Social Gender Construction in Political Context: A Corpus-Based Study of Lexical Differences across Genders

Anchored on the public stereotype towards the dichotomization of gender and the social gender construct theory, the study examines the gender differences in terms of lexical choice manifested by the selected 20 U.S. presidential candidates from the year 2012 to 2020 and presents the changes of each gender group in a male-dominated political context. The corpus of this study consists of 60257 lexis of 10 male politicians and 63095 words of 10 female politicians which are extracted from their announcement and campaign speeches. Findings obtained from the results based on the quantitative research design and the application of CLAWS, AntConc and chi-square test reveal statistically significant gender-based differences. The findings support that even though male and female presidential candidates have almost the same priorities of usage in general lexical categories, the lexical choice of male candidates was relatively close to feminization while women tended to be neutral. Finally, it further speculates that the candidates of differing gender enjoy a distinctive consciousness on the social gender construction in public discourse to challenge or neutralize public stereotypes of gender identity. In view of the findings, the study recommends a questionnaire survey to verify the inference on gender color displayed by different language variables and an extensive database to enable a greater validity of the results in the future.


Introduction
Language is one of the important mechanisms for human beings to express their thoughts. It is through language that people express their perception of the real world. In other words, language choice may be a diagnosis of the speakers' hidden feelings. Likewise, language may be related to gender. Since the mid-1960s, with the prosperous expansion of sociolinguistics and feminist movements, systematic studies on gender language have gone through about 60 years from their beginning. Following the development of feminism, different mainstream theories interpreting the relationship between gender and language have emerged in the field of linguistics. The leading theory among them is deficit theory/ deficiency theory which was reflected in Lakoff's influential work "Language and Women's Place" (1975), and shows that in a male-dominated society, the way women talk is considered as a violation of and deviation from the norm that refers to the way men talk. As a result of the inequality in power between men and women, the subsequent difference theory which emphasized different subcultures and different socialization processes experienced by men and women when they grew up gave birth to the modern gender theory or the social construction gender theory that pointed out the dynamic constructed relationship between social gender identity, and discourse practice. These diverse linguistic gender theories provided a good theoretical foundation for contemporary linguistic gender studies.
In the field of social science, more and more consensus showed that men and women tend to use different language forms and discourse styles. For example, women are more likely to use verbal interaction for social purposes, while men are inclined to transfer information (Brownlow, Rosamond, & Parker, 2003;Colley et al., 2004). In 2014, Hanafiyeh and Afghari conducted a quantitative empirical study based on the research results of Lakoff (1975) to test the use of hedges, tag questions, intensities, and empty subjects. The results verified the significant usage differences among the groups, and partially confirmed Lakoff's view on gender-constrained language. Mulac, Bradac, and Gibbons (2001) tested the two-culture model proposed by Maltz & Borker (1982) to explain the differences in language use between two genders. Sixteen language features have been proven to indicate the gender of the communicators according to the authors' investigation of the past empirical researches, and the results supported the relationship between assumed 15 linguistic features and stylistic dimensions.
In the past few decades, although there was a sharp increase in research on the nature and existence of gender differences, the lack of a unified conclusion on gender differences in language use has been established. Newman, Groom, Handelman, and Pennebaker, (2008) integrated the research methods of computer linguistics and corpus linguistics and supported a systematic difference existing in language of two genders that includes the content and the way to express. At the same time, some other scholars put forward a pretty different view that there was insufficient evidence to prove significant differences between two genders in the language use (Weatherall, 2002).

