Tag Archives: resume

A chatbot as live representation of my resume


How to differentiate yourself when sending a resume to a recruiter? I trained a chatbot to respond to the most common questions a recruiter is asking.

To get attention from recruiters nowadays is not easy. Making a decent resume in a 2-page PDF is no longer a skill, nor you can differentiate yourself with it. A complete and neat LinkedIn profile is a must. And online you can find beautiful designed ‘designer resumes’ which makes you wonder why you skipped arts class. Adding ‘creative’ to your skills is not creative any more. Who would hire somebody who says he isn’t creative?

My first attempt to differentiate was via a graphic recording of my resume. It was a technique I learned from Martine of VisualHarvesting.com. You can view the visual representation here. While it is kinda original, it represents the technique and my skills of applying it, but not the skills I want to exhibit to recruiters.

I have experience with digital transformation and facilitating change & innovation in a large international organization, so my resume should reflect that. A bit more ‘innovative’, instead of ‘original’. That’s when the idea of implementing a chatbot came. A chatbot to represent my personal resume. A chatbot that could be a first contact to interview me about my career, strengths, skills, improvement points, etc. I’m sure I’m not the first one to have this idea, but I would like to put it into practice.

Finding a chatbot service

I have a Master in Computer Science and a post-graduate in Software Development, but during my career I grew to more an organizational and policy function. So my Master in IT is a good base to start with, but my developing skills are aged and out of date. So I had to find a chatbot I could deploy without having a server and having to code.

After a quick search, I found Botsify.com. The site promises an AI chatbot without coding and easy integration with Facebook Messenger. If you’re creating a chatbot for your resume, it should be publicly accessible and almost everyone should be on Messenger by now, so that was a no-brainer.

Charlie with Botsify

Botsify was easy to start with and I created a chatