Adjective-Noun Placement Variation in French Twitter Corpora Joan Palmiter Bajorek University of California, Davis One Shields Ave, Davis, CA 95616 jpbajorek@ucdavis.edu French tweets through the examination of two Abstract French Twitter timelines of users Le Monde and Cyprien. It is hypothesized that French French written and spoken corpora reveal tweets will display higher levels of adjective that variation of adjective-noun placement placement restriction than Thuilier’s written occurs more often within written corpora and spoken corpora findings. This may be than in spoken ones (Lightfoot, 1979; attributed to Twitter’s character limit and the Prévost, 2009; Thuilier, 2013). I examine curt, spoken-like treatment of language on the adjective lemma variation of placement, social network and overall online usage of adjective categories, frequency, and language. syllable length as a means of assessing current Twitter trends of adjective 2 French Adjectives placement. In this preliminary study, a cross-section of 30 adjective lemmas In French, attributive adjectives “adjectifs within a corpus of approximately 6,000 épithètes" can be placed before or after the noun as French tweets demonstrates that parallels seen in (1) (Grevisse & Goosse, 2011; Laenzlinger, exist between the usage of adjectives on 2005), an example modified from (Thuilier, 2013). Twitter and norms of spoken French. (1) a. un beau requin b. un requin sympa 1 Introduction a beautiful shark a shark nice This article examines French adjective-noun In example (1a), the adjective “beau,” found in the placement in the determiner phrase within prenominal position and modifies the noun written, spoken, and Twitter corpora. “requin.” Conversely, in (1b), the adjective Placement of French adjectives in the “sympa” is postnominal (Alexiadou et al., 2007; determiner phrase has long been controversial Delbecque, 1990; Thuilier, 2013). French and misconstrued within the literature adjectives are widely believed to exist in one of (Benzitoun, 2014; Grevisse & Goosse, 2011). three categories: fixed prenominal, fixed It is generally accepted that “[i]n Romance postnominal, or accepted alternator, with languages such as French and Spanish, recognized semantic shift attributed to either postnominal adjectives are the rule rather than position. the exception” (Alexiadou, Haegeman, & Although these categories are frequently used, Stavrou, 2007). “[i]t has proven notoriously difficult to define the However, contemporary written and spoken functions of the two positions” (Sleeman, Van de corpora indicate that adjective placement has Velde, & Perridon, 2014) and variation is oft greater flexibility and ambiguity than has attributed to semantic shifts (Alexiadou et al., previously been understood. Statistical analysis 2007; Laenzlinger, 2005). For example, the of corpora indicate that adjective placement is adjective “brave” means good or decent in the more restrictive in spoken usage than written preposition, but courageous or brave in the usage by a large margin (Thuilier, 2013). This postposition. In this way, positions are correlated research investigates adjective placement in with meaning. Yet, Thuilier, a native speaker of 7 French, provides several examples of exceptions of postnominal dominance in Modern French words that are considered as regular alternators (Boucher, 2006; Sleeman et al., 2014). with specified meanings correlated to positions Several theories postulate reasons for this (2013). Furthermore, many adjectives retain the phenomenon. According to Glatigny, same meaning regardless of placement, such as the “preposed adjectives belong in very great adjective “charmant,” meaning charming (Thuilier, majority to the ancient foundations of the 2013). In many ways, this variation in placement language” (translated by the author (Thuilier, thus is a choice of the user and varies widely on 2013)). The shift may resulted from the the usage context. It should be noted that the influence and exposure of French and primary concern of this research is variation in Romance languages to Germanic languages placement and patterns of adjective corpora rather (Grevisse & Goosse, 2011). Yet some reject than semantic interpretations, which are outside the these “influenced by another language” scope of this study. interpretations due to the lack of documented evidence (Lightfoot, 1979). Another theory 2.1 Syntactic Theory indicates that prenominal adjectives have marked positions that licenses specialized In considering the syntactic basis of adjective reading and greater frequency (Ledgeway, placement, modern frameworks explain the 2012). Evidence for this theory comes from the variation through Branching Direction Theory stronger syntactic bond of prenominal (Song, 2012). It is posited