Intriguing ML applications!
The second session of the Empower project included a discussion around various ML applications and the technologies behind them.
Candidate Shortlisting system using ML
Due to the digitalisation of candidate selection procedure, companies are receiving an enormous number of job applications. The manual process of shortlisting candidates has become very tiring and the process also introduces some biases, as the recruiters have to go through so many resumes and they can’t go through all the resumes at once and compare them and pick the best candidates.
Companies have shifted to a more optimised method, which is fast and free of biases. Companies have introduced machine learning models to solve the problem of candidates shortlisting. Companies make the candidate fill all his/her information like education, previous work experiences, skills, etc in a form, and apply machine learning to score the resume on the basis of targeted job description.
Various machine learning algorithms have been used to create smart candidate shortlisting systems. Like document matching using cosine similarity along with keywords matching. Keyword matching doesn’t consider only the exact words but also considers the synonyms or similar words. For example — Object Oriented Programming(OOPs) is the word stated in the job description, all the words related to OOPs mentioned in the candidate’s job application will be counted like C++, Java etc.
Cosine similarity example:
In job description(Sx): Require a web developer, skilled in CSS,HTML and JS.
In Resume(Sy): Book Library : Web App created using PHP,CSS,HTML and JS.
Count of words:
Require web developer skilled CSS HTML JS App created using PHP
Cosine similarity in Sx and Sy = 5/(2.64*2.82) = 0.669
Some companies have also used supervised learning to train their model. They have collected the training data by extracting the information from the applications of their current employees.
Actions on Google
Actions on Google is a developer platform that helps us to leverage the functionality of Google Assistant.
With Actions on Google you can create a Game between user and Google Assistant and can make users control smart home devices with Google Assistant.
Technology behind Actions on Google
● The technology behind this personalised voice assistant is Deep Neural Network(DNN) and Natural Language Processing.
● When an input is given to voice assistants, the voice is converted to text, analysed to come up with a reply in text and is then converted back to voice .As soon as the voice is transcribed to text using Natural Language Processing (NLP),it is analysed based on the context and tone of the data.
● NLP gives the ability to understand languages human speak and write. Apart from this Natural Language Generation which generates text and speech using predefined data is also used to respond to the data/input given by the users.
What happens during a conversation
Users have to give a name for the Action . So, when a user gives it as input Google starts responding for that.
Intent:A task that users want to do like finding a piece of music.In Actions on Google, that’s represented as a unique identifier and the corresponding user utterances that can trigger the intent.
Fulfilment:A conversation that handles an intent and carries out the corresponding action
I have created a game on C Basics and I have attached responses of the assistant with the test version.