Political Language
In the field of gender language research, politics plays an indispensable role. Politics is the reflection of power (Chilton & Schaffner, 1997;Lakoff, 1990). Any act might become potential political act once it expresses authority (Chilton& Schaffner, 1997). According to them, it was impossible to implement politics without language on the one hand, and it was very likely that the application of language results in a broad sense of politics on the other hand.
Political language equals to political discourse or text. Linguists have different definitions on political language. Geis (1987) limited political language to political news reports. McNair (1999) believed that political communication was purposeful communication of political affairs, and all political discourses were included in this definition. On the other hand, Tian (2002) made a comprehensive description of political discourse in his article which includes three types. Broadly speaking, political language consists of a large range of politicians' discourses involving (but not limited to) political speeches, political debates, political advertising slogans and publicity materials on political events. This study focuses on public political speeches during the 2012-2020 U.S. presidential election campaigns, mainly including presential campaign announcement speeches and campaign speeches. Anyone who meets the requirements of the U.S. Constitution for presidential candidates can declare his or her candidacy for the coming year's election of president. The candidate's announcement speech is at a surfacing stage described by Trent and Friedenberg (2004), and it provides the candidate with a valuable opportunity to introduce themselves to voters and the news media. Although it is possible to change the initial impression, the first impression has a considerable influence on people's perception (Bromley, 1993, p. 36).

Gender Language in the Political Context
Although women have gained more democracy after the feminist movement, such as the right to vote and stand for election, and the number of women speaking in public has increased significantly, the struggle for power is a field naturally occupied by men. Politics is described as more masculine than other social practices (Brown, 1988). It has been claimed that male's personality characteristics in public's stereotype including strength, knowledge, self-confidence and directness were the prototype of politics (Huddy & Terkildsen, 1993;Kahn, 1996). And in this male-dominated field, women face the problem that the political environment emphasizes tough forms of competition and conflict. According to Liu (2012) women were latecomers of political participation, and the number of women participating in politics is still insufficient even in modern times when the slogan of gender equality has spread all over the world. Likewise, in Božić Lenard's (2016) study of the 113th United States Congress, when women were a minority in the US Senate and House of Representatives, they would have long been politically marginalized.
As women gain a higher position containing power and authority in the political world, their biological gender becomes more and more prominent that urges them to pay more attention to their behaviors including linguistic behaviors than men occupying the same position. Several studies have also verified that women are more easily challenged by gender cognition in political activities than men. The research conducted in 2005 supports that "inferences of competence based solely on facial appearance predicated the outcomes of US congressional elections better than chance" (Todorov, Mandisodza, Goren, & Hall, p. 1623). Nevertheless, Dolan (2008 believed that voters have a fixed prejudice against male and female candidates, which might lead voters to prefer men to women without being influenced by other factors. Jones (2017) held a same stand with Dolan that voters hold stubborn expectations towards the personality traits of ideal political leaders which are opposite to their impressions on women. She pointed out that female leaders are trapped by contradictory difficulties. If she demonstrates obvious feminine style, she might be considered to be lacking in leadership. If she performs like a man entirely, she has violated the public's positive expectations on women. In general, no matter how they behave, they are more likely to violate the fixed expectations of the public than male politicians. Jones' research (2016) confirmed the assumption that female politicians may try to present themselves in a way that minimizes gender prominence. In short, gender is clearly a strategic consideration in campaign communication even if it is not clear how gender affects voters' perceptions on Election Day.

Social Gender Construction Theory
In the past few decades, the development of language gender theory has experienced from the early biological determinism to the theory of gender difference, and the later theory of dynamic gender construction. The Social Gender Construction Theory was anchored on the concept of dominance as posited by Zimmerman, West andFishman (1970, 1980). The concept of dominance states that men obviously showed strong dominance in conversation style and strategy. Men often interrupt women's conversations and tend to control the choice and number of topics. Therefore, they draw the conclusion that gender was linked to social power, and the gender difference between men and women was the relationship of domination, oppression, subordination and domination of women. They asserted that men should be in a dominant position while women should be in a subordinate position in the conversation between men and women. In other words, the unequal power between men and women eventually leads to the imbalance between men and women in language communication. As Qian stated, different from the defect theory, the domination theory advocates disagree with that the language of women was inferior to that of men, but emphasized the domination of men rather than the deficiency of women (2004, p. 48).
Another faction that supports the difference in language use between genders is the difference theory (or called two-culture model), and Maltz and Borker (1982) were the first to put forward it who analyzed the causes of gender differences in language from the perspective of social language subculture. Cultural difference theorists denied the defect idea of women's speech style and considered it just as a difference from men's speech style. However, the theory ignores the existence of power relations and patriarchy, and cannot show the complexity and delicacy of gender issues. Just as Li (2001) questioned, since this theory is also based on the dichotomy between male and female, and makes a rigid statement on the verbal communication style of men and women, it implies that each gender has a single and fixed verbal communication style, which cannot explain why people of both sexes may adopt the way of talking to each other.
With the development and rapid changes of human society, the above-mentioned traditional gender language theories have difficulty in explaining some language phenomena that appear and exist today. In the 1990s, with the rise of women in various fields around the world and the emergence of the third wave of feminism, a diversified and dynamic view on gender came into being which is called social gender construction theory. Constructivists strongly oppose the traditional dualism, that is, the two genders are diametrically opposed, and they support that gender can be changed. If the dominant and difference theorists are concerned about how women differ from men in language use, then the constructionists' orientation is that how specific language behaviors contribute to the generation of men and women. Constructivists regard gender as a combination of behavioral elements, rather than behavior itself as a direct result of gender. Cameron has pointed out the core idea of the theory that it is the social practices that people engage in which create human gender characteristics, not the opposite (1996). To be simplified, if a person talks like a woman, it is not an inevitable result of being a woman, but a way of becoming a woman. Social members continuously enter new communities and participate in new social practice activities so that they must constantly adjust his or her behavior to suit the requirements of the groups to which they belong, thus forming their own gender characteristics. In the meanwhile, it is possible that they may refuse to carry out the acts stipulated by the relevant associations to show resistance to the current gender regulations. Gender constructivism explains gender differences in language more objectively and dynamically than others, and it is more convincing in interpreting some of the more significant gender language problems of the contemporary era such as differences between the same sex and similarities between the opposite sex.

Research Objectives and Questions
This study might contribute to a supplementary evidence for previous studies on language differences between the gender groups. At the same time, it also makes up for the gap of comparative empirical researches in the direction of gender differences and political discourse in the context of pursuing the higher political position. The current study generally aims to describe the political speeches of male and female US presidential candidates from the year 2012-2020 in terms of lexical usage. From the perspectives of lexical usage, the author looks forward to expound the role of language in distinguishing, maintaining or changing the political gender identity of candidates in campaigns, and to analyze the social causes of its formation, and verifies the significance of social gender construction theory. Specifically, it tries to answer the following questions: 1. What differences of lexical choice are evident in the political speeches of male and female presidential candidates?
2. How to explain these linguistic differences across genders from the perspective of social gender construction theory?

Sampling and Participants
The present study adopts the method of random sampling to select candidates and their speeches to produce more objectives and comprehensive research results. The author selected public political speeches during the 2012-2020 US presidential election campaigns mainly including presidential candidacy announcement speeches and campaign speeches. The corpus consists of 20 speeches by 10 female US presidential candidates and 20 speeches by 10 male US presidential candidates from 2012 -2020.

Instruments and Analysis
To analyze data, the present study applies several instruments which include text collator, CLAWS, AntConc and SPSS. Text collator is a free text editing software which is used to clean and collate raw text. The text should be cleaned up to avoid various nonstandard symbols or formats in the text collected through the network. The unprocessed texts might lead to errors or even failure in automatic part-of-speech coding, which would affect the retrieval statistics. Free CLAWS webtagger is a marker which is used for part-of-speech (POS) marking in the current study (Liang, Li & Xu, 2010, p.51). The C7 label set used in this study is the current standard label set. In this tag set, it classifies the words with different parts of speech, namely nouns, verb, adjectives, adverbs, articles, prepositions, conjunctions, numerals and miscellaneous, with a total of 137 specific coding categories. AntConc can be called an index tool. It enjoys a multiple function such as being used to observe many text and index lines, thus enabling the author to effectively analyze word collocation and class join.
At last, the researchers use 2x2 table of chi-square test in SPSS to test the significance of frequency difference between the grouped data. In corpus data analysis, chi-square test is an effective method to test the statistical significance of frequency difference (Liang, Li & Xu, 2010). Researchers judge whether there is significant difference between the two groups of data according to P value which can be calculated by 2x2 table of chi-square test. There is no statistical difference between two groups when p is more than 0.05. When p is less than 0.05, there exists statistically significant frequency difference between two groups. Consequently, the less the value is, the greater the difference is.
In addition, the remaining research tools like Word and Excel were used to aiding in recording data and drawing analysis.