that nouns move adjectives to nouns as compared to the leftward and up the syntactic tree with “pied- relationship of their postnominal counterparts piping” and “snowballing” movement and interact and nouns. The phonological connections with linear and mirror image matching among between prenominal adjectives and nouns are languages (Cinque, 1994, 2010; Laenzlinger, 2005; demonstrated through the use of liaisons and Stavrou, 2012). However, much is left to be irregular adjective agreements (Ledgeway, desired in these theoretical understandings of 2012). syntactic underpinnings. If adjectives that alternate in position and have semantic shifts relegated to 3 Usage Norms in Corpora those positions, why does the noun moving up and down the tree account for the adjective’s semantic For decades, newspapers, literary texts, university change (Alexiadou et al., 2007)? How does this essays, and spoken corpora have been analyzed for theory account for languages that display varied their adjective placement norms. While patterns adjective-noun ordering systems? In the words of arise, gaps in information and uniformity of Alexiadou et al., “ [a] question that remains open is analysis abound in the research. Most studies cite why various languages have to resort to different raw counts of adjectives found to be prenominal or means (order, excessive articles, morphology) in postnominal. They do not take into account order to encode different interpretations of adjective lemma repetition, aspects of frequency, adjective-noun combinations” (parentheses and the possibility of alternation of position. original (2007)). Do to the recent growth in Recognizing the importance of the source of Optimality Theory (Song, 2012), there is hope that corpora, Delbecque writes that prenominal and elegant, more fully explanatory theories can be postnominal proportions of placement are “mainly developed. to be attributed to the text genre” (1990). Echoing these sentiments, Thuilier writes that intense 2.2 Historical Background difference exist according to the type of production (2013). French evolved from Latin, a language with no fixed adjective placement (Sleeman et al., 3.1 Literary Texts 2014). Between the 13th and 19th Century, adjective-noun placement shifted from In Wydler’s study of the famous text “La Chanson prenominal preference in Old French to de Roland,” 70% of overall adjectives were prenominal (Gerard J. Brault, 1978; Lightfoot, 8 1979). “The Song of Roland” is the written version 80.8% were exclusively postnominal (1980). of the epic tale that was once performed orally, Wilmet did not account for categories of though its precise origins are clouded in adjectives, but what can be interpreted from this “considerable speculation” (G.J. Brault, 2010). study is the propensity for the work of university Delbecque notes that modern French novels students, who focus their studies on literary demonstrate the same proportional split in criticism, history, and linguistics, to use literary, placement, but provides no data to support the worked, polished language that displays claim (1990). postnominal preferences, an idea supported by In another analysis of French adjectives, Wagner Thuilier’s work (2013). separated adjectives into three basic categories. “Cardinals” are basic, essential, and high 3.3 Newspapers frequency adjectives throughout the French language. “Populaires” are mainstream, Postnominal preference is also found in most of the colloquial, and wide spoken adjectives. Finally, studies in newspaper corpora. Five studies of “savants” are “learned” adjectives of scholarly French adjective placement in newspapers registers (translations by the author (Lightfoot, published between 1911 and 1978 by Damourette 1979)). For clarity in the following section, these and Pichon, Hug, and Forsgren returned three categories will be referred to as Frequent, homogenous results of 65% postnominal Colloquial, and Scholarly. In an examination of placement. With an average of 3,000 adjectives per 13th and 14th century prose, Wagner found that study, all of this research compared raw counts of Frequent adjectives strongly favored prenominal adjective-noun placement (Delbecque, 1990). positions: 2,393 prenominal and 11 postnominal occurrences, a 99.5% to 0.5% split (Lightfoot, 3.4 Spoken and Modern Corpora 1979). 