The project aims to develop a prototype of a platform that eradicates the traditional way of employment which comes with the need to manually go through numerous applications and CVs to find out what suits the particular requirement of the job being offered. WeEmploy seeks a more efficient way to short-list submitted candidate CVs from a large number of applicant providing a consistent and fair CV ranking policy, which can be legally justifed. This system will help the HR department to easily short-list the candidate based on the CV ranking policy.
- Demo
- Prerequisites & Development Libraries
- Installation
- Instructions
- Demo
- Background
- Components of the Application
- Aptitude Assessment
- Personality Test
- CV Analysis
- Disclaimer
Our hosted app had some features removed due to limited storage by hosting provider. We eliminated the personality feature due to the size of the models trained. However, all other features are fully functional.
The video below shows a live demo of how the web application works.
The video below shows a live demo of how the web application works.
To install WeEmploy you'll need pip and Git. It also uses a some Python packages (NumPy, SciPy and Matplotlib) but these should be taken care of by the installation process.
| Software | Version |
|---|---|
| Python 3 | 3.9.6 |
| Pandas | 0.25.1 |
| Snscrape | 0.4.3 |
| Flask | 2.1.2 |
| Google Chrome | 102.0.5005.115 |
You can install WeEmploy by cloning the repository:
git clone https://github.com/philkam/AI_Personality-Prediction-System-Through-CV-Analysis.git-
Clone the repository
-
Set up a virtual environment. See here for more details. Don't forget to activate the virtual environment.
-
Install all required libraries through requirements.txt
pip install requirements.txt-
Run your local server ( WAMP, XAMPP etc)
-
Now run the Flask app app.py.
python app.py- In your browser open http://localhost:5000 (or :{port-number} as specified by the Flask's development server)
There are various tests that help to determine personality types such as the Big Five, Rorschach test, and MBTI test. In this project, prediction of personality is done by considering the MBTI test.
The MBTI personality classification system grew out of Jungian psychoanalytic psychology as a systematization of archetypal personality types used in clinical practice. The system is divided along four binary orthogonal personality dimensions, altogether comprising a total of 16 distinct persons.
-
Extraversion (E) vs Introversion (I): a measure of how much an individual prefers their outer or inner world.
-
Sensing (S) vs Intuition (N): a measure of how much an individual processes information through the five senses versus impressions through patterns.
-
Thinking (T) vs Feeling (F): a measure of preference for objective principles and facts versus weighing the emotional perspectives of others.
-
Judging (J) vs Perceiving (P): a measure of how much an individual prefers a planned and ordered life versus a flexible and spontaneous life
- Login and Registration
- Aptitude Assessment
- Personality test
- CV analysis
The aptitude assessment helps understand the underlying patterns of candidates interest and predict the stream that the candidate is interested in. Understanding a candidates inherent aptitude is very crucial for an organisation. Candidates can test their aptitude after which a report is generated which can assess a candidate's interest. Based on this, the Human Resource Manager can place a candidate in the right team and point out the right candidate for a particular job.
This section of the application allows the registered candidates to attend the aptitude assessment. The candidate registers through a portal, fills in the compulsory profile information and logs on to the application. The student views the assessment report. The application allows the human resource personnels to prepare the set of questions. Assigning of questions to the candidates is done automatically, though the question paper is prepared by a human resource personnel. It also allows the personnels to manage students and also to edit their own profile. The personnel is registered by the Human Resource Manager.
All the questions are MCQ. Four types of questions are supported by this application.
- Normal MCQ questions
- Question with image and option
- Question and option with images.
Questions can be from four main sections namely Science, Commerce, Humanities, and Aptitude. To prepare the question paper the human resource personnel chooses about fifteen questions from each section. Each question paper contains sixty questions. The questions are given a weightage according to the category that particular question falls. The weightage is made useful for the evaluation of candidate assessment.It is mandatory for the students to attempt all the sixty question.
- Registration
- Login
- Answers questions
- Views report
- Only the Human Resource Manager can add a personnel to the Staffing team.
