What Web Users Copy to the Clipboard on a Website: A Case Study

Ilan Kirsh

The Academic College of Tel Aviv-Yaffo, Tel Aviv, Israel

Keywords:

Clipboard, Copy and Paste, Web Analytics, Web Usage Mining, Human-Computer Interaction (HCI), User

Behavior, Personalization, Adaptive Websites, Text Simpliﬁcation, Text Summarization, Plagiarism, Search

Engine Optimization (SEO).

Abstract:

The clipboard is a central tool in human-computer interaction. It is difﬁcult to imagine a productive day-to-

day interaction with computers, tablets, and smartphones, without copy and paste functionalities. This study

analyzes real usage data from a commercial website in order to understand what types of textual content users

copy from the website, for what purposes, and what can we use such user activity data for. This paper advocates

treating clipboard copy operations as a bidirectional human-computer dialogue, in which the computer can

gain knowledge about the users, their preferences, and their needs. Copy operations data may be useful in

various applications. For example, users may copy to the clipboard words that make the text difﬁcult to

understand, in order to search for more information on the internet. Accordingly, word copying on a website

may be used as an indicator in Complex Word Identiﬁcation (CWI) and help in text simpliﬁcation. Users

may copy key sentences in order to use them in summaries or as citations, and accordingly, the frequency

of copying full sentences by web users could be used as an indicator in text summarization. Ten different

potential uses of copy operations data are described and discussed in this paper. These proposed uses and

applications span over a wide range of areas, including web analytics, web personalization, adaptive websites,

text simpliﬁcation, text summarization, detection of plagiarism, and search engine optimization.

1 INTRODUCTION

This study examines the content copied to the clip-

board by users of a commercial website and proposes

possible uses of copy operations data, in a wide range

of applications.

Web analytics tools are widely used in almost all

industries to better understand the interests, prefer-

ences, needs, and actions of website users. Premium

web analytics services track client-side user activity

at a very low level and go as far as to record all user

mouse movements. However, tracking what the users

copy to their clipboards is currently not part of the

web analytics toolbox.

This paper is organized as follows: Section 2 re-

views related work. Section 3 introduces the company

website examined in this study and the data that were

collected and used in this research. Section 4 presents

examples of textual contents that were copied from

the website and taxonomy of copy operations types

and subtypes. It also discusses possible user motiva-

tions to copy each type of textual content. Section 5

https://orcid.org/0000-0003-0130-8691

discusses ten different directions for potential applica-

tions and uses that can beneﬁt from copy operations

data. Section 6 concludes the paper.

2 RELATED WORK

Various studies show that web analytics is effective in

understanding how visitors use websites and in im-

proving and optimizing websites and web applica-

tions. This effectiveness has been demonstrated for a

wide range of industries, including, for example, on-

line news (Tandoc, 2015), online learning (Luo et al.,

2015), e-commerce (Hasan et al., 2009), and digital

marketing (Chaffey and Patron, 2012; J

arvinen and

Karjaluoto, 2015). Web analytics concepts, princi-

ples, and methods are described in detail in various

books (Kaushik, 2007; Kaushik, 2010; Dykes, 2014;

Alhlou et al., 2016).

The mouse cursor position is often used to esti-

mate which areas of the website capture the user’s at-

tention. Studies show a correlation between the posi-

tion of the mouse cursor on the screen and the user’s

Kirsh, I.

What Web Users Copy to the Clipboard on a Website: A Case Study.

DOI: 10.5220/0010113203030312

In Proceedings of the 16th International Conference on Web Information Systems and Technologies (WEBIST 2020), pages 303-312

ISBN: 978-989-758-478-7

303

eye gaze (Huang et al., 2012), with a higher correla-

tion when the user clicks the mouse or moves it (Chen

et al., 2001; Rodden and Fu, 2007). Mouse cursor

position information has been found to be an effec-

tive indicator of user’s attention in various web ap-

plications, including in e-commerce (Schneider et al.,

2017), web marketing (Tzaﬁlkou et al., 2014), online

surveys (Cepeda et al., 2018), task execution (Mil-

isavljevic et al., 2018), text reading analysis (Kirsh

and Joy, 2020b), and web search (Guo and Agichtein,

2008; Huang et al., 2012; Navalpakkam et al., 2013;

Rodden and Fu, 2007).

The cumulative attention of all the visitors in dif-

ferent areas of a web page can be visualized by

heatmaps (

Spakov and Miniotas, 2007; Lamberti and

Paravati, 2015; Lamberti et al., 2017). Attention

heatmaps are also offered by many commercial web

analytics services, which track, record, store, and vi-

sualize user mouse activity (Kirsh and Joy, 2020a).