75% of Colloquial adjectives were prenominal whereas roughly 70% of Scholarly Contemporary analyses provide more thorough and adjectives were postnominal. nuanced approaches to corpora studies as seen When the Frequent and Colloquial categories through the work of Thuilier (2013). For reduced were combined, the average prenominal to redundancies, section 2.8, refers to Thuilier’s 2013 postnominal division was 95% to 5%. This work. In 2013, Thuilier compiled a corpus of over combined category was of extremely high 1.3 million French words from written and spoken frequency, almost 16 times as common as the French. The written corpus was composed of Le Scholarly category (Lightfoot, 1979). It is Monde newspaper articles dating from 1989-1993 important to note that all of the data literature cited sourced from the French Treebank (FTB). For the in this section explores the raw counts of the data spoken corpus, Thuilier used the 2005 edition of and categorization of lemmas, but does not specify the CORAL Romance corpus of transcribed the frequency or other attributes of the adjectives. speech. Thuilier’s data reveals placement norms Wagner’s study underlines the importance of between the French written and spoken usages. adjective categorization when considering Spoken data displayed placement that was more adjective placement. In addition, Lightfoot notes rigid and more likely to follow prescriptivist norms that the difference between source adjectives and than written language. In explanation, Thuilier those that are “fairly recent” borrowings hypothesizes that written language is “travaillée,” reinforcing Glatigny’s theories of ancient French worked and polished, and therefore may have versus modern linguistic acquisitions (1979). greater opportunity for flexibility, nuance, and variations of style (translation by the author). 3.2 University Essays Conversely, it was hypothesized that spoken language is spontaneous and instinctive tendency In 1980, Wilmet investigated 90 philology essays follows mainstream norms. of university students dating from 1977 to 1978 Thuilier’s corpus comprised of 1,750 adjective (Delbecque, 1990). Of the 3,835 adjectives found, lemmas of which 170 were found to alternate in Wilmet reported that 2.3% were exclusively position. Oral corpora lemmas demonstrated a prenominal, 16.9% alternated in position, and prenominal preference of 74% while their written 9 corpora counterparts demonstrated a prenominal are retained on a user’s profile page, called a preference of 67%. “timeline” (Lomicka & Lord, 2012). Additionally, Thuilier coined a term about the A global phenomenon, Twitter has 288 million syllable length patterning demonstrated by the users as of 2015 (Popper, 2014; Statistica, 2015). data, “court avant long,” short before long While the company is American and based in San (translated by the author). Prenominal adjectives Francisco, 77% of Twitter accounts are outside of were frequently monosyllabic and “adjectives of the United States and over 35 Twitter company one syllable are more than 80% in anteposition” offices are outside of the United States, including (translated by the author). one in Paris (Twitter, 2015a). Despite its American origins, which might imply high volumes of English language tweets, only 34% of tweets are in English (Smith, 2015). Twitter users from France tripled from 2009 to 2013 (Statistica, 2013). 4 Defining the Corpus Approximately 6,000 tweets were analyzed from the Twitter feeds of the French newspaper Le Monde (@lemondefr) and the French comedian Cyprien (@MonsieurDream). These users were chosen because of 1) their comparable four million Twitter followers ("Twitter statistics for France," 2015), 2) Le Monde’s connection to Thuilier’s (2013) and Wilmet’s (1980) corpora studies, and 3) the users have the 2nd and 3rd largest audiences of Twitter in France. The top 24 French Twitter accounts are attributed to men or groups (Socialbakers, 2015) thus the Twitter timelines Table 1: Summary of Literature Review Percentages analyzed in this study are a sample of the most popular, mainstream content on French Twitter. From these studies and especially that of Thuilier’s work, further experiments can be 4.1 Le Monde designed to investigate several factors in adjective- noun placement in French corpora. To summarize Launched in March 2009, Le Monde’s Twitter the above sections succinctly the percentages timeline feed of tweets reports events related to quoted in this section are summarized in Table 1. France, the world, and politics. Written by The following factors have been previously journalists Pauline Croquet and Clément Martel, demonstrated to be significant in classifying the timeline has over 132,000 tweets and over French adjective placement: lemmas as a better 6,000 photos and videos (Twitter, 2015c). As the gauge of corpora rather than raw counts of online presence of one of the largest French adjectives, variation of placement, categorization newspapers, the tweets from this source are of adjectives, frequency, origin of corpora, and objective, political, newsworthy, and diplomatic in syllable length. nature. 