english language

The english language essay below has been submitted to us by a student in order to help you with your studies.

The Linguistic Analysis For Online Deception English Language Essay

Abstract—Previous studies on deception detection use verbal and nonverbal language but it is less predictable who is telling lies in online communication. This preliminary study, therefore, aims to investigate what linguistic features are used by truth-tellers and deceivers in Thai online chat. All chat messages were then analyzed by using frequency word count, categorizing word classes, and comparing other typical features. The results show that some word classes (e.g., gender particles and auxiliary verbs) and other typical features (e.g., initiation, conversation details, sentence types, and flirtation) seem to be a possible indicator of cyber deception.

Keywords-deception detection; online communication; Thai language; linguistic analysis

Deception in Online Communication

The rapid growth of Information Technology provides a new mode of communication as well as opens up new possibilities for online crimes: phishing personal information, sexual harassment, and other online deceits [1], [2]. These problems have led to a new research trend in Social Psychology and Information Technology (IT) to investigate deceptive cues in online communication, for example, e-mail, online discussion and instant messaging. Traditionally, research into deception detection has investigated face-to-face communication using verbal language (spoken language) and nonverbal language, such as gesture, eye movement, and facial expression [3]. However, it is less predictable who is telling lies in online communication because only the text-based message is available [4]. The deception detection in online communication then can be examined by analyzing verbal language (linguistic cues) [5]. Yet, most previous studies are limited to English language and these deceptive linguistic cues might not be applicable to other language. This preliminary study, therefore, attempts to expand the scope of analysis in another language by finding out what linguistic features are used by truth-tellers and deceivers in Thai online chat.

Linguistic AnalysIS for Online Deception

Previous studies on detecting deception in online deception center on English language use by analyzing linguistic features. The methods of analysis can be divided in two major categories: traditional linguistic analysis and automated linguistic analysis. The former analysis is influenced by the idea from Social Psychology that categorizing language use (parts of speech, linguistic style, positive/negative words) in communication can reflect the speaker’s inner feelings and motives [6], such as Linguistic-Based Cues (LBC) [7]. The latter one, influenced by Information Technology, emphasizes using computer-based software, such as Linguistic Inquiry and Word Count (LIWC) [8], [9] and Agent99Analyzer [8]. The limitation of both analyses is the overemphasis on English language at the surface level because of these previous research focuses: establishing deception detection software and measuring its reliability of detecting deception.

Linguistic Indicators Used by Deceivers in Online Communication

The findings in the previous studies reveal that some distinctive linguistic features can be an indicator of deception in online communication, as summarized in the following:

Fewer self-references (1st personal pronoun)the deceivers try to dissociate themselves from the online conversation or group discussion [9].

More negative emotion wordsthe deceivers may feel anxious and guilty of telling lies so the use of negative words display their inner feelings [10]

Less lexical diversity and complexityit is difficult for deceivers to use complex words when telling lies [5].

Frequent use of modal verbsthe deceivers use more modal verbs showing uncertain feeling [4]

Greater number of word count and turn-takingthe deceivers express themselves more in online communication so they give immediate response while the truth-tellers delay their response using more cognitive complexity [5], [7].

Strategic Scholarships for Frontier Research Network from the Office of the Higher Education Commission, ThailandThese deceptive indicators, derived from English language, may not be applicable to other language use. To the best of our knowledge, there has been no investigation on deception detection in Thai online communication. So this preliminary study aims to examine the differences in word quantity (e.g., total number of words, total number of turns, and average number of word per turn), word classes, and other typical verbal cues between truth-tellers and deceivers in Thai online chat. Three research questions are investigated:

Are there any differences in word quantity used by truth-tellers and deceivers in Thai online chat?

Are there any differences in word classes used by truth-tellers and deceivers in Thai online chat?

Are there any differences in other typical features used by truth-tellers and deceivers in Thai online chat?

Research Methodology

The participants were five voluntary undergraduate students (4 female, 1 male) studying in the Faculty of Sciences at Mahidol University, Thailand. All participants were Thai native speakers and were accustomed to chatting on MSN Messenger. The experimental study was done in September of 2009. These participants were informed that they would engage in two chat sessions with another random chat partner on MSN Messenger using pseudonyms. Those who agreed to take part in the study submitted the consent form before the data collection. The participants were also assigned to be either a truth-teller giving only truthful information or a deceiver pretending to be someone else in each 5-minute chat session. The participants then reported what role (truth-teller or deceiver) they had taken after completing the task. The data of each chat session was automatically recorded in the computer.