Unlike the mouse cursor position and the mouse

activity, copy operations are hardly used to gain

knowledge on users. A recent study proposed and

demonstrated the visualization of cumulative user

copy operations using heatmaps, similarly to the

mouse cursor heatmaps (Kirsh and Joy, 2020a), but

commercial web analytics services do not offer such

heatmaps yet.

This study is the ﬁrst to analyze web user copy

operations of different types, to deﬁne a taxonomy of

copy operations by types and subtypes, and to propose

various possible uses of copy operations data. Initial

work on two of the ten proposed uses have already

started: Using copy operations in automatic text sum-

marization (Kirsh and Joy, 2020c) and in automatic

text simpliﬁcation (Kirsh, 2020a). Further work is re-

quired to explore the other eight proposed uses and

applications.

3 COPY OPERATIONS DATASET

Data for this study were collected by tracking and

recording copy operations of users on a commercial

website of a software company, ObjectDB Software.

The examined website contains hundreds of docu-

mentation pages with technical information on the

company’s products and tools, and about related sub-

jects. These web pages are visited daily by thousands

of visitors.

Many of the visits to the website are short. Soft-

ware developers, who make up most of the trafﬁc to

the website, use the website as a learning and knowl-

edge source. They frequently arrive from search en-

gines for short visits, after searching for speciﬁc tech-

nical solutions and code examples. When the desired

code example is found, it is often copied to the user’s

clipboard in order to be pasted and integrated into the

user’s software project. Although most of the copy

operations on this website are of code, this study ex-

plores also copy operations of text. Copy operations

of images and other resources, however, are out of the

scope of this study. Most of the uses that this paper

discusses are more relevant to copy operations of text.

Figure 1 shows the architecture that was used to

collect usage data. A reference to a Tracking Script

is embedded in the relevant web pages. As a re-

sult, every request for one of these web pages re-

turns a revised page that triggers an additional request

to load the Tracking Script from the Copy Tracking

Server. This script tracks clipboard copy operations

(by listening to JavaScript events) and reports them to

the Collector component in the server, which stores

the data (following anonymization) in a dedicated

database. The database contains the input dataset for

this study, while the Reporter component supports re-

trieval of copy operations based on speciﬁc param-

eters (e.g. type, length, and language of the copied

content).

Recording copy operations of web users is re-

lated to session recording, which is a common prac-

tice in modern web analytics of tracking and record-

ing user activity on websites, including mouse move-

ments and keystrokes. This practice raises questions

related to personal data protection and user privacy,

because of the risks of collecting sensitive personal

data intentionally or unintentionally (Gilliam Haije,

2018). Session recording does not necessarily require

prior user consent under personal data protection reg-

ulations, such as GDPR (under certain terms, as ex-

plained by the privacy and IT lawyer Arnoud Engel-

friet (Gilliam Haije, 2018)). Unnecessary personal

data should not be collected. If the collected data are

completely anonymized, which is a standard practice

in web analytics, then they are no longer considered

personal data (e.g. according to GDPR).

The copy tracking system was run on 231 pages

of the website. Usage data were collected for a pe-

riod of several months, ending in March 2020. During

this period, 654,399 page-views from 241,644 unique

visitors (estimated), and 53,131 copy operations were

recorded. This dataset of copy operations is the input

for this study.

Obtaining data for web usage mining research is

usually much more challenging than for web content

mining research, and therefore, most of such studies

focus on a single website. This study is no different,

and further work is required in order to explore copy

operations on other websites.

WEBIST 2020 - 16th International Conference on Web Information Systems and Technologies

304

Visitor

Researcher

o o o

+ Tracking Script Ref

Website Web Server

Tracking Script

Collector

Reporter



Copy Tracking Server

HTTP

Figure 1: High-Level Architecture of the Copy Tracking System.

4 WHAT USERS COPY TO THE

CLIPBOARD: EXAMPLES AND

A TAXONOMY

Users copy from websites different things for dif-

ferent reasons. This section presents real examples

of copied textual content from the studied dataset of

copy operations. The goal is to classify different types

of copy operations and to understand possible user

motivations for copying each of these types of con-

tent (although there is no attempt to cover every pos-

sible motivation). Each of the following subsections

presents one type of copied content, with a further di-

vision into subtypes.

4.1 Copying Single Words

We start by examining copying of single text words.

Table 1 presents the most frequently copied words

(taking into account only copy operations of a single

word). The examples of copied words are divided into

three categories of word types.

The ﬁrst category contains acronyms of technical

terms. The three examples provided represent central

concepts in the website domain knowledge. A possi-

ble reason for copying these terms into the clipboard

is that users need more information about these con-

cepts, so they copy the terms to search them on the in-

ternet. They might search these terms on their own or

as part of a longer (and more speciﬁc) search queries

that contain additional words.