Stastic Significant Difference in Lexicial Choice across Genders
The lexical categories which show significant gender differences include verb, preposition, noun, number and adverb. Through further comparison of the occurrence frequencies of these groups in the male and female corpora, it can be concluded that men are predominant in the use of verb, number, and adverb generally, while women show a preference on the use of preposition and noun rather than men. However, there is no significant difference between different genders in the use of determiner, adjective, conjunction, article and pronoun as a whole. In addition, in order to further find out how two groups differ in lexical usage specifically, the authors extract Chi-square test results and distributions of gender-based frequency of specific tags belonging to the C7 label set in CLAWS which show a significant difference (sig.<0.05).
Specifically, the significant gender frequency difference in the use of preposition mainly lies in the use of IF ("for") and IO ("of") across genders, and the usage frequency of women is significantly higher than that of male candidates. Of 22 tags belonging to noun defined by CLAWS, there are significant gender differences in frequency of use of four tags in total. Men are predominant in the use of singular noun of direction (ND1) and plural numeral noun (NNO2). Women, on the other hand, use more plural common noun (NN2) and singular locative noun (NNL1) than men. In the lexical group of numbers, only the chi-square test of MC1"one" shows a significant result, and men use it predominantly rather than women. Also, through further analysis, the author finds that in the word categories of adverbs, the significant gender differences mainly appear in the use of the following four kinds of adverbs including degree adverb (RG), wh-ever degree adverb "how to" (RL), general adverb (RR) and comparative general adverb (RRR) such as better. Among them, men use more intensifying adverb such as "very", general adverb and comparative general adverb than women, while women use more wh-ever degree adverb "however" than men. Although the authors have not found significant gender difference in the usage frequency of determiner as a whole (p=0.079>0.05), the significant gender differences in the use of some tags belonging to determiners appear. As shown in Table 2, they are namely DB (p = 0.030 < 0.05), DD2 (p = 0.011 < 0.05), and DDQV (p=0.000<0.001). Moreover, all three items are used more frequently by women. The tags refer to before determiner ("all", "half"), plural determiner ("these", "those"), and wh-ever determiner ("wherever", "whenever") respectively. Similarly, chi-square test shows that there is no significant gender-based usage difference of adjective in general (p=0.255>0.5). However, significant gender-based frequency difference of general superlative adjective (JJT) and catenative adjective (JK) exist, and men are predominant in the use of it rather than women. Besides, chi-square test shows statistically significant results in the frequency of CCB (p=0.031<0.05), CSA (p = 0.011 < 0.05), CST (p = 0.025 < 0.05), and CSW (p = 0.020 < 0.05). The above four codes respectively represent adversative coordinating conjunction "but"， "as", "that" and "whether". By further comparing the frequency of use of both genders in Table 2, it is found that men use more "but", "that" and "whether", while women use more "as". In the use of singular article, there is a significant gender frequency difference between men and women (p=0.036<0.05), although the authors do not find significant gender differences in the use of general articles in political discourse (p=0.645>0.05). According to the specific frequency analysis, men use more indefinite articles than women. As for the use of subjective wh-pronoun "who" (PNQS), 3rd person plural objective personal pronoun (PPHO2), 3rd person plural subjective personal pronoun (PPHS2), 1st person singular subjective personal pronoun (PPIS1), and 1st person plural subjective personal pronoun (PPIS2), there are significant differences between female and male politicians.
In conclusion, a statistically significant difference between the male and female speakers was recorded for 36 lexical variables out of the 137 specific coding categories. Out of the 36 tags, 23 lexical categories are predominantly used by male politicians in their discourses while 13 are more used by the female group.  Jones (2017) has shown that women prefer to use the high rate of social words which include verbs that suggest human interaction; thus, women's speech tends to be personalized and socialized. This assumption can also be confirmed by most studies based on different research backgrounds (Abbas, 2014;Argamon et al., 2003;Argamon et al., 2007;Argamon et al., 2009;Gleser, Gottschalk, & Watkins, 1959;Hamdi & Dabaghi, 2012;Koppel, Argamon, & Shimoni, 2002;McMillan et al., 1977;Mehl & Pennebaker, 2003;Newman et al., 2008;Schwartz et al., 2013). Thus, it is assumed that the predominant use of verb refers to a feminine speaking style.
In this study, the results show that men are more inclined to use verbs than women in the political election environment which overturns all the above research results, but confirms the latest results of Yu (2013) and Božić Lenard (2016) who conducted the gender language researches in political discourse. Male speakers in this study pay more attention to social and other aspects, while women pay more attention to tasks.