All chat messages (10 online conversations) were then analyzed to examine whether there are any differences in word quantity between truth-tellers and deceivers using frequency count [7]: the total number of words, the total number of turns, and the average number of words per turn. For linguistic features, the chat messages were also categorized based on Thai word classes to find out whether there are any linguistic differences used by truth-tellers and deceivers in Thai online chat. Iwasaki and Ingkaphirom [11] identify fourteen categories of Thai word classes as shown in Table 1.

Thai Word Classes

Categories

Thai Word Classes

Noun-related words

1. Noun 2. Pronoun

3. Demonstrative 4. Preposition

5. Classifier 6. Numeral

Verb-related words

7. Verb 8. Auxiliary verb

9. Negator

Modifying words

10. Adjective 11. Adverb

Miscellaneous words

12. Linker 13. Particle

14. Exclamative

However, the online communication had an influence on spelling variation [12], so these variants were counted as one word. Example 1 shows how Thai linguistic features were used to analyze truth-tellers and deceivers in Thai online chat.

Example 1 Anaylzing Chat Messages

In this online conversation, Chatter 3 was a deceiver pretending to be male, while Chatter 4 was truth-teller

Chatter 3: หวัดดีคับ [Wat Di-Khap]

Hi Noun SLP

Chatter 4: ดีค่ะ [Di-Kha]

Hi Noun SLP

Chatter 3: ชื่อไรคับ [Chue-Rai-Kap]

Name what? Noun QP SLP

N – Noun; SLP – Speech Level Particles/Gender Particles; QP –Question Particles

After that, all chat messages were then compared within groups and across groups (truthful vs. deceptive) to find out the similarities and differences of typical features used in online conversation such as initiation, sentence types, and the content of chat messages.

Results

Word Quantity

Table 2 provides a total number of words and turns used by truth-tellers and deceivers in Thai online chat. Participants as deceivers produced fewer turns than truth-tellers, while there was no clear pattern in the number of words produced. This resulted in the average number of words per turns being consistently higher for deceivers than truth-tellers.

Total Number of Words and Turns

No.

Sex

Types

Number

of words

Number

of Turns

Average

Word/Turn

Chatter 1

Female

Truth-teller

159

48

3.31

Deceiver

115

19

6.05

Chatter 2

Female

Truth-teller

88

19

4.63

Deceiver

109

14

7.78

Chatter 3

Female

Truth-teller

94

26

3.61

Deceiver

165

24

6.87

Chatter 4

Female

Truth-teller

149

24

6.20

Deceiver

94

14

6.71

Chatter 5

Male

Truth-teller

162

27

6.00

Deceiver

200

24

8.33

Thai Word Classes

Table 3 reports the descriptive statistics of Thai linguistic features used by truth-tellers and deceivers while participating in synchronous online chat. All participants as truth-tellers frequently used verbs, adverbs, and exclamative words. By contrast, the female participants as deceivers frequently produced pronouns, auxiliary verbs, and particles in online conversation.

A Categorization of Thai WorD Classes Used by Truth-Tellers and Deceivers in Thai Online Chat