The second category contains real-life words re-

lating to the speciﬁc examples used in the website tu-

torials. The tutorials explain how to develop simple

“Guestbook” applications in several environments.

Users can follow the step by step instructions in these

tutorials to create their own versions of these ap-

plications in their IDEs. The words: “Guestbook”,

“guest”, “GuestDao”, and “GuestListener” are the

suggested names for the created projects and classes.

Users copy these names from the tutorials to their

clipboards and then paste them in the relevant dialog

boxes in their IDEs.

The third category is complex words. These are

regular but less frequent words, which may be less

familiar for some non-native English speakers. A

plausible explanation as to why users copy complex

words rather than simple words, such as “of” and

“the”, is that they need more information about these

words and therefore paste them in online dictionar-

ies or search engines, searching for more information

or a translation. This is discussed and analyzed in

a follow-up study (Kirsh, 2020a), which shows that

copied words are more likely to be evaluated as com-

plex than uncopied words and words that are copied

more frequently are more likely to be evaluated as

complex than words that are copied less frequently,

by three distinct word complexity measures.

4.2 Copying Phrases and Clauses

In this subsection, we progress from single words to

sequences of several words. Table 2 presents the most

frequently copied phrases and clauses (which are not

complete sentences).

The phrases category contains short sequences of

two or three words. These are mainly technical terms

and concepts, similar to the acronyms category in sub-

section 4.1. Users may copy them to the clipboard in

order to search for more information on these subjects

(possibly with additional search words, in order to re-

ﬁne and focus the search). They may also be copied in

order to paste them later in texts that the users write,

What Web Users Copy to the Clipboard on a Website: A Case Study

305

Table 1: Examples of Frequently Copied Words.

Type Examples (the numbers of copy operations are indicated in the parentheses)

Concepts (Acronyms) JPQL (229), JPA (36), JDO (10)

Tutorial Words Guestbook (70), guest (32), GuestDao (20), GuestListener (14)

Complex Words criteria (36), transient (24), embeddable (21), embedded (20), persistence (17),

composite (16), redundent (14), retrieved (14), explicit (11), cascaded (11)

Table 2: Examples of Frequently Copied Phrases and Clauses.

Type Examples (the numbers of copy operations are indicated in the parentheses)

Phrases Composite Primary Key (86), JPA Criteria API Queries (53), Embedded Primary Key (39),

JPA Criteria API (32), The Sequence Strategy (23), JPA Named Queries (21)

Clauses This page covers the following topics (7),

A value will be automatically generated for that ﬁeld (5)

including summaries, documentation, answers to on-

line questions, etc.

The clauses category contains longer sequences

of words. Usually searching for longer sequences

of words on search engines is less effective, because

such search attempts may be too restrictive. The most

frequently copied clause “This page covers the fol-

lowing topics” is very general and not related specif-

ically to the subject of the website. It is more likely

that users copy such clauses in order to use them in

their own text (they may search them on the internet

ﬁrst, to verify that these clauses are commonly used).

The frequency of these copy operations are relatively

low, and the information that they provide may be of

less value in the context of this study.

4.3 Copying Sentences

From phrases and clauses, we proceed to full sen-

tences. Table 3 presents some of the frequently copied

sentences and sequences of sentences.

Some of the examples in Table 3 are single sen-

tences and the others are multiple sentences. In both

cases, it is quite unlikely that users copy these sen-

tences in order to use them in searches on the Internet,

as such long strings are not effective for searching.

It is much more likely that full sentences are copied

in order to use them in other texts, e.g. as citations

or in summaries. Possible uses for the copied sen-

tences could be in documentations, summaries, pre-

sentations, blogs, websites, answers on forums (such

as StackOverﬂow), or even private communications

between colleagues who work on a project together.

Copying full sentences is further demonstrated and

discussed in a follow-up study (Kirsh and Joy, 2020c).

4.4 Copying Translated Text

The website under investigation contains only texts in

English. Therefore, it was a surprise to ﬁnd many

hundreds of copy operations of text in other lan-

guages (mainly Spanish, but also French, Portuguese,

and other languages). It seems that many users used

Google Translate or other translation services to read

translations of the pages, and subsequently copied

translated text as well. Table 4 shows some examples

of copied translated text, which are relatively short

and could ﬁt the limited space in the table.

Google Translate usage is usually transparent to

websites and web applications, as ordinary requests

are sent from the browser to the website server, and

ordinary responses are sent back from the server to the

client. Translations are done by communications be-

tween browsers and Google Translate, in which web-

sites are not involved. However, client-side JavaScript

copy events are aware of translated text in the browser

(as shown to the users), and therefore, copying of

translated text is tracked and recorded.