Preposition
Pennebaker (2013) explained that men naturally classify things and assign objects to spatial positions, so men use more prepositions. Argamon et al. (2007) also confirmed that prepositions are a powerful indicator of the style of male writers. Thus, it is assumed that few prepositions belong to women's fixed language style.
The result of the current research shows a remarkable significant gender frequency difference of preposition, which contradicts the research results of Božić Lenard who did not find gender-related preposition usage differences in political discourse (2016). Specifically, ten female candidates use more prepositions than men, and this means that the female candidates in the study do not show the traditional female language style in the use of prepositions, but moved closer to a masculine language style. Biber, Conrad, and Reppen (1998) found that female writers use a more complicated style, characterized by more pronouns and present tense verbs, while male writers tend to use a more informative style, characterized by more nouns and long words. On average, men use nouns more frequently than women. In this corpus, there are significant differences in the frequency of nouns used by different genders, and women use more nouns than men. More specifically, out of the 22 noun categories classified by CLAWS, women presidential candidates use more content nouns than men in the political campaign speeches which belong to notional words, and denote entities according to Radford (2009). This result once again overturns the findings on gender differences in language (Argamon et al. 2003;Argamon et al. 2007;Biber, Conrad, & Reppen, 1998;Koppel, Argamon, & Shimoni, 2002;Newman et al. 2008;Schwartz et al. 2013), but confirms the results of the study on gender differences in language in the same field (Yu, 2016). Yu (2016) studied gender differences in language use in political debates and suggested that female legislators use more nouns, which is a feature of a male language that conforms to the norms.

Number
The word class, number, has not received much attention in gender language studies, but it also plays a vital role in political speeches. First of all, the use of numbers reflects whether the speaker's point of view is neutral or critical. The numbers are more effective than quantifiers or abstract adjectives to prove their point of view, and also enhance the objectivity and credibility of politicians. Moreover, it hints the possible background investigation or research made by the speaker for the speech, which may enhance the audience's favorable impression of them. Several studies on gender language differences classify the use of numbers into the masculine language style (Argamon et al., 2003;Koppel, Argamon, & Shimoni, 2002;Manjavacas, 2015;Newman et al., 2008). In this study, men use numbers more frequently than women; however, the difference between male and female on the use of numbers is concentrated in the use of single cardinal number "one". This result conforms to the previous mentioned studies but it is different with the result of Božić Lenard (2016) that no significant differences in the use of numbers by political participants of different genders was identified.

Adverb
The difference in the use of adverbs in gender language was first proposed by Lakoff (1975). He believed that women tend to use intensive advisors, because it is related to women's tentative and powerless language. Since then, more and more research results have supported verified the hypothesis (Biber, Conrad, & Reppen, 1998;Eliason, 2007;Heath, 2006;McMillan et al., 1977;Mehl & Pennebaker, 2003;Mulac, Seibold, & Farris, 2000;Newman et al., 2008). Therefore, the authors assume that the preferred use of adverbs is the unique language style of women.
In the present study, men show a great tendency in its usage rather than female. This result is consistent with the findings of the studies of Yu (2013) and Božić Lenard (2016) that male politicians generally use adverbs more than female, but overturns the pervious assumption. Therefore, the authors speculate that male politicians use adverbs in their election speeches in order to construct a feminine linguistic style. Lakoff (1975) pointed out that female would like to use more empty adjectives, intensifiers and determiners than men, but Koppel et al. (2003) supported that the discourse containing a higher proportion of determiners reveals a masculine language style. In this study, although there is no significant gender-based difference in the use of general determiner, female speakers predominantly use before determiner, plural determiner and wh-ever determiner rather than men. This verifies Lakoff's view that women use determiners more often than men. Lakoff (1975) recognized in the early days that empty adjectives are usually used by women because they are usually considered more expressive and emotional than men. In the present study, there is no significant gender-based usage difference in overall adjective frequency, but it is found that the significant differences between male and female candidates exist in the use of general superlative adjective and catenative adjective. Both kinds of adjectives can act as intensifying emotion in the sentence. They not only enable the speaker to express the views and attitudes more incisively and vividly, but also can effectively stimulate the listener's emotion. Male candidates in the study use more frequently both categories than women. It can be speculated that they tend to soften the biological forceful image and create an approachable political image by controlling the adjectives used in their own words.