3.5 Twitter 4.2 Cyprien Twitter is a real-time social networking and microblogging service (Lomicka & Lord, 2012). While Cyprien has significantly fewer tweets than Users can read, post, and repost messages called the Le Monde timeline, Cyprien’s popularity was “tweets.” Tweets are limited to 140 characters that determined to be a more important factor than number of tweets. Launched in June 2007, 10 MonsieurDream has over 10,000 tweets and is access to the general public, roughly 3,200 tweets written by Cyprien Iov who goes by Cyprien were downloadable per Twitter timeline. Each user (Twitter, 2015d). Cyprien is a prolific 24-year-old has a homepage, a unique page of their tweets and French blogger, illustrator, comedian, and poster of retweets called a “timeline,” synonymous with the videos on YouTube.com (Iov, 2015a, 2015b). His term “Twitter feed” (Twitter, 2015b). The scripts work ranges from the banal to the political, created csv files of the tweet’s date published and blogging about the experience of moving text content. Due to the prolific tweeting of Le apartments, gay marriage, and general self- Monde, all tweets analyzed were published promotion. Cyprien has more Twitter followers between April 1st, 2015 and May 4th, 2015. For the than the Twitter accounts of movie celebrities, for Cyprien corpus, the tweets from his timeline were example Gad Elmaleh with 300,000 fewer published between December 9th, 2009 and May followers (Socialbakers, 2015). As a young 2nd, 2015. Hereafter for reasons of clarity, the comedian, Cyprien’s humor is colloquial, corpora will be termed “Le Monde” and “Cyprien” sensationalized, dramatic, and irreverent in nature. and will no longer be italicized. His most popular YouTube video is a rap song For the scope of this study, in the Twitter corpus where Cyprien responds a man who criticized his 30 adjective lemmas were investigated, see jokes (Cyprien, 2011). As of May 2016, the video Appendix. These lemmas were chosen based on has almost 37 million views and includes jokes their prevalence in the linguistic literature and about sex, online reputation, clothing style, and were placed into categories: High Frequency, poor education, concluding with a joke about Colors, Contemporary/Neologisms, Famous killing handicapped people (Cyprien, 2011). Alternators, and Others, see Appendix A. By juxtaposing the linguistic trends from these Tweets containing the specific adjective lemmas two users, adjective placement norms on Twitter were appended to separate csv files to isolate all can be examined. It was hypothesized that French instances of the adjective. Due to the added tweets would display higher levels of adjective complexity of multiple adjectives, see mirror placement restriction than Thuilier’s corpora imaging (Cinque, 1994, 2010), only tweets with a findings (2013). It was expected that Cyprien’s single adjective were analyzed. Tallies were usage of language would demonstrate even greater created of how many instances were found for each restriction in adjective placement than Le Monde’s adjective lemma within the two Twitter feeds. tweets do to the higher register and scholastic These tallies were then sorted to categorize nature of the latter’s content. adjectives found in a determiner phrase. From those that fit the correct syntactic environment, 5 Methodology determiner phrase, they were annotated as prenominal and postnominal. Tallies were noted The research looked into the placement norms of for all the above values. In Python, a Chi square Le Monde and Cyprien on Twitter to determine test was conducted on the raw counts of the 30 whether corpora more closely resembled written adjectives found in the determiner phrase of either or spoken corpora norms. In addition, questions of timeline were statistically significant. distribution, alternation, syllable length and frequency were considered in the data collection 6 Findings and analysis. Tweets were downloaded from twitter using 6,406 tweets were data-mined from the Cyprien Python scripts and the Python module “Tweepy” and Le Monde Twitter feeds. After isolating the 30 (Roesslein, 2009). To access the tweets, a Twitter adjective lemmas, 1,181 adjective occurrences account was created with a corresponding Twitter were analyzed. Roughly 82% of these single website giving the user access to personalized adjectives were found in a determiner phrase or Application Program Interface (API) credentials. noun phrase, a total of 949 adjective occurrences. Modified from scripts posted on GitHub Some adjective lemmas were not attested within (Yanofsky, 2013), roughly 6,000 tweets were the corpora. In the analysis of the data, variation in downloaded from Twitter users Le Monde and placement, categorization of adjectives, adjective MonsieurDream. Due to restrictions on tweet frequency, syllable length and origin of the corpora 11 were analyzed. The