The copied translations include phrases and

clauses (similar to the copied English phrases and

clauses, shown in subsection 4.2), and sentences (sim-

ilar to the copied English sentences, shown in subsec-

tion 4.3). Translated single words (as shown for En-

glish in subsection 4.1) have not been copied.

Three types of copied words have been described

in subsection 4.1 (see Table 1). Words of the ﬁrst two

types, acronyms and names from the tutorials, might

have been copied by Google Translate users, but since

these words are never translated, they cannot be de-

tected as copied translated text. Words of the third

type, the complex words, are translated (and therefore

WEBIST 2020 - 16th International Conference on Web Information Systems and Technologies

306

Table 3: Examples of Frequently Copied Sentences.

Type Examples (the numbers of copy operations are indicated in the parentheses)

Single

Sentences

The JPA Criteria API provides an alternative way for deﬁning JPA queries, which is mainly

useful for building dynamic queries whose exact structure is only known at runtime. (24)

Transient entity ﬁelds are ﬁelds that do not participate in persistence and their values are never

stored in the database. (9)

Changes to detached entity objects are not stored in the database unless modiﬁed detached

objects are merged back into an EntityManager to become managed again. (7)

Multiple

Sentences

The IDENTITY strategy also generates an automatic value during commit for every new entity

object. The difference is that a separate identity generator is managed per type hierarchy, so

generated values are unique only per type hierarchy. (63)

The @Id annotation marks a ﬁeld as a primary key ﬁeld. When a primary key ﬁeld is deﬁned

the primary key value is automatically injected into that ﬁeld. (16)

Table 4: Examples of Copied Translated Text.

Type Examples (the numbers of copy operations are indicated in the parentheses)

Phrases Requerimientos de plataforma (4), Fiabilidad y Estabilidad (3)

Sentences Adecuado para archivos de bases de datos que van desde kilobytes a terabytes. (1),

Las relaciones son campos persistentes en clases persistentes que hacen referencia

a otros objetos de entidad. (1)

could be detected if copied), but the translation to the

user’s preferred language eliminates the need to copy

these words for the purpose of translation.

4.5 Copying Programming Code

Elements and Fragments

All the types of copy operations that are discussed

above are related to copying text, rather than code.

Table 5 shows some of the more commonly copied

code elements and fragments, which are short enough

to be included as examples.

Code elements and fragments copied by users

range from single words (query literals, class names,

method names, ﬁles names, etc.), through lines of

codes (e.g. queries), to larger fragments containing

complete examples. This reﬂects the extent of assis-

tance that the copying user needs. In some situations,

only a single word is missing. In other situations, a

complete example is needed. It is reasonable to expect

that in most cases code elements and fragments that

are copied on the website are then pasted in the users’

IDEs and integrated into software projects. Single

code words may also be used in further searches.

5 POSSIBLE USES OF COPY

OPERATIONS DATA

As shown in the previous sections, copy operations of

website visitors can be tracked, stored, and analyzed.

This potential source of web usage data could be use-

ful in various applications. Each one of the follow-

ing ten subsections proposes and discusses a different

possible use of such data.

5.1 Understanding the Website

Audience’s Interest

Knowing the users and understanding how they use

the website is a core element of user-centered design

and a key to business success. Copy operations reveal

valuable information about the audience of the web-

site, which is not easily obtained by other means.

The frequencies of copying code indicate the

importance of speciﬁc code fragments to the audi-

ence of the website. For example, Table 5 shows

a high interest in the date and time literals, CUR-

RENT DATE and CURRENT TIMESTAMP. Simi-

larly, the frequencies of copying text indicate the im-

What Web Users Copy to the Clipboard on a Website: A Case Study

307

Table 5: Examples of Frequently Copied Code.

Type Examples (the numbers of copy operations are indicated in the parentheses)

Words CURRENT DATE (243), CURRENT TIMESTAMP (195), UPPER (112), LOWER (99),

persistence.xml (93), EntityManager (80), SUBSTRING (57), Transient (54),

orphanRemoval=true (47), YEAR{d’2011-12-31’}) (41), @GeneratedValue (36),

MEMBER OF (32), getSingleResult (25), PersistenceException (24), orphanRemoval (16).