Conjunction
Many scholars have claimed that women use more conjunctions, especially "but" (Biber, Conrad, & Reppen, 1998;McMillan et al., 1977;Mehl & Pennebaker, 2003;Mulac, Bradac, & Gibbons, 2001). But in this database, men predominantly use the conjunctions including "but", "that" and "whether" rather than women, while women are predominant in the use of "as". Therefore, the authors come to the conclusion that male candidates support a feminine language style via controlling the use of conjunctions, and they show a great consciousness in keeping a complicated and logical speech style by using more "but", "that", and "whether". Also, it is speculated that women deliberately reduce the usage frequency of "but" in the political discourses in order to make their speeches smooth.

Article
Most scholars consider using few articles as a typical linguistic feature of women (Argamon etal., 2003;Brownlow, Rosamond, & Parker, 2003;Flekova & Gurevych, 2013;Gleser et al., 1959;Kapidžić & Herring, 2011;Koppel, Argamon, & Shimoni, 2002;Mehl & Pennebaker, 2003;Mulac & Lundell, 1986;Newman et al., 2008;Schler et al., 2006;Yu, 2013). Pennebaker (2013) explained that when men think and talk about objects in a clearer way, men will naturally use more nouns and thus more articles. However, similar to the results of Ludu (2014) and Božić Lenard (2016), the authors have not identified any gender-based differences in the use of articles as a whole in the political discourse. Even though in the use of indefinite articles, male candidates show a statically significant predominance rather than the female group. This cannot represent the tendency of male politicians in the use of feminine linguistic style since women use more nouns than men in the study. The relationship between the few use of articles and feminine speaking style is hard to make sense in this situation when it cannot be related with the use of noun; thus, it will not be identified as a symbol of femininization.

Pronoun
A large number of studies have reported that women use pronoun more frequently than men (Argamon et al., 2003;Argamon et al. 2007;Biber, Conrad, & Reppen, 1998;Gleser, Gottschalk, & Watkins, 1959;Heath, 2006;Koppel, Argamon, & Shimoni, 2002;Koppel et al., 2003;Mullac & Lundell, 1986;Newman et al., 2008;Schwartz et al. 2013). Besides, considering the overall category of personal pronouns, women are usually higher users, and higher frequency of the use of personal pronouns is also an indicator of women's speech style (Argamon et al., 2003;Gleser et al., 1959;Herring & Paolillo, 2006;Mulac et al., 1988). Holmes explained this phenomenon in 1993, saying that the linguistic means used by women are to emphasize the social linguistic universality of unity between the speaker and the listener. Friginal (2009) explained in his research that women use more personal pronouns in order to participate more actively in the interactive process. Jones (2017) gave a subjective explanation for this phenomenon. Her research results showed that women use more pronouns and social words which include all non-first-person-singular personal pronouns and verbs that involves human interaction than men. She believed that the formation of female speech style is mainly based on women's emphasis on society. Based on these findings, the authors assume that the frequent use of pronouns except for the word "I" is closely related to women's speech style. In the author-constructed corpus, there is no significant gender-based difference in the use of general pronouns, but as for the use of "who", 3rd person plural personal pronoun, 1st person singular subjective personal pronoun, and 1st person plural subjective personal pronoun, there are significant differences between female and male politicians.
Through the authors' research, there is the significant gender-based difference in the use of "I" and "we". It is worth mentioning that, compared with "I", candidates of different genders have more significant differences in the use of "we". In public political speeches, because the speaker is the direct participant in the speech and always mentions his/her own standpoint and viewpoint, the form of the statement is always the first person. In this case, the speaker is directly involved in the dissemination of information, enhancing the credibility of the speech. To a large extent, first-person pronouns can indicate the speaker's position, that is, how the speaker views his relationship with the listener. Normally, the use of "I" could convey more authority and force while the use of possessive "we" is contributing to a closer relationship with the listener than "I". In this study, female candidates use more first singular subjective personal pronouns, while male candidates use more possessive "we". This totally overturn the above findings; thus, it is evidential that both two genders have consciousness on breaking through the traditional bias in their biological gender roles.
Besides, men candidates use more the 3rd person plural personal pronoun than women. This result overturns Božić Lenard's (2016) research results once again who claimed that there is no significant difference in the use of "they" between male and female politicians. The 3rd personal subject personal pronoun can enhance the objectivity and fairness of the text at the lexical level and mark the speaker's participation in social interaction and orientation towards others (Božić Lenard, 2016). Additionally, in a negative context, the male politicians use third person plural personal pronoun as a resource to identify groups to which they do not belong, emphasizing the collective identity when criticizing each other's political parties and foreign political actions, which can significantly crack down on political opponents or foreign political groups and legalize them. This result supplies the evidence on male candidate's tendency to speak in feminize style.

