General Instructions:
The annotation task consists of first looking at a group of words (i.e., representing as a group or cluster) and answer the first question. Then, look at two groups of words to answer the second question. The two questions are shown below. In many cases the second question may not appear.
The word cluster is represented in a form of a word cloud, where the frequency of the word in the data represents a relative size of the word in the word cloud. To understand the context of each word in the cluster it is important to hover over the a word, which will show the associated sentences for that word below. In addition to looking at the sentences, it is also needed to know the meaning of some words. Double clicking a word will allow you to open a browser tab to show the meaning in a google search results.
Please see the detailed instructions for each example below.
Detailed Instructions:
A word group is meaningful if it contains semantically, syntactically}, or lexically similar words. The example in the following figure has only numbers with two digits, hence, this is a meaningful group. The labels for this question can be one of the following:
Labels:
If the answer to this question is Yes, then the task would be to assign a name to the word cluster using one or more words. While assigning the name, it is important to maintain the hierarchy. For example, for the above word group in the Figure we can assign a name: semantic:number.
This needs to be written in the text box right below this question. While deciding the name of the cluster, the priority has to be to focus on semantics first, syntax second, and followed by any other meaningful aspects. While annotating it is also important to consider (i) their relative frequency of the tokens in the cluster, which is clearly visible in the word cluster, (ii) context in a sentence where the word appears in.
Note that the annotation interface also suggests a set of labels, which are automatically obtained. You can choose from them, however, please make sure that suggestive labels make sense. Otherwise introduce a new name depending on what property a cluster is representing.
The idea of this question is to understand whether two sibling clusters can be combined to form a meaningful super-cluster. In these two clusters, the left one is the same cluster annotated for the first question. The answer (i.e., labels) of this question is similar to Q1.
Labels:
Depending on the answer to this question the task is to provide a meaningful name similar to Q1.
Any comments?
Please write your comment in the text box that can help for further analysis.