Introduction
There is a growing interest on studying subjects’ personality, specially among the natural language processing (NLP) community. This is because through tech- niques developed by psychologists, identification of one’s personality has been proved efficient for predicting thought patterns, emotions and behaviour [1].
In order to study this area, the NLP community needs to have resources, i.e. labelled corpora. While there is a large number of resources in English about a great number of problems, very few resources exists for Spanish, and even less for the Personality Identification task in Spanish.
To tackle this problem we have been collecting, during 2 years, a corpus of handwritten short essays of undergraduates Mexican students. The personal- ity information of each subject was obtained using a psychological instrument called TIPI (Ten Item Personality Inventory) [2]. Our corpus, called HWxPI (Handwritten text for Personality Identification), contains information from 836 subjects. Recently was used in the Multimedia Information Processing for Per- sonality & Social Networks Analysis Challenge at ICPR (International Confer- ence on Pattern Recognition)1.
HWxPI corpus
The corpus consists of handwritten Spanish essays from undergraduate Mexican students.2 For each handwritten essay we have two sources of information: the manual transcription and the scanned image of the handwritten essay. The cor- pus is available at https://competitions.codalab.org/competitions/18362. An example of these two modalities can be seen in Table 1.
Table 1. Example of a scanned image of a handwritten essay’ fragment and its manual transcription with added tags.
Una vez sali <FO:salí> con un amigo no muy cercano, fuimos a comer y en la comida el chico
se comportaba de forma extraña algo como <DL> desagradable <DL> <DL> con un <MD>
aire de superioridad <MD> algo muy desagradable tanto para <DL> mi <FO:mí> ...
Ground truth. During the gathering process we asked each subject to an- swer a psychological instrument called TIPI to identify its personality according to the Big Five Model (i.e., Extroversion, Emotional stability, Agreeableness, Conscientiousness, and Openness to experience traits). The TIPI allows to di- vide each trait into four classes: high, medium high, medium low, and low. For HWxPI corpus we binarized the personality information of each trait, such as, high and medium-high classes are converted into 1 and low and medium-low are converted into 0.
Manual transcriptions and annotations. An important aspect of this corpus, beside its manual transcription, is a set of seven tags used to labelled handwriting phenomena: insertion of drawings or emojis <D:desc.>, insertions of a letter into a word <IN>, modification of a word <MD>, elimination of a word <DL>, two words written together <NS>, syllabification <SB> and misspelling <FO:word>. To the best of our knowledge there is no other corpus for personality identification with this kind of information. Preliminary analysis suggests that some tags might be positively correlated with a personality trait.
We keep working on gathering more subjects to participate on this research project. Therefore, eventually we can add more instances to our corpus.
References
- Funder, D.C.: Personality. Annual Review of Psychology 52(1), 197–221 (2001). https://doi.org/10.1146/annurev.psych.52.1.197
- Gosling, S.D., Rentfrow, P.J., Swann, W.B.: A very brief measure of the big-five personality domains. Journal of Research in Personality 37(6), 504 – 528 (2003). https://doi.org/10.1016/S0092-6566(03)00046-1
- Ramírez-de-la-Rosa, G., Villatoro-Tello, E., Jiménez-Salazar, H.: TxPI-u: A resource for personality identification of undergraduates. Journal of Intelligent & Fuzzy Sys- tems 34(5), 2991–3001 (2018). https://doi.org/10.3233/JIFS-169484
1 http://chalearnlap.cvc.uab.es/challenge/27/description/
2 A subset of this corpus and the complete gathering methodology is described in [3].