Conclusions
It might be criticized that the discussion of gender difference in the use of language is premised on public's gender discrimination and the simple conceptualization of dichotomy. However, whether there is a difference is not decided by the science, on the contrary, the difference itself is actually produced by the public stereotypes which is non-negotiable. Based on the dichotomization of gender and from the perspective of social gender construction theory, the current dissertation provides a clear map of language differences between male and female candidates, and is conducive to present the changes of each gender group in a male -dominated political context. The authors construct an example diagram to show the correlation (see Figure. 1). In view of the fact that the public's understanding of others depends on their language to a large extent, the dichotomy of gender differences in previous studies and popular publications is likely to play a central role in maintaining gender stereotypes. Therefore, it can be said that the mutual interaction between public stereotypes and previous research concerning gender differences has contributed to the perpetuation of the dichotomy of gender color in the language use in public cognition. According to the theory of social gender construction, gender is not static, but dynamically constructed in social practice. Therefore, the dichotomy of gender color shown in the use of language existed in the public's cognition provides a basis for men and women candidates in political campaigns to use language strategically in order to shape a favorable gender image. In this way, different gender identities constructed by language enable challenges on the public's perception of speakers, thus it is more likely to make up for some of the negative perceptions of the biological gender, which is conducive to win the election. In conclusion, a statistically significant difference between the male and female speakers is recorded for 36 lexical variables out of the 137 tags defined in CLAWS. Out of the 36 tags, 23 were predominately used by the male politicians. Among the 23, there are 2 categories involving obvious masculine language style, while the 19 items contain strong feminine color which is defined by the past empirical studies. And out of the 36 variables, 13 were predominately used by the female politicians. Out of 13 variables, 10 categories are observed containing strong feminine color while the remaining 3 items are considered as a representation of masculine language style.
To sum up, the above research findings support that the male candidates tend to be more feminine in the use of words obviously, while the female group cannot show an obvious tendency to speak in the way of male but is more cautious in the choice of words in general. Thus, it can be speculated that female candidates control their conscious choices to a greater extent than male in the use of language in the political context. This echoes the female dilemma influenced by "double binds" (Jones, 2017, p.19). On the one hand, the female candidates need to avoid the doubts of voters about their leadership caused by too feminization; on the other hand, the group cannot fail to live up to the voters' expectation of women's high characteristics including high politeness, and sympathy, so that they still need to reserve a part of the feminine style of language. A neutral style of expression enables the diminishment of the voters' potential stereotypes on genders and contributes to a greater possibility to win support from public. Compared with female, men mainly pay attention to the negative considerations of voters on their own biological gender in the election, thus applying a part of feminized lexical categories become the first choice in their speeches.