following analysis considers the Le Monde and Cyprien datasets separately as 6.2 Adjective Categories well as combined, to consider the implications of Twitter as a usage context. In consideration of the initial categories of the 30 It is crucial to note that despite the 6,406 tweets adjective lemmas, adjectives were analyzed within data-mined, the resulting adjectives found in their categorizations as a comparison point to determiner phrases was lower than initially Wagner’s categories of Frequency, Colloquial, and anticipated. It is therefore recognized the findings Scholarly, see Table 2 (Lightfoot, 1979). Of the from this corpus must be considered a preliminary Famous Alternators, most of these were study that has recommendations for further prenominally placed, though many of this category investigations. were adjective lemmas that were unattested in the corpus. Contemporary/Neologisms and Others 6.1 Placement Variation displayed the greatest amount of variation in placement within the corpus. Mirroring Wagner’s Of adjectives found within the corpora, 8% of results, High Frequency adjectives were placed adjective lemmas in Cyprien tweets and 30% prenominally for the most part, on average 92% of adjective lemmas in Le Monde’s tweets varied in the adjective lemmas. Colors were exclusive found placement, Figure 1. in postnominal positions. This complements Of the lemmas which alternated in this corpus, Wilmet’s findings which demonstrated that “blanc, they demonstrated prenominal preference at an bleu, noir et rouge” were found postnominally average of 80%/20% prenominal/postnominal 97.4% of the time (1980; Thuilier, 2013). Thus, split. The distribution of this variation ranged from category of adjective was helpful for grouping 60% to 99.67% prenominal placement. These some adjectives, but not for all the categories of finding are in line with Thuilier’s finding that the study. alternators are more inclined to be prenominally placed (2013). Figure 1: Raw Counts and Percentages of Tweets Table 2: Combined Cyprien & Le Monde Data Sets Trends of the placement of alternators differed between the Cyprien and Le Monde corpora. Le 6.3 Frequency Monde’s adjectives displayed more variation in placement whereas very few of Cyprien’s Adjectives used with highest frequency are adjectives varied in placement. Supporting the correlated with those that are prenominally placed. initial hypothesis, adjective placement is more Of the top 10 adjective lemmas in the Cyprien rigid in Cyprien’s tweets than in Le Monde’s tweets, all were prenominal dominant at an average tweets. The margin of difference in usage norms is of 97.79% overall. Of the top 10 adjective lemmas a significant finding of this study and would be a in the Le Monde tweets, 70% were prenominal strong starting point, to analyze Twitter feeds of dominant, see Table 3. These high frequency news sources as compared to the tweets of adjectives in the Le Monde tweets were individuals, for future investigations. exclusively postnominal, “politique,” “blanc,” and “rouge.” As seen also with the above section, 12 adjectives describing colors are famously While two syllable adjectives display prenominal postnominally placed (Grevisse & Goosse, 2011). preference, this is not a major finding of this study. Of the highest frequency adjectives from both timelines, 6 out of 10 overlap between the Cyprien 6.5 Twitter Corpora Overall and Le Monde tweets. As seen in the variation of placement, Le Monde’s adjective lemmas demonstrated greater flexibility of placement than Cyprien’s, which supports the hypothesis. Le Monde’s flexibility was more in line with written corpora norms as seen in Thuilier, while Cyprien’s follows more spoken norms of restriction (2013). As compared to the corpora in Table 1, section 2.8, this Twitter corpora demonstrated significant stratification at a 76%/31% prenominal split. From a raw count of the 30 adjectives found in either Table 3: Combined Cyprien & Le Monde Data Sets position within the corpora, Cyprien’s adjective placement was more stratified than Le Monde’s In the Cyprien corpus, there was a significant placement, Table 4. When combining the counts in divide between the high frequency adjectives both Twitter feeds, there was a strong tendency prenominal dominance to low frequency toward prenominal placement: 504 prenominal to postnominal dominance. The Le Monde corpus is 100 postnominal, 5:1 a ratio. more nuanced in this separation of variables. In the Le Monde corpus, 7 of the 10 high frequency adjectives displayed prenominal dominance. For the low frequency adjectives, 10 of 13 also displayed prenominal preference, with only 3 that were exclusively postnominal in placement.Overall, the Le Monde corpus has a larger number of adjectives that alternate