Lines @NamedQuery(name=”Country.ﬁndAll”, query=”SELECT c FROM Country c”) (279)

SELECT c1, c2 FROM Country c1 INNER JOIN c1.neighbors c2 (158)

Fragments TypedQuery<Country> query =

em.createQuery(”SELECT c FROM Country c”, Country.class);

List<Country> results = query.getResultList(); (294)

portance of speciﬁc concepts. For example, Table 2

shows that certain types of primary keys (Compos-

ite Primary Key and Embedded Primary Key), and a

certain type of queries (Criteria API Queries) are of

particular interest to many users.

Existing web analytics tools, including premium

services that track every user’s mouse movement, fall

short of providing this precious information. This

study advocates the use of copy operations data as a

new tool in the web analytics toolbox. As discussed

in section 2, web analytics is effective in improving

and optimizing websites. Extending the boundaries of

web analytics to include copy activity of users could

make it even more effective.

5.2 Web Personalization and Adaptive

Websites

The insights copy operations provide about user inter-

est could also be used for personalization. Knowing

the user’s interest in real-time paves the way to com-

mercial opportunities, such as presenting customized

advertisements and special offers, as well as improv-

ing user experience by adjusting the website user in-

terface, for example, by presenting new relevant links.

An adaptive website can also change the content

that is presented to users based on their copy oper-

ations. In the context of a software company web-

site with technical documentation, if a user copies

only sample code fragments of a speciﬁc program-

ming language, content that is relevant to that pro-

gramming language should be presented where avail-

able, rather than content that is only relevant to other

programming languages.

5.3 Text Simpliﬁcation

Shardlow deﬁned text simpliﬁcation as “the process

of modifying natural language to reduce its complex-

ity and improve both readability and understandabil-

ity” (Shardlow, 2014). In lexical text simpliﬁcation,

complex words are replaced with simpler words with

similar meanings (Shardlow, 2014). This could be

done manually or automatically. In any case, the iden-

tiﬁcation of complex words is the ﬁrst step.

Word complexity is subjective. Information about

which words users of a given website consider com-

plex could be very helpful in improving the website

content and making it more readable and understand-

able. Some users move the mouse cursor while read-

ing to mark their reading position, so slowing down

or stopping near words might indicate difﬁculties in

reading or understanding (Kirsh, 2020b).

Table 1 shows examples of single word copy op-

erations. The third category in that table contains reg-

ular words that have been copied by the users of the

website. As discussed in subsection 4.1 and shown in

a follow-up study (Kirsh, 2020a), these are relatively

complex words, and apparently, some users have to

search their meanings on search engines or online

dictionaries. Therefore, simplifying these complex

words might be particularly beneﬁcial to the website’s

users.

5.4 Tooltips and Glossary

Similarly to complex words, users also copy profes-

sional and technical terms to the clipboard in order

to search for more information about them, either on

the website or externally on the internet. The “Con-

cepts” category in Table 1 and the “Phrases” category

in Table 2 contain terms that express concepts. Unlike

WEBIST 2020 - 16th International Conference on Web Information Systems and Technologies

308

complex words, these terms cannot be replaced with

simple words in order to improve readability.

However, there may be other possible techniques

to help readers. One option is to underline these terms

and display tooltips when the mouse cursor hovers

over them. Another option is to present a focused,

local glossary of relevant terms beside the main text.

Frequently copied terms may also be deﬁned and ex-

plained on other pages of the website. However, most

users arrive from search engines directly to speciﬁc

pages, missing deﬁnitions found on other web pages,

so they can beneﬁt from employing such techniques.

The frequencies of copy operations in the “Concepts”

and “Phrases” categories could indicate where on the

website such assistance is most needed.

5.5 Text Summarization

In the era of information explosion, text summariza-

tion is essential in bridging the gap between computer

capabilities to store texts and human abilities to read

them. A common approach to automatic text summa-

rization is to compose a summary from key sentences,

which are extracted from the original text. Various

statistical metrics can be used to evaluate the impor-

tance of sentences in the original text. The most im-

portant sentences, based on the results of this evalua-

tion, are selected and included in the summary (Sajjan

and Shinde, 2019).

However, copy operations data might provide a

better indicator of the importance of sentences. Ta-

ble 3 shows examples of sentences that are frequently

copied by users. As discussed in subsection 4.3, users

may copy sentences for various reasons, including for

summaries, and it is reasonable to expect that more

important sentences would be copied more frequently.

This novel approach is further developed and

demonstrated in a follow-up study (Kirsh and Joy,

2020c), which shows that users tend to copy impor-

tant key sentences more frequently and that a good

summary can be built from the most frequently copied

sentences.

Even if fully automatic text summarization is not

used, copy operations data can help a human summa-

rizer to make difﬁcult decisions regarding what to in-

clude in a summary. Sentences copied by users more

frequently are likely to be more important to the au-

dience of the website, and therefore, should probably

be included in summaries.

