My Mooc log

The work on the MOOC implies to chose a research topic that will be studied using one of the datasets provided.

I chose to study the association between current romantic relationships and ideal relationships among adolescents.

Here are the links for my literature review on this topic (or close ones):

- Adolescent romantic relationships: a look from the future http://www.sciencedirect.com/science/article/pii/S0140197101904099

- The Emerging Field of Adolescent Romantic Relationships http://www.psy.miami.edu/faculty/dmessinger/c_c/rsrcs/rdgs/peers_social_general/furman.adol_romance.curddir2002.pdf

- Adolescent Romantic Relationships – Why are they important? And should they be encouraged or avoided? http://headspace.org.au/assets/Uploads/Resource-library/Health-professionals/romanticrelationships-adolescent-romantic-relationships-why-are-they-important-headspace-evsum.pdf

- Adolescent Romantic Relationships and Experiences http://www.du.edu/psychology/relationshipcenter/publications/furman_collins_2007.pdf

- The role of romantic relationships in adolescent development  http://www.du.edu/psychology/relationshipcenter/publications/furman_shaffer_2003.pdf

- Romantic Relationship Patterns in Young Adulthood and Their Developmental Antecedents http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3830676/

- Relations amoureuses à l’adolescence : une étude sur des jeunes appartenant aux couches populaires cariocas https://www.cairn.info/revue-dialogue-2008-1-page-103.htm

- Les relations amoureuses des lycéens http://www.persee.fr/doc/socco_1150-1944_1995_num_21_1_1417

- L’attachement à l’adolescence http://www.cairn.info/revue-dialogue-2007-1-page-73.htm

As a research project, I decided to use the AddHealth dataset. I want to study romantic relationships of adolescents as they are described in this dataset. I plan to work on the relation between the ideal romantic relationships of respondants and their actual relationships: does the fact that a youngster is involved in a romantic relationship change his/her vision of an ideal relationship? does the form of the actual relationship change the ideal vision (for example, is an adolescent currently involved in a romantic relationship more likely to dream of marriage?)?  

For my literature review, I used the following terms:

- romantic relationship adolescent current ideal

- adolescent romantic relationships

- romantic relationship adolescent current marriage

- relations amoureuses adolescents (French)

Literature (see references here: http://mooclog.tumblr.com/post/135068953486/mooc-data-management-and-visualization-literature) focuses on the importance of the adolescence romantic relationships in adult life. It also notices that young adults do not perceive their past adolescent relationships in the same way as they do their current romantic relationships.

As a hypothesis, I want to confirm (or not) this assumption: adolescents involved in a current romantic relationship are more likely to dream of a more engaging relationship (marriage, children...) and even more as their current relationship is older.

To create the query for a research:

Create a pilot corpus with a simple query and find the keywords of this pilot corpus against a general corpus;

Intuitively decide the accuracy of the KW

Video 4

Video 1

Sets of +/- 5 words is the most useful span for collocates

Minimum frequency threshold: 10 co-occurrences

Take into accounts sentence boundaries?

Mutual information: how often 2 words co-occurate, relative to how often they occur without one another.

Mutual information may not be very accurate with data low in terms of frequency

Other measures are available that can give different results: for example the Dice coefficient

Video 2

Colligation: a word collocates with a particular grammatical class

Semantic preference: collocation with words in a specific semantic family

Discourse prosody characterize speaker's attitude. Example of the verb 'cause' which is very often associated with negative events.

Video 3

Keyword list < comparing 2 frequency lists

Defining the words diamond and cause.

Diamond: English noun. Refers to a very hard gem, mostly used for jewellery, and also in tools to cut glass. Also used for a typographic sign.

A token is any given word in the corpus.

A type is the number of unique word forms present in a corpus.

As an exercice, I tried to reply to a question I made in one of my previous posts: how many words are there in Pride and Prejudice by Jane Austen. So I loaded a raw text (unicode) version of the novel on Project Gutenberg, edited the file to delete the Gutenberg license and information and load it into AntConc. So, the answer is: 6259 words types and 122817 word tokens.

A question occurs to me then: is this distribution (about 20 times more tokens than types) standard in English? Let's see with the LOB corpus: 39524 types, 1020445 tokens, so about 26 times more tokens than types.

I notice that the LOB corpus consists of 15 files of various sizes with names ending with a letter: LOB_A.txt, LOB_B.txt etc. What do these different files correspond to?  How and why is the corpus split in different files?

[a few minutes later] Silly of me! Each file correspond to the genres covered by the corpora, as explained in the section of the course about them.

Resources

Laurence Anthony’s YouTube site

Types of corpora

Specialised corpus (by genre, time, place)

General corpus

Multilingual corpus (same kind of texts in various languages)

Parallel corpus (exactly the same texts translated to several languages)

Learner corpus

Historical or Diachronic corpus

Monitor corpus (extended daily or almost)

Definitions

Concordance: words with left and right contexts

Collocation: tendency of words to co-occur

Questions

In one of the videos, an example given is a corpus of Jane Austen's works. Thinking about it, I realize that I can't figure how words there are in a novel, for example. So how many words are there in all Jane Austen's works? In Pride and Prejudice? [XXX to search later]

Is a corpus of several translations of the same text in only one language (for example, all the English translations of Sophocles' plays from the XVIIth century to now) a parallel corpus  or a diachronic corpus?

I tried to figure out how I could use corpora for some of my interests: history, literature, folklore... One special topic I am interested in is western esotericism.

So how could I use corpus linguistics to explore such a topic? I think a lot of work can be done exploring diachronic esoteric corpora. For example, alchemical latin literature is mostly 'technical' and 'scientific' at its beginning, but about the XIIIth century, it begins to be more 'spiritual' or 'mystical'. I've been willing to study this point for a long time, to assess how this tendancy emerged. I think corpus linguistics could give interesting hints about the evolution of the vocabulary used and the semantics of the words.

Another interesting work could be the study of a parallel corpus, for example of translations of a common text (for example the Corpus hermeticum) to compare reception of this text in various language areas and in various times.

I am a former teacher in what we call in France "Lettres classiques", i.e. French, Latin and Greek literatures and languages, in secondary schools. I later changed my mind and have been working for more than 10 years as a ICT for education specialist (doing consulting, programming and system administration). I follow this course to get some general culture about corpus linguistics and, maybe, to get ideas to combine some of my main interests: computer science, languages, history (especially history of wester esotericism).