|This page describes a feature in AmCAT 3.3|
|View other version: 3.3 - 3.4 - 3.5|
The output option Summary in the Query search is the first of three main functions of automatic content analysis in AmCAT. The Summary function is very useful to get a first impression of the content of the debate concerning your topics of interest in the documents. It lists all articles from the selected article set(s) that contain one or more of the search terms you entered in the 'Keyword search' field, as well as an overview of the textual context of the search keywords in each article.
The example shown in the screenshots on this page illustrates this. When you select the output option 'Summary' and click on the 'Submit' button below the output options, AmCAT provides you with the number of articles in which one or multiple of your search terms occur, followed by a list of these documents under 'Found # articles'. Of each article, a number of characteristics is displayed: the headline of the article, the date, the number of words, the medium (media outlet) and the context in which your search terms were mentioned in the text (with your search terms in red, so they stand out from the rest of the text). In this example, the search term "nuclear acciden*" is used. As you can see, a total of 659 articles was found that used the terms ‘nuclear accident’ or ‘nuclear accidents’. The first article in the list was published on March 13th, 1958 and mentions the search term at least once (since the words are printed in red once in the article preview). When you click the headline of an article in this list, AmCAT opens the full text article in a new tab, with the search terms highlighted in yellow, and some additional article details.
The Summary function has four additional possibilities: Assign as codingjob, ClusterMap, Associations and Save as Set (see Figure 6.3.1 upper right corner). Each of these possibilities is discussed below.
The Summary Assign as codingjob function will save the documents that were found in the article set using your search string as a new article set and will create a codingjob at the same time. To do so, you name the new article set, assign a coder to the codingjob, and select the unitscheme and articlescheme (see Figure 6.3.2).
The Summary ClusterMap function provides you with the possibility to create a venn diagram showing the overlap between different search strings. In order to be able to create a venn diagram, you need to enter at least two different search strings in the 'Keyword Search string(s)' field. A venn diagram is useful for the visualization of the extent to which different concepts are related to one another. In other words, a venn diagram presents associative frames.
Select ClusterMap, select 'Show in navigator' and click 'Ok'. AmCAT now displays a venn diagram with the conceptual overlap of your search strings. When you have a relatively small number of articles, the venn diagram displays little dots within larger colored shapes. These dots represent particular articles in the article set. When you click on one of these dots you will see the particular article in which your concepts overlap. If you have a relativey large number of articles, the venn digram displays a single large dot.
If we want to measure how often terrorism is related to either Israel or Palestine, for example, we can enter (1) Terrorism#terror*, (2) Israel#israel* and (3) Palestine#palesti* in the 'Keyword Search string(s)' field. Please note that each of these search terms started at a new line. Figure 6.3.3 shows this particular venn diagram. A total of 1190 articles was found. You can see that the majority of these articles (1016 articles) do not associate terrorism with either Palestine or Israel. Moreover, you can see that only two articles were found in which Palestine was mentioned, but not in combination with terrorism of Israel. Ninety articles mentioned all three concepts together. Below the venn diagram, AmCAT shows a table with the exact number of articles that mention your search terms, either alone or combined with (another) search term(s). Note that '0' indicated that the concept is not mentioned in the article and that '1' indicates that a concept is mentioned in the article).
When you select one of the other formats as the output of the ClusterMap rather than 'Show in navigator', either CSV, Excel or SPSS, you can save the quantitative data on which the venn diagram is based in another format and open and process the data in another program. The CSV file contains a datamatrix which shows, for each combination of your search strings, how often a particular combination occurs.
The Summary Associations function is somewhat similar to the ClusterMap. The main difference between these two functions is that the Associations function allows you to quantify the extent to which the overlap between different concept occurs. Select Associations and select a network outpuy by clicking on the drop down list. Here you can display the associations between concepts either as a table or as a network graph. With the table function you can create a table showing per combination of the search strings you entered in the 'Keyword Search string(s)' field what the chance is that concept B appears in an article given the fact that concept A is mentioned in the article. Concept A, the condition, is displayed in the rows, wheras concept B is displayed in the columns. The network graph function displays the same associations, but does so graphically rather than quantitatively. Each of search term is a node within the network. Associations from concept A to concept B (or reversed) are displayed with arrows. By selecting the 'Graph: include associations in label' (see Figure 6.3.4) the strength of the association is displayed in the graphic network. With 'Graph: treshold' you can select a minimal strength an association must have in order to show up in the network graph (weak associations are not displayed). By 'Number Format' you can select how you want to display the associations (proportion [presented here as .12 or .123] versus percentage [presented here as 12% or 12.3%]).
If we want to measure associations between terrorism, Israel and Palestine, for example, we again use (1) Terrorism#terror*, (2) Israel#israel* and (3) Palestine#palesti* in the 'Keyword Search string(s)' field (separate lines). We can display a table (see Figure 6.3.5) or a network graph (see Figure 5.6) either displaying the associations as proportions or percentages (we selected the latter in this example). In both Figure 5.5 and Figure 6.3.6 you can see that if the concept terrorism is mentioned in an article, the chance is 9% that the concept Palestine is mentioned as well. If, on the other hand, the concept Palestine is mentioned, the chance is 90% that the concept terrorism is mentioned as well. If the concept Isreal is mentioned, the chance is 91% that the concept terrorism is mentioned. We could thus conclude that the association between Palestine and terrorism has approximately the same strength as the association between Israel and terrorism.
The Summary Save as Set function enables you to save the results of your article search as a new article set in AmCAT. Select Save as Set and name the new article set in the 'Setname' box. You can use this function to easily store your own populations and/or samples of articles in AmCAT without having to search for the relevant articles every time you want to use them. If you want to add the results of your search to an already existing article set you can select this particular set by selecting it in the drop down list of 'Existingset'.