Keywords

1 Introduction: Does Emphasizing of Important Text Parts Help Understanding?

In recent years, we have become forced to read many documents in our workplace and daily life. However, we cannot read all documents carefully because of time limitation. Some recent researches, therefore, has investigated methods that can automatically summarize documents to aid the comprehension of long documents [1, 2].

A problem takes place after the extraction of important parts: we have to consider about how to display of the texts.

In some cases, less important parts are simply hided, while important parts are left. Even though some parts are evaluated as unimportant by the system, they still have certain information that might be crucial. So we should avoid hiding them.

Another way is visual emphasis that highlights important parts of texts without hiding unimportant parts.

Today we can extract important parts from texts by using some automatic software. Also we do not feel difficulty to emphasize important parts by applying conventional typographical techniques. The problem is that we are not sure about cognitive effects of emphasis of important parts: are they really helpful for comprehension or not?

The purpose of this paper is to confirm our expectations that the visual emphasis increases the degree of comprehension and shortens the reading time. We also examine side-effects of visual emphasis.

2 Preparation: Extraction and Emphasis of Important Parts

2.1 Automatic Methods to Detect Important Parts of Texts

For the experiment, we can consider the following three methods to automatically evaluate and extract important texts.

  • Method A: Extracting sentences containing terms that appear most frequently in the document.

  • Method B: TF-IDF method, which extracting sentences with a high Term Frequency Inverse Document Frequency (TF-IDF) sum.

  • Method C: Improved TF-IDF method [3]. Extracting sentences containing not only many instances of frequent terms, but terms that tend to appear together with the frequent terms, and terms evaluated by the TF-IDF measure.

We used Total Environment for Text Data Mining (TETDM) [4] to implement Method C.

In the experiment, we set 30 % as the ratio of compression of the text length, so that 70 % of original sentences will be evaluated unimportant.

Our research interest does not aim at the discovery of an excellent method to automatically evaluate and extract important texts. We use these three methods to confirm that the visual emphasis increases the degree of comprehension and shortens the reading time even if the extracting methods are any methods.

2.2 Ways of Visual Emphasis on Important Parts

We apply following three types of visual emphasis simultaneously on the parts evaluated as important (Fig. 1).

  1. 1.

    Important sentences are displayed with red letters.

  2. 2.

    The important sentences are enlarged from 11pt font-size to 13pt.

  3. 3.

    The most frequent terms used in the important sentences are typed in blue.

The purpose of this paper is to confirm that the visual emphasis increases the degree of comprehension and shortens the reading time. Therefore, we do not care about the difference of ways of visual emphasis on important parts (Fig. 2).

Fig. 1.
figure 1

(a) Text with visual emphasis and (b) plain text

Full size image
Fig. 2.
figure 2

The example of the admission tests

Full size image

3 Experiment: Visual Emphasis and Degree of Understanding

3.1 Source Texts and Applied Visual Emphasis

We select eight documents for this experiment. Those documents are used for admission tests on comprehension ability in Japanese (ISBN 978-4-86248-938-8) (Table 1).

Table 1. Text extraction method for each document
Full size table

For each document, we prepare two versions of appearance: texts with visual emphasis and plain texts without emphasis (Fig. 1).

3.2 Subjects and Order of Sessions

Twelve university students participate in our experiment. We assign the two document groups to two corresponding groups of subjects (Fig. 3). The order in which the subjects read the texts is changed randomly for each subject to eliminate the order effect.

Fig. 3.
figure 3

Example of a random document order presented to a subject

Full size image

First, each subject seat 45 cm away from a screen and is instructed to read the document on the screen quickly. After reading it, the document disappears from the screen. The subject answers four questions to check the degree of understanding of the contents.

3.3 Results and Discussion

Comprehension Performance.

Table 2 shows the result of the rate of correct answers. When the questions relate contents written in the emphasized area, the subjects answered more correctly. The attentions of the subjects were strengthened toward the emphasized area. This tendency was verified as significant with t-test (p < 0.05).

Table 2. Correct answers rate
Full size table

This suggests the proposed method can support reading comprehension for this problem type (Fig. 4).

Fig. 4.
figure 4

Means of correct answers out of the four questions

Full size image

Table 2, however, also shows the result of the rate of correct answers. When the questions relate contents of not-emphasized area, the subjects answered less correctly. The attentions of the subjects were strengthened toward the emphasized area, so they may not pay attention toward the not-emphasized area carefully. Though this difference is not proven as significant by using the t-test, we value this tendency as a negative effect (Table 3).

Table 3. Means and standard deviations of number of correct answers
Full size table

Reading Time.

The reading times of three documents with visual emphasis were slightly shorter than those of plain text (Fig. 5), but this difference is not proven as significant by using the t-test.

Fig. 5.
figure 5

Rate of change of reading time for each document

Full size image

Summarizing all of the experimental result, we find that the proposed method cannot shorten reading time greatly, but we can use our method to support reading comprehension for visually emphasized areas. That means proper selection of emphasis area is crucial to proper comprehension of readers (Table 4).

Table 4. Reading time and reading speed, normalized by the number of document characters
Full size table

4 Conclusion and Future Work

These results of this experiment suggest that the proposed method could support reading comprehension in visually emphasized areas.

We, however, also found a negative effect of visual emphasis that the readers less understand contents of not-emphasized area.

These two results suggest that we should increase the emphasized area on important parts of texts to support readers’ comprehension.

We, however, know that too many emphasized areas disturb reader’s comprehension.

In future work, we plan to investigate that appropriate quantity of the emphasized area of each user.

We plan to investigate that appropriate quantity of the emphasized area of each user. Therefore we will attempt the quantification of an effect of visual emphasis on important parts of texts. We plan to use two variables to quantify an effect of visual emphasis on important parts of texts of each user.

  • Variable 1: The ratio of the emphasized areas to the total text length.

  • Variable 2: The grammatical classification of the emphasized area, e.g. word, phrase, sentence and paragraph.

In future work, we plan to investigate that whether these variables are appropriate. Also we plan to look for the other variables.