5.6 Reference Cards and Tip of the Day

Sentences that are copied frequently by users may

also emphasize important points. A further exami-

nation of the examples of copied sentences in Table 3

reveals that there are two different types of frequently

copied sentences.

The ﬁrst type is summary sentences, such as key

deﬁnitions and conclusions. For example: “Tran-

sient entity ﬁelds are ﬁelds that do not participate in

persistence and their values are never stored in the

database.”

The second type highlights important details that

have to be considered and may be less familiar, e.g.

caveats and edge cases. For example: “Changes to de-

tached entity objects are not stored in the database un-

less modiﬁed detached objects are merged back into

an EntityManager to become managed again.”

A summary can beneﬁt from both types of key

sentences. In addition, sentences of the second type

may be selected (manually or automatically) and

highlighted using reference cards (or cheat sheets) or

as part of a Tip of the Day system.

5.7 Detecting Plagiarism

The vast majority of copy operations are legitimate,

e.g. embedding short pieces of copied text as cita-

tions, with references to the source, is usually consid-

ered legal. On the other hand, some copy operations

violate copyright rules, e.g. copying and publishing

long sections of text, without explicit permission and

with no reference to the source, is illegal in most cir-

cumstances.

Copy operations of isolated words and phrases are

usually less concerning in this context. Plagiarism is

associated mainly with copying larger sections of text

and code (as well as images and other resources, but

this study focuses on copying textual content).

Simple Google searches of sentences and lines of

code that were copied frequently by the website users

reveal that many of them appear on other websites.

Some occurrences, such as answers to questions on

StackOverﬂow with proper references to the sources,

are not only legitimate but even desirable. On the

other hand, using large original sections of the web-

site content, without any attribution to the source, are

pure plagiarism.

This process of searching for unauthorized uses

of the content of a website, based on copy operations

data, can be either manual or automatic. An auto-

matic plagiarism detection implementation can track

suspicious copy operations, search for the copied text

on the internet frequently, and alert the website owner

What Web Users Copy to the Clipboard on a Website: A Case Study

309

when instances of plagiarism have been found.

5.8 Tracking User Progress in Tutorials

Table 1 shows several words that are frequently

copied from the website tutorials. The “Guest-

book”, “guest”, “GuestDao”, and “GuestListener” are

the suggested names for the tutorial project and its

classes. Users are expected to copy these names from

the tutorial’s web pages and paste them in their IDEs.

Similarly, users are expected to copy fragments of

code from the tutorials and integrate them into their

projects.

Tracking these copy operations (possibly in com-

bination with other indicators) may be useful in ana-

lyzing user progress in tutorials. It could be used to

detect possible breaking points, i.e. points at which

many users abandon the tutorials. Some reduction in

the number of copy operations throughout a tutorial,

due to users’ decisions to quit the tutorial, is expected.

However, if the copy operations data show an extreme

drop at a certain point it may indicate an issue with

that speciﬁc section of the tutorial and may require

further investigation.

5.9 Understanding Language

Translation Needs

As discussed in subsection 4.4, some users view web-

sites through Google Translate. The translation pro-

vided by Google Translate is normally invisible in the

website statistics and web analytics data. The website

is accessed ordinarily from the browser, and the trans-

lation is done by the browser and Google Translate.

In this case study, the copy operations expose a com-

munity of users that use Google Translate to translate

pages on the website to Spanish.

Examining copy operations data is a simple way

to detect the level of usage of Google Translate on

a website. This method has the additional beneﬁt of

detecting also speciﬁc interests of these users on the

website (see subsection 5.1 above).

Deciding on investing in the translation of a web-

site or speciﬁc web pages can take into account copy

operations data of Google Translate users. That data

is much more relevant for decisions on translation

than the distribution of users by country (which is

ordinarily available by web analytics services) since

many non-native English speakers do not need trans-

lation and may even prefer reading the content in the

original language.

5.10 Search Engine Optimization

Copy operations data can be useful in Search Engine

Optimization (SEO), i.e. in making a website perform

better in search engine results, and consequently in

increasing the trafﬁc to the website (which is usually

desired for commercial websites). Copy operations

indicate which topics are more popular among users

of the website. Investing in new content on these sub-

jects may be a cost-effective way to increase the trafﬁc

to the website.

As part of an SEO process, new pages can be cre-

ated for phrases that users frequently copy for the

purpose of further searches. A new dedicated page

can provide additional content about the subject of a

frequently copied phrase, with a title containing that

phrase, as well as other optimizations that help search

engines to establish the relevancy of the new page to

the phrase.