in Table 4: Raw Counts and Percentages of Tweets placement than those that are exclusively It should be noted that the lemma “nouveau” was postnominally placed, which leads to lower levels removed from the raw count comparisons due to its of stratification in frequency and placement. When nature as an outlier. While the lemma posed no the corpora were combined, the mixed nature of problem as an example of categories, high the Le Monde corpus was demonstrated. frequency, and syllable length, its high raw count Frequency was strongly correlated with skews the data set. Within the Cyprien data set, prenominal adjective placement in the Cyprien “nouveau” was found 301 instances in the corpus where it was highly stratified, but this trend determiner phrase compared to all the other was less significant in the Le Monde corpus. adjectives with values ranging from 0 to 59 occurrences. Due to the limited scope of this article 6.4 Syllable Length “nouveau” was not analyzed independently, a case for future study. The results of this study suggest that syllable count was not a significant factor in adjective placement. 7 Conclusions Previous literature indicated that “short before long” patterns were observed in larger data sets Twitter corpora displayed adjective placement (Thuilier, 2013), however these trends were not norms that were more closely aligned with spoken observed in these data, see Appendix B. The corpora norms than written corpora norms when average percentage for the average prenominal compared to the patterns explored in Thuilier placement for lemmas with one syllable was 60%, (2013). As hypothesized, the Cyprien corpus was two syllables at 82%, and three syllables at 61%. 13 more restricted in adjective placement alternation adjective-noun prenominal and postnominal than the Le Monde corpus. In the Le Monde alternation; Spanish does as well (Alexiadou et al., tweets, 30% of the adjectives were found in both 2007; Delbecque, 1990). An analysis of the prenominal and postnominal positions, much adjectives in Spanish tweets, or other languages higher than the 8% alternating in the Cyprien with comparable alternation trends could be corpus. This suggests the greater flexibility of compared to this study. This study provides a adjective placement in the Le Monde corpus. jumping-board for further adjective-noun Twitter Syllable count was not a significant factor in corpora studies. adjective placement, though there was a tendency for adjectives of two syllables to be prenominally Appendix A place. Frequency was a significant factor in Full Combined Data Sets for Prenominal and adjective placement and was highly correlated with Postnominal Placement of Adjectives found in prenominal placement in the Cyprien corpus. Determiner Phrases Limitations of this study include small final numbers of adjectives attested per lemma. Despite the approximately 6,000-tweet sample size, some adjectives occurred less than 5 times. In the future, larger data sets would be preferred. In addition, the 30 adjectives chosen may have influenced the data collected and in the future, different sampling methods could remedy this issue. Overall, these findings provide preliminary insight into the French adjective placement norms on Twitter. 7.1 Further Research Sampling: Using a parts of speech tagger, a larger analysis of all the adjectives in the corpora could be conducted. Tweets could be data mined over a specific time interval rather than from user timelines. In addition, this study showed the longitudinal trends of specific users. Further research could also examine cross-sectional data of streams of tweets and isolate specific collocations or adjective usage. Collocations of adjective-noun pairings and a study of ngrams of the corpus could illuminate stylistic preferences of Twitter users. Gender and other demographics data could be compiled to generate studies that compare different types of Twitter users. Nouveau: Further study might investigate specific adjective lemmas, such as nouveau, for a microanalysis interpretation of the data. Dialects and Languages: While it is clear that French is spoken around the world, varieties of the language based on location were not chosen for the this study (Thuilier, 2013), but could be analyzed to compare language norms of this corpus. Semantic shifts in placement were not analyzed within this study, but could be an aspect of future work. French is not the only language that displays 14 Appendix B Chomsky, N. (1986). Knowledge of Language: Its Nature, Origin, and Use: Praeger. Full Combined Data Sets of Syllable Counts Cinque, G. (1994). On the evidence for partial N- movement in the Romance DP. Paths towards universal grammar, 85-110. Cinque, G. (2010). The Syntax of Adjectives: A Comparative Study: MIT