- The Human Resource personnels determines the questions a candidate should take.
- The Human Resource personnels sets the weight each question should carry
- Student profile
- Answers given by students
- Instructor details
- Question details
- Answer details which comprises of possible answers and the right answer
Social media establishes uninterrupted connectivity, between its users and external world through revealing personal details and their viewpoints in every aspect of life. The focal aim of this here is to analyze how twitter (dataset) can be used to predict personality. We show processes of developing Machine Learning models on textual data. With this candidates can see their predicted personality types according to Myers-Briggs Personality Types Indicator using their twitter user names since twitter is the classic entry point for practicing machine learning. With Twitter data, you got an interesting blend of data (tweet contents) and meta-data (location,hashtags, users, re-tweets, etc.) that opened up paths for analysis. This training contains following topics:
- Exploratory Data Analysis
- Handling Imbalanced Dataset
- Vectorization of Text Data
- Model Creation
- Model Training
- Model Evaluation
The publicly available Myers–Briggs personality type dataset from Kaggle, containing 8675 rows of data, was used in this research. In this dataset, each row consists of two columns. The first column is for the MBTI personality type of a given person, and the second column includes fifty posts obtained from the individual’s social media. Each post has been separated by three pipe characters . This data has been collected from the users of an online forum, where in the first step, users take a questionnaire that recognises their MBTI type; and in the second step, communicate with other users.
In this step, matplotlib which is a Python 2D plotting library were used for data preview and to determine the distribution of the MBTI personality types in the dataset.
The image above show a non-uniform representation of MBTI types in the dataset that is not commensurate with the actual proportions of MBTI types. As a result, it was clear that some cleaning in the dataset would be necessary in order to improve the accuracy of the proportional representation of each MBTI type.
Four different categories were created for the type indicators in order to understand the distribution of types indicators in the dataset.
The first category was for Introversion (I)/Extroversion (E), the second category was for Intuition (N)/Sensing (S), the third was for Thinking (T)/Feeling (F)and the fourthcategory was for Judging (J)/Perceiving (P). As a result, for each category, one letter will return and at the end there will be four letters that represent one of the 16 personality types in the MBTI. For instance, if the first category is returning I, the second category is returning N, the third category is returning T and the fourth category is returning J, the relevant personality type would be INTJ.
For the first category of Introversion (I)/Extroversion (E), the distribution of Introversion (I) is much greater than Extroversion (E). Similarly, for the second category which is Intuition (N)/Sensing (S), the distribution of Intuition (N) is much higher than Sensing (S). for the third category which is Thinking (T)/Feeling (F), the distribution of Feeling (F) is slightly more than Thinking (T). Finally, for the fourth category which is Judging (J)/ Perceiving (P), the distribution of Perceiving (P) is greater than Judging (J).
Word Cloud is a data visualization technique used for representing text data in which the size of each word indicates its frequency or importance. Significant textual data points can be highlighted using a word cloud. Word clouds are widely used for analyzing data from social network websites. Word cloud was used to analyze the most frequently used words for each of the personalities.
As discussed earlier, data in this dataset was collected from an Internet forum and after analysing the content of the dataset, it was clear that some word removal is necessary. Non-uniform representation of MBTI types in the dataset that is not commensurate with the actual proportions of MBTI types in the general population was the most important reason for this. It was determined that this is because the data was collected from an Internet forum created for discussion about personality type and MBTI types were repeated too many times in the posts. This may also affect the accuracy of the model. As a result, NLTK was used to remove the MBTI types from the dataset. After this step, the distribution of MBTI personality types in the dataset was determined again. In addition, all urls and stop words were removed from the dataset. Finally, in order to make the dataset more meaningful, the text was lemmatised, i.e., inflected forms of the words were transformed into their root words. Imbalanced data was also handled where the random over sampler function was used. This ensured that the categoraization of type indicatores in four dimenstions were balanced.
