I am doing open coding of a huge database of images. Unfortunately -or fortunately- this has to be done manually as the auto tagging software(s) are not yet reliable on this matter.
My question is that I am ending with a corps of around 90,000 words so what is the best approach or technique to associate each word to one or more of the pre-defined 25 categories I already made?
Are there any time-saving solutions rather than going the manual one by one way?
Thanks.
Maybe use Google hit count for each word in combination with each category?
I am not sure but probably you can use Wordnet (http://wordnet.princeton.edu/) to find a link/score of similarity between each word and each predefined category. Then you can associate each word to the category with the highest score.
Hey Bilal,
not sure what software you have used until now or if you have already decided on one, but atlas.ti provides the possibility of creating "families" of codes or "supercodes," they differ slightly. Also, I don't know if 90,000 words can be grasped? However, if yes for both, atlas.ti also can help you analyzing any co-occurrences or create a word count matrix and more...
@Ian: In Joseph Turian's map of 2500 English words (machine learning), several country-names are clustered together geographically, as one would expect. But according to the 'machine', Irak and Vietnam have about the same meaning :-) !
Thank you all for the contribution.
@Ian, I am afraid i have to agree with @Lambert
SO it seems there is no escape from creating dictionaries for the 25 categories i have based on the corps of words/units i have now. Maybe this will be a helpful start for someone who comes latter. But as i can see, no other way.
So, I will be using a combination of helpful tools through this process, I will explore things by " cluster the individual terms" as @Ian suggested but then will go through "analyzing any co-occurrences or create a word count matrix" as advised by @Katrine.
I agree with the general consensus that you want to use software that allows you to create "dictionaries, " and MAXQDA has a very nice tool for that called MAX Dictio.
Also, you want to avoid calling this "open coding" if you have pre-exting code categories. Open coding is a term from Ground Theory, and it means that you develop your codes from a strictly inductive, ground-up immersion in the data.
WordStat will help you categorize quickly a word and all related words (synonyms, antonyms, hyponyms, hypernyms, words with same stem, etc.) into a specific category: http://provalisresearch.com/products/content-analysis-software/wordstat-whats-new/
- Badges
- Science topic