facebook graph api - C# - Personality Analysis based on on Word/Language -

i trying develop c# program retrieves user's facebook posts , analyse person's personality based on posts. mean analysing words , languages/sentence structures determine type of person is. eg: overconfident, timid, angry etc.

however, have no idea how should go using words analyse personality. seems difficult since can have both positive , negative words in post. ideas?

what intend accomplish quite complex (mainly delivering acceptable accuracy) , there no simple solution.

basic ideas of approach take in position:

1. it can exclusively used given language (e.g., english), 1 consider while performing development. relying on translations bring other languages account affect reliability lot. new "starting analysis" required each new language want consider.
2. the important part of development database , have put big focus on design, connectivity, data storage/retrieval, etc.
3. minimum content database: list of words, associated "value" (timid, angry, etc.), importance within given value (the f word getting 8 out of 10 "aggressive"), how word have different meaning (as suggested in comment above, "sick" might warning flag check exact context), further considerations (e.g., how number of repetitions affects value), etc. additionally, have include different levels of storage (per post, per user, per time, etc.).
4. you have create algorithm adaptable , scalable enough (lots of changes, improvements, additions, etc. expected here) deliver want. basic idea come is: assessing each post on account of values each word defined in database (by considering each word alone, number of repetitions of given word, context of given word, etc.), is, checking words analysis-worthy , ones not; parsing system not capable extract individual words analyse context (words before , after target 1 in same sentence/paragraph or in different one); setting rules avoid "misunderstandings" (e.g., minimum number of posts consider behaviour aggressive, otherwise ignore (perhaps humour); accounting complex moods formed on account of different types of posts (e.g., angry in 3 posts + timid in 9 posts = *); etc.); in summary, has capable of converting "discrete posts reality" desired output (an assessment post/user) accurately possible.

as said, not easy. if things step step , make sure structure adaptable enough allow modification/extension, might pretty reliable piece of software (by understanding ideal result can expect, is, ranking of type of language used; extrapolating real personality analysis sounds perhaps ambitious) might of interest quite few people.