Participants

Truth-tellers

Deceivers

Chatter 1

Chatter 2

Chatter 3

Chatter 4

Chatter 5

Total

Chatter 1

Chatter 2

Chatter 3

Chatter 4

Chatter 5

Total

Noun-related words

Female

Female

Female

Female

Male

Female

Female

Female

Female

Male

1. Noun

26/

16.35

10/

11.3

16/

17.2

26/

17.44

30/

18.51

108

16/

13.9

27/

24.0

29/

17.57

19/

13.57

30/

15.0

121

2. Pronoun

9/

5.66

5/

5.68

6/

6.38

2/

1.34

8/

4.93

30

13/

11.5

8/

7.33

10/

6.06

5/

3.57

8/

4.0

44

3. Demonstrative

--

--

4/

4.25

6/

4.02

4/

2.46

14

--

1/

0.91

1/

0.60

1/

0.71

3/

1.50

6

4. Preposition

4/

2.51

2/

2.27

2/

2.12

4/

2.68

3/

1.85

15

4/

3.47

2/

1.83

3/

1.81

--

2/

1.0

11

5. Classifier

--

--

1/

1.06

--

--

1

--

--

--

--

--

--

6. Numeral

2/

1.25

--

2/

2.12

2/

1.34

--

6

--

--

1/

0.60

--

2/

1.0

3

Verb-related words

7. Verb

37/

23.27

24/

27.2

14/

14.85

30/

20.1

40/

24.69

145

22/

19.1

18/

16.5

33/

20.0

13/

9.20

42/

21.0

128

8. Auxiliary Verb

5/

3.14

4/

4.54

3/

3.19

5/

3.35

7/

4.32

24

6/

5.21

8/

7.33

11/

6.66

2/

1.42

15/

7.50

42

9. Negator

4/

2.51

--

4/

4.25

6/

4.02

--

14

3/

2.60

2/

1.83

5/

3.03

3/

2.14

6/

3.0

19

Modifying words

10. Adjective

2/

1.25

3/

3.40

5/

5.31

7/

4.69

9/

5.55

26

7/

6.08

4/

3.66

4/

2.42

6/

4.28

13/

6.5

34

11. Adverb

9/

5.66

6/

6.81

6/

6.38

17/

11.40

13/

8.02

51

4/

3.47

5/

4.58

16/

9.69

8/

5.71

8/

4.0

41

Miscellaneous Words

12. Linker

13/

8.17

6/

6.81

2/

2.12

10/

6.71

7/

4.32

38

11/

9.56

8/

7.33

10/

6.06

10/

7.41

8/

4.0

47

13. Particle

29/

18.2

20/

22.72

16/

17.2

21/

14.09

28/

17.28

114

20/

17.3

21/

19.2

35/

21.21

17/

12.14

45/

22.5

138

14. Exclamative

19/

11.94

8/

9.09

13/

13.54

13/

8.72

13/

8.02

66

9/

7.82

5/

4.58

7/

4.24

10/

7.14

18/

9.0

49

Word Total

159/

100

88/

100

94/

100

149/

100

162/

100

652

115/

100

109/

100

165/

100

94/

100

200

/100

683

Note: the above number is the frequency count of word classes and below is percentage

Several Thai word classes remained uncertain since there were slight differences in total number of prepositions, classifiers, numerals, negators, and linkers used by truth-tellers and deceivers in online conversation.

C. Other Typical Features

Table 4 presents four major characteristics emerging from this experiment: (1) Initiationhow truth-tellers and deceivers initiated in online conversation; (2) Conversation detailsthe content of the online conversation used by these two groups of chatters; (3) Sentence typesthree kinds of sentences including affirmative, negative, and interrogative which were used in truthful and deceptive chat messages; and (4) Flirtationthe flirtatious behavior that occurred in the online conversation.

Other Typical Features

Characteristics

Truthful Chat Message

Deceptive Chat Message

1. Initiation

Participants as truth-tellers initiated the conversation by introducing themselves first.

Chatter 5 (T): เดคับ [Day-Khap]

Hi [variant] [GP-male]

Hi

Chatter 3 (T): ดีค่ะ คัยอ้ะ [Di-Ka-Kai-A]

Hi [gender particle-female] Who? PP

Hi. Who are you?

Chatter 5 (T): ตี้คับ ชื่อรายคับ [Ti-Kup-Chue-Rai-Khap]

Ti [GP-male]. Name what? [GP-male]

My name is Ti. What’s your name?

Participants as deceivers started with greeting in Thai online chat.

Chatter 3 (D): หวัดดีคับ [Wat-Di-Khap]

Hi [GP-male]

Hi

Chatter 4 (T): อ่ะดีค่ะ [A-Di-Kha]

PP Hi. [GP-female]

Hi

Chatter 3 (D): ชื่อไรคับ [Chue-Rai-Khap]

Name what? [GP-male]

What is your name?

2. Conversation

Details

Truthful Chat Message

Deceptive Chat Message

Truth-tellers generally talked about campus life, location, and activities in order to make friends and share common information.

Chatter 1 (T): แร้วเทอล่ะเรียนไหน[Laeo-Thoe-La-Rian-Nai]

And you study where?

And you? Where do you study?

Chatter 5 (D): เรียนจุฬาคับ[Rian-Chula-Khap]

Study Chula [GP-male}

I’m studying at Chulalongkorn University

Deceivers tried to get as much as personal information as they could, such as, asking for photo.

Chatter 4 (D): บุ๊คมีแฟนยังอ่ะ [Mi-Fan-Yang-A]

Book got a boyfriend ? [PP]

Book have you got a boyfriend?

Chatter 2 (D): สงสัยจะเนื้อคู่ [Songsai-Cha-Nueakhu]

Guess will soul mate

Guess that you will be my soul mate

3. Sentence Types

Truthful Chat Message

Deceptive Chat Message

Truth-tellers frequently used the affirmative sentences to share general information and wh-questions to continue online conversation.

Chatter 5 (T): เเง่วววววววววเรียนไหนอ่ะ [Ngaeo-Rian-Nai-A]

Study where? [PP]

So where do you study?

Chatter 3 (T): วิดเคมี มหิดล [Wit-Khem- Mahidon]

Sciences Chemistry Mahidol

Chemistry, Fac. Of Sciences at Mahidol

University

Chatter 5 (T): ที่เดียวกานเรย [Thi-Diao-Kan-Loei]

The same

The same university

Deceivers mostly used affirmative sentences to catch up the online conversation and used yes-no questions higher than wh-questions.

