# Context ## Survey data (2014, 2016, 2017 and 2018) The aim of this dataset is to provide access to the raw survey data from the 2016, 2017 and 2018 OSMI mental health in technology surveys used to facilitate analysis e.g [my kernel fusing the OSMI surveys across time periods](https://www.kaggle.com/ekwiecinska96/dataset-creation-fusing-surveys-from-2014-2018). This is due to the fact that the popular 2014 dataset uploaded onto Kaggle has already been pre-processed and cleaned (and the only other 2016 upload does not play nice with kernels). Whilst this is useful, many columns were renamed into simple attributes e.g 'Are you self-employed?' is standardised to 'self_employed'. As none of the surveys from the following years have had this treatment, it was difficult to reverse-engineer the processing steps to make the attributes match. Also, it's great to have all the data in one place. ## Similarity matrix The associated similarity matrix, stored as a numpy-readable file (.npy) is a supplementary file for the previously mentioned kernel. This was uploaded due to the unfortunate fact that any [SpaCy models ](https://spacy.io/usage/models) that are contain word vectors (aka any model other than *sm*) are not supported by Kaggle on the date of writing (Jun 2019). Please see the associated kernel for more information on how this matrix was created. ## Acknowledgements The original data collection and hosting has all been provided by [Open-Sourcing Mental Illness (OSMI).](https://osmihelp.org) you can find all of the datasets (including 2016, 2017 and 2018) [here](https://osmihelp.org/research). # Inspiration The inspiration for uploading these datasets was to allow Kaggle users such as myself to have greater control over the pre-processing and standardisation of attributes.