This could attract more visitors from search en-

gines. It may even be possible to catch a rebound

of users that leave the website with a phrase in their

clipboard to search for more information on the inter-

net. These users may ﬁnd themselves after that search

back on the same website, possibly on a page that was

created in this SEO process.

6 CONCLUSIONS AND FURTHER

WORK

This study explored what users of a website copy to

their clipboard. Copy operations of users of the web-

site were recorded and analyzed, and different pat-

terns were identiﬁed. Accordingly, copy operations

were classiﬁed into types and subtypes, and possible

motivations for each of the copying types were dis-

cussed.

Copy operations data may be valuable in various

situations. This paper proposes ten potential applica-

tions and uses that can beneﬁt from that data. Addi-

tional work is needed in order to explore these pro-

posed directions further and to develop and evaluate

the usefulness of these potential applications.

Most of the proposed applications focus on spe-

ciﬁc types of copy operations. Therefore, auto-

matic classiﬁcation of copy operations types (possi-

bly based on the taxonomy that this paper deﬁnes),

may be needed as the ﬁrst step in most of these ap-

plications. An automatic classiﬁer may also help in

estimating the distribution of copy operations among

the different types, and the frequency of copy opera-

tions of each type.

WEBIST 2020 - 16th International Conference on Web Information Systems and Technologies

310

Work on two of these applications, the use of copy

operations in text summarization and the use of copy

operations in Complex Word Identiﬁcation (CWI) and

text simpliﬁcation, has already started with promis-

ing initial results, as discussed above. Further work

is needed in order to explore the other eight proposed

uses and to extend the initial work on the two ﬁrst

applications.

Although it is reasonable to expect that the results

of this research are not unique for the selected web-

site and can be extrapolated to other websites, due to

the study focusing only on one website, further work

on other websites is needed in order to establish and

generalize these results.

REFERENCES

Alhlou, F., Asif, S., and Fettman, E. (2016). Google An-

alytics Breakthrough: From Zero to Business Impact.

John Wiley & Sons, USA.

Cepeda, C., Rodrigues, J., Dias, M. C., Oliveira, D.,

Rindlisbacher, D., Cheetham, M., and Gamboa, H.

(2018). Mouse tracking measures and movement

patterns with application for online surveys. In

Holzinger, A., Kieseberg, P., Tjoa, A. M., and Weippl,

E., editors, Machine Learning and Knowledge Extrac-

tion, pages 28–42, Cham. Springer International Pub-

lishing.

Chaffey, D. and Patron, M. (2012). From web analytics to

digital marketing optimization: Increasing the com-

mercial value of digital analytics. Journal of Direct,

Data and Digital Marketing Practice, 14:30–45.

Chen, M. C., Anderson, J. R., and Sohn, M. H. (2001).

What can a mouse cursor tell us more? correlation of

eye/mouse movements on web browsing. In CHI ’01

Extended Abstracts on Human Factors in Computing

Systems, CHI EA ’01, page 281–282, New York, NY,

USA. Association for Computing Machinery.

Dykes, B. (2014). Web Analytics Kick Start Guide: A

Primer on the Fundamentals of Digital Analytics.

Peachpit, Pearson Education, USA.

Gilliam Haije, E. (2018). Are session recording tools a risk

to internet privacy.

Guo, Q. and Agichtein, E. (2008). Exploring mouse move-

ments for inferring query intent. In Proceedings of the

31st Annual International ACM SIGIR Conference on

Research and Development in Information Retrieval,

SIGIR ’08, page 707–708, New York, NY, USA. As-

sociation for Computing Machinery.

Hasan, L., Morris, A., and Probets, S. (2009). Using google

analytics to evaluate the usability of e-commerce sites.

In Proceedings of the 1st International Conference on

Human Centered Design, pages 697–706, Berlin, Hei-

delberg. Springer Berlin Heidelberg.

Huang, J., White, R., and Buscher, G. (2012). User see,

user point: Gaze and cursor alignment in web search.

In Proceedings of the SIGCHI Conference on Hu-

man Factors in Computing Systems, CHI ’12, page

1341–1350, New York, NY, USA. Association for

Computing Machinery.

arvinen, J. and Karjaluoto, H. (2015). The use of web

analytics for digital marketing performance measure-

ment. Industrial Marketing Management, 50:117 –

127.

Kaushik, A. (2007). Web Analytics: An Hour a Day.

SYBEX Inc., USA.

Kaushik, A. (2010). Web Analytics 2.0. SYBEX Inc., USA.

Kirsh, I. (2020a). Automatic complex word identiﬁca-

tion using implicit feedback from user copy opera-

tions. In Proceedings of the 21st International Confer-

ence on Web Information Systems Engineering (WISE

2020), Lecture Notes in Computer Science, forthcom-

ing, Cham. Springer International Publishing.