Sklearn library was used to recognize the words appearing in the posts. It builds a vocabulary that only considers 100000 features ordered by term frequency in the dataset. In the first step, posts were placed into a matrix of token counts. In the next step, the model learns the vocabulary dictionary and returns a term-document matrix. The count matrix then transforms into a normalised TF–IDF representation.
Models were created and predictions were made using the classification methods below:
-
Random Forest: Random Forest is a classification method that combines many decision trees based on individual sets of examples from the dataset. Each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. In the Random Forest algorithm, the results of decision trees are combined to select the most popular class.
-
Naive Bayes: It is a classification technique based on Bayes' Theorem with an assumption of independence among predictors. Simply, a Naive Bayes classifier assumes that the presence of a particular feature in a class is unrelated to the presence of any other feature.
-
Support Vector Machine: A support vector machine (SVM) is a supervised machine learning model that uses classification algorithms for two-group classification problems. SVM training algorithm builds a model that assigns new examples to one category or the other, making it a non-probabilistic binary linear classifier (although methods such as Platt scaling exist to use SVM in a probabilistic classification setting). SVM maps training examples to points in space so as to maximise the width of the gap between the two categories.
-
Decision Tree : It is a classification method where the paths from root to leaf represent classification rules. A decision tree is a flowchart-like structure in which each internal node represents a test on a feature, each leaf node represents a class label (decision taken after computing all features) and branches represent conjunctions of features that lead to those class labels. .
-
XGBoost : XGBoost (eXtreme Gradient Boosting) is a popular supervised-learning algorithm used for classification on large datasets. It uses sequentially-built shallow decision trees to provide accurate results and a highly-scalable training method that avoids overfitting.
The accuracy, precision, recall, f1-score and ROC area were the metrics used to evaluate the performance of the model. In the figure below, the metrics helps us determine the best classification algorithm to use for prediction of the different indicator types. From the results, in making prediction for the first category which is Introversion (I)/Extroversion (E), it is best to use the Random forest algorithm in the prediction. The random forest algorithm had an accuracy of 0.952 which exceeds all the other algorithms, hence the model is selected for prediction. In making prediction for the second category which is Intuition (N) / Sensing (S), it is best to also use the Random forest algorithm in the prediction. The model had an accuracy of 0.993. In selection of a model for the third category which is Judging (J)/ Perceiving (P), the XGBoost is the most preferred model to use due to its performance against the other classification models. For prediction of fourth category which is Judging (J)/ Perceiving (P), the SVM is the most preferred model to use due to its performance against the other classification models.
From training of the dataset, we were able to come to a conclusion on the best model to use in each dimension. By using pickle, the models are imported and used. The snscrape which is a scraper for social networking service is imported to scrape a user's tweets from their twitter account after the user has entered their username in the application. The Human Resource team requires all users to have a twitter account. Users are required to be engaging in tweets and post three hours before they take the personality test. The models are loaded, the users tweets are pre-processed and predictions are made thereafter.
Type the twitter handle without the @ symbol.
Predicted personality of the user and some tweets.
Though curriculum vitae analysis is necessary to ensure recruiters do not miss good candidates or put forward incorrect candidates; but it is also important to process all applications quickly to ensure a fast response for clients and move to the interview and placement stages. In our project, our goal was to parse the entire CV and search for keywords from the CV.
For parsing CVs, we have used pyresparser which is a simple resume parser used for extracting important features such as name, email id, description, skills from CVs. Pyresparser supports PDF and DOCx files.
- A registered candidate can upload the CV.
- The candidate can visualise the score based on resume uploaded
- Candidates gets resume writing tip suggestions if necessary.
We liken the score of the CV to that of an ATS score. The ATS does not know how to handle tables or images. It is best to use a chronological or hybrid curriculum vitae format. Since the CV will be scanned, headers and footers are not necessary. An improperly formatted CV cannot be scanned hence the candidate could be disqualified.
WeEmploy is still an experimental prototype however instances fit for specific use cases can be spawned and developed for your use. In order to contact us for such an endeavor please check out the contributors for this project.