Press. Cyprien. (2011, May 21, 2016). Cyprien répond à Cortex. Retrieved from https://www.youtube.com/watch?v=dKwzZZKIbUs Delbecque, N. (1990). Word Order as a Reflection of Alternate Conceptual Construals in French and Spanish. Similarities and Divergences in Adjective Position. Cognitive Linguistics, 1(4), 349-416. Grevisse, M., & Goosse, A. (2011). Le bon usage : grammaire française : 75 ans. Bruxelles; [Paris]: De Boeck-Duculot. Iov, C. (2015a). Presse. Retrieved 30 May 2015, from http://www.cyprien.fr/index.php/presse/ Iov, C. (2015b). YouTube Channel: Cyprien.fr. Retrieved 30 May 2015, from https://www.youtube.com/user/MonsieurDream Laenzlinger, C. (2005). French adjective ordering: perspectives on DP-internal movement types. Lingua, 115(5), 645-689. doi: 10.1016/j.lingua.2003.11.003 Ledgeway, A. (2012). From Latin to romance : morphosyntactic typology and change. Oxford; New York, NY: Oxford University Press. Lightfoot, D. (1979). Principles of Diachronic Syntax: Cambridge University Press. Lomicka, L., & Lord, G. (2012). A tale of tweets: Acknowledgments Analyzing microblogging among language learners. Thank you to the reviewers for their insightful System, 40(1), 48-63. comments and to Dr. Raúl Aranovich, Dr. Susan Matthews, P. H. (2007). Syntactic relations: A critical Palmiter, and Alan Wong for their remarks on earlier survey (Vol. 114): Cambridge University Press. versions of this work. Popper, B. (2014). Twitter Now Has 255 Million Users, but Activity Has Dropped Year over Year. References Retrieved 20 Mar. 2015, from http://www.theverge.com/2014/4/29/5665752/twitter Alexiadou, A., Haegeman, L., & Stavrou, M. (2007). -q1-2014-earnings Noun Phrase in the Generative Perspective. Noun Prévost, P. (2009). The Acquisition of French: The Phrase in the Generative Perspective, 71, 1-664. doi: Development of Inflectional Morphology and Syntax 10.1515/9783110207491 in L1 Acquisition, Bilingualism, and L2 Acquisition: Benzitoun, C. (2014). The place of the attributive John Benjamins Publishing Company. adjective in French: what we learn speech corpora. 4e Roesslein, J. (2009). Tweepy Documentation. from Congres Mondial De Linguistique Francaise, 8, 16. http://docs.tweepy.org/en/v3.2.0/ doi: 10.1051/shsconf/20140801066 Sinclair, J., & Carter, R. (2004). Trust the text : Boucher, P. (2006). Mapping Function to Form: language, corpus and discourse. London; New York, Adjective Positions in French. Lingvisticae N.Y.: Routledge. Investigationes, 29(1), 43-60. Sleeman, P., Van de Velde, F., & Perridon, H. (2014). Brault, G. J. (1978). The Song of Roland. University Adjectives in Germanic and Romance: John Park: Pennsylvania State University Press. Benjamins Publishing Company. Brault, G. J. (2010). Song of Roland: An Analytical Smith, C. (2015, 20 May. 2015). By the Numbers: 150 Edition: Introduction and Commentary: Pennsylvania Amazing Twitter Statistics (May 2015). 30 May State University Press. 2015, from 15 http://expandedramblings.com/index.php/march- Twitter. (2015a). About Twitter, Inc. Retrieved 30 2013-by-the-numbers-a-few-amazing-twitter- May 2015, from https://about.twitter.com/company stats/10/ Twitter. (2015b). Help Center: What's a Twitter Socialbakers. (2015). Most Popular Twitter Accounts in timeline? Retrieved 04 Jun 2015, from France. Twitter Statistics for France. Retrieved 30 https://support.twitter.com/articles/164083-what-s-a- May 2015, from twitter-timeline http://www.socialbakers.com/statistics/twitter/profile Twitter. (2015c). Le Monde.fr Twitter. Retrieved 30 s/france/ May 2015, from https://twitter.com/lemondefr Song, J. J. (2012). Word order: Cambridge University Twitter. (2015d). Monsieur Dream Twitter. Retrieved Press. 30 May 2015, from Statistica. (2013). Facebook, Twitter and Google https://twitter.com/MonsieurDream Penetration 2009-2013 | France. Retrieved 30 May Twitter statistics for France. (2015). Retrieved 2015, from 20.March.2015, from s/france/ Statistica. (2015). Social Networks: Global Sites Wilmet, M. (1980). Anteposition and Postposition of Ranked by Users 2015. Leading Social Networks Qualificative Epithets in Contemporary French. Worldwide as of March 2015, Ranked by Number of Travaux de Linguistique, 7, 179-201. Active Users (in Millions). Retrieved 30 May 2015, Yanofsky. (2013, 1 Nov. 2013). A Script to Download from All of a User's Tweets into a Csv. http://www.statista.com/statistics/272014/global- Yanofsky/tweet_dumper.py. Retrieved 30 May social-networks-ranked-by-number-of-users/ 2015, from https://gist.github.com/yanofsky/5436496 Stavrou, M. (2012). The syntax of adjectives: A comparative study. Language, 88(2), 419-423. doi: 10.1353/lan.2012.0024 Thuilier, J. (2013). Syntaxe du français parlé vs. écrit : le cas de la position de l’adjectif épithète par rapport au nom. TIPA. 16