Kirsh, I. (2020b). Using mouse movement heatmaps to vi-

sualize user attention to words. In Proceedings of the

11th Nordic Conference on Human-Computer Inter-

action (NordiCHI 2020), Tallinn, Estonia, forthcom-

ing, New York, NY, USA. Association for Computing

Machinery.

Kirsh, I. and Joy, M. (2020a). A different web analyt-

ics perspective through copy to clipboard heatmaps.

In Proceedings of the 20th International Conference

on Web Engineering (ICWE 2020), Lecture Notes in

Computer Science, vol 12128, pages 543–546, Cham.

Springer International Publishing.

Kirsh, I. and Joy, M. (2020b). Exploring Pointer Assisted

Reading (PAR): Using mouse movements to analyze

web users’ reading behaviors and patterns. In Pro-

ceedings of the 22nd HCI International Conference

(HCII 2020), Lecture Notes in Computer Science,

Cham. Springer International Publishing.

Kirsh, I. and Joy, M. (2020c). An HCI approach to ex-

tractive text summarization: Selecting key sentences

based on user copy operations. In Proceedings of

the 22nd HCI International Conference (HCII 2020),

Communications in Computer and Information Sci-

ence, Cham. Springer International Publishing.

Lamberti, F. and Paravati, G. (2015). Vdhm: Viewport-dom

based heat maps as a tool for visually aggregating web

users’ interaction data from mobile and heterogeneous

devices. In Proceedings of the 2015 IEEE Interna-

tional Conference on Mobile Services, MS ’15, page

33–40, USA. IEEE Computer Society.

Lamberti, F., Paravati, G., Gatteschi, V., and Cannav

A. (2017). Supporting web analytics by aggregating

user interaction data from heterogeneous devices us-

ing viewport-dom-based heat maps. IEEE Transac-

tions on Industrial Informatics, 13:1989 – 1999.

Luo, H., Rocco, S., and Schaad, C. (2015). Using

google analytics to understand online learning: A case

study of a graduate-level online course. In Proceed-

ings of the 2015 International Conference of Educa-

tional Innovation through Technology, EITT ’15, page

264–268, USA. IEEE Computer Society.

Milisavljevic, A., Hamard, K., Petermann, C., Gosselin, B.,

Dor

e-Mazars, K., and Mancas, M. (2018). Eye and

What Web Users Copy to the Clipboard on a Website: A Case Study

311

mouse coordination during task: From behaviour to

prediction. In International Conference on Human

Computer Interaction Theory and Applications, pages

86–93, Set

ubal, Portugal. SciTePress.

Navalpakkam, V., Jentzsch, L., Sayres, R., Ravi, S., Ahmed,

A., and Smola, A. (2013). Measurement and model-

ing of eye-mouse behavior in the presence of nonlin-

ear page layouts. In Proceedings of the 22nd Inter-

national Conference on World Wide Web, WWW ’13,

page 953–964, New York, NY, USA. Association for

Computing Machinery.

Spakov, O. and Miniotas, D. (2007). Visualization of eye

gaze data using heat maps. Elektronika ir Elektrotech-

nika - Medicine Technology, 115.

Rodden, K. and Fu, X. (2007). Exploring how mouse move-

ments relate to eye movements on web search results

pages. In Proceedings of ACM SIGIR 2007 Workshop

on Web Information Seeking and Interaction, pages

29–32, New York, NY, USA. Association for Com-

puting Machinery.

Sajjan, R. and Shinde, M. (2019). A detail survey on au-

tomatic text summarization. International Journal of

Computer Sciences and Engineering, 7:991–998.

Schneider, J., Weinmann, M., vom Brocke, J., and Schnei-

der, C. (2017). Identifying preferences through

mouse cursor movements – preliminary evidence.

In Proceedings of the 25th European Conference

on Information Systems (ECIS), pages 2546–2556,

Guimar

aes, Portugal. Research-in-Progress Papers.

Shardlow, M. (2014). A survey of automated text simpliﬁ-

cation. International Journal of Advanced Computer

Science and Applications(IJACSA), Special Issue on

Natural Language Processing 2014, 4(1):58–70.

Tandoc, E. C. J. (2015). Why web analytics click. Journal-

ism Studies, 16(6):782–799.

Tzaﬁlkou, K., Protogeros, N., and Yakinthos, C. (2014).

Mouse tracking for web marketing: Enhancing user

experience in web application software by measuring

self-efﬁcacy and hesitation levels. International Jour-

nal on Strategic Innovative Marketing, 01.

WEBIST 2020 - 16th International Conference on Web Information Systems and Technologies

312