social sciences authors and journals are more likely to contain a mixture of different move patterns as the result suggests.
This study re-examined Hartley and Betts
In addition, the present study re-confirmed that missing moves in traditional abstracts seemed unavoidable as authors are left to write traditional abstracts in a free writing form. Regardless of the move distribution differences among the studies that were examined, this study suggests that some amounts of missing moves can be expected in a given field. Missing information indicated the way authors view pertinent information in a traditional abstract may differ from the type of expected information in a structured abstract.
Questions might be raised as to the way this study is relevant to the improvement of the abstract format for the field of social sciences. As aforementioned, Hartley and Betts
Yet, the analysis of move patterns in this study has some implications on the sub-headings of structured abstracts in the field of social science. In a sense, rather than selecting the sub-headings by experienced academics, move analysis is a more reasonable way to match and select the sub-headings for the structured abstracts. This study has demonstrated the commonalities of structural and functional components between the two types of abstracts. Altogether, this study confirmed that move ordering of a traditional abstracts tend to be consistent with the sub-headings in structured abstracts.
The type of moves in a given field or domain can raise a compatibility issue between structural abstracts and traditional abstracts. From the perspective of mapping, missing moves in traditional abstracts and the non-sequential move patterns would pose a challenge to traditional abstracts. In devising methods to map the sentences in traditional abstracts to structured abstracts automatically, additional procedures to resolve the incompatibility issues concerning moves in traditional abstracts to sub-headings in structured abstracts need to be investigated.
On the other hand, an informative abstract by definition suggests major common moves such as BAMRC. The move analysis might be an effective method in identifying appropriate sub-headings for a particular field. More compatible move types with sub-headings of structured abstracts could allow smoother conversion between two forms of abstracts. To this end, determining appropriate sub-headings for a particular field needs to be done by academics in a particular field.
By focusing on moves, this study explored ways to analyze the research abstracts in social science. This study’s result was compared to the results in Hartley and Betts
Some abstracts in the datasets did not follow the BAMRC move sequentially due to missing moves. Thus, relying on the location is not adequate as some sentences might not follow the particular order. It is an indication that some authors prefer to write abstracts more freely instead of writing in the BAMRC order. Some researchers even argued against this particular order. For instance, Noris
The present study analyzed move patterns in a particular datasets using different ways. A number of possible future studies shall be suggested. Firstly, a possible future research would be the investigation of pattern of moves with a specific type of traditional abstracts. For instance, a possible research of high interest could be the investigation of move ordering even further in order to identify additional type of traditional abstracts (e.g., indicative abstract). Such a suggestion is based on the assumption that patterns exist between the subtype of traditional abstract and the move order.
In addition, using a much larger sample size, a future study could be devised to compare move orderings and move distributions among a number of disciplines within the social sciences. The move distribution in this study differed considerably from the results in Hartley and Betts.
Moreover, a larger variety of different types of move patterns should be examined for the field of social science. A move analysis based on move orders and move distribution could be applied to other fields as well. It would be more intriguing to compare move orders in social sciences to other fields such as humanities in an attempt to discover the differences. Although this study compared the results to the applied linguistics by using Tseng’s
Lastly, methods to resolve missing information by extracting information from the paper’s contents need to be investigated. In particular, extractive techniques such as Aliguliyev
In a broader scope, the result of move analysis has practical applications in the areas of library and information science and could be used by students from this particular field. For example, students could be encouraged to examine different move patterns as a conscious effort in writing an effective abstract. Also, recognizing common patterns could perhaps become a feasible means to evaluate the quality of an abstract.
In conclusion, other findings that could be inferred from this study might also indicate that researchers are limited to accessing relevant abstracts; and relevant abstracts are not viewable to researchers, due to the limitations identified in the present study. Consequently, the results of this study also suggests the need for publications and publishers in social sciences to establish a common ‘writers guidelines’, similar to an ISO