Chatter 1 (D): ก็ตอนไปหาเพื่อนที่มหาลัย

[Ko-Ton-Pai-Ha-Phuean-Thi- Maha-Lai]

When see friend at university

Chatter 2 (T): เคนเรียนทีไหนหรอ [Khen-Rian-Thinai-Roe]

Ken study where?

Ken, where do you study?

Chatter 1 (D): ไก่เรียนมหิดลนิ ชิป่ะ? [Kai-Rian-Mahidon-Ni-Chi-Pa]

Kai study Mahidol, right?

Kai, do you study at Mahidol, don’t you?

4. Flirtation

Truthful Chat Message

Deceptive Chat Message

Truth-tellers avoided flirtation by not sharing personal information with their chat partner. In this conversation, Chatter 2 indirectly refused to give her phone number.

Chatter 1(D): ขอเบอร์อ่ะไว้คุยกันทางโทสัพ

[Kho-Boe-A-Wai-Khui-Kan-Thang-Thosap]

Request number [PP], talk on the phone

Can I get your phone number? we can talk on

the phone.

Chatter 2 (T): งั้นเราไปแระ [Ngan-Rao-Pai-Lae]

I go

I have to go

All deceivers showed flirtatious behavior in Thai online chat by asking for too much personal information: photo, mobile number, and love life.

Chatter 3 (D): มีรูปมั้ยคับ[Mi-Rup-Mai-Khap]

Got photo? [GP-male]

Have you got a photo?

Chatter 4 (T): ทำไมต้องอยากดูรูปด้วยอ่ะ

[Thammai-Tong-Yak-Du-Rup-Duai-A]

Why want to look photo? [PP]

Why do you want to look at the photo?

(T) –Truthful; (D)-Deceptive; (GP)-Gender Particle; (PP)-Pragmatic particle

Discussion

The differences in word quantity reveal that the participants as deceivers used longer sentences with a delayed response. One possible explanation is that the deceivers take more time generating untruthful information when they were chatting online so the total number of turns used by deceivers was lower than the truth-tellers. However, previous studies of deception detection in English [5], [7] claim that deceivers express themselves more than truth-tellers in online chat. This might be a cultural difference between Thai and English internet users in online communication.

For linguistic features, the deceivers in this study frequently produced more particles and auxiliary verbs than truth-tellers in Thai online chat. It is possible that the deception strategy may result in frequent use of Thai word classes. The participants as deceivers used gender swapping so they might produce more gender particles and auxiliary verbs to convince their partners that they were pretending to be someone else in online communication.

The deceivers in this study also initiated the conversation and showed flirtatious behavior in Thai online chat. It seems that these chatters with deceptive purpose used flirtation to avoid talking about themselves in online communication, whereas the truthful chatters with realistic purpose provided more general information to keep the conversation going.

Conclusion

One of the main limitations was a small sample size so the linguistic features cannot be generalized as representative of truthful and deceptive characteristics in Thai online chat. Factors such as gender differences, task design, chatter’s role, and pairing (truth-teller with truth-teller; truth-teller with deceiver; deceiver with truth-teller; and deceiver with deceiver) may influence the data collection procedures and the results of the study. It should also be noted that the methodology centered on linguistic analysis so other conversational features (e.g., average pause and response latency) were not included in the study.

The results of the study suggest that examining linguistic features used by truth-tellers and deceivers in Thai online chat is likely to be productive. If consistent differences in the use of linguistic features between truth-tellers and deceivers can be identified, a Natural Language Processing application for detecting deception in Thai online chat can be designed from the findings in this preliminary study, some of the features distinguishing language use by deceivers can be programmed relatively easily, notably, word class differences and sentence types. Other features are more problematic, either because automated detection is problematic (e.g. the content of the conversations) or because it is difficult to determine a baseline (e.g. word count).

This study therefore shows that linguistic analysis of Thai online chat has potential for automated deception detection in Thai. However, given the low generalizability of the small sample in this study, we intend to continue our study on cyber deception with a greater number of participants and a deeper level of data analysis (e.g., discourse analyses and corpus-based analysis) to get a better understanding of deception in Thai computer-mediated communication.

Acknowledgment

We would like to thank the Office of the Higher Education and Commission, Thailand for supporting by a grant fund from the program Strategic Scholarships for Frontier Research Network for the Ph.D. Program Thai Doctoral degree for this research. Our thanks also go to the science students who volunteered to participate in this experimental research.


Request Removal

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please click on the link below to request removal:

Request the removal of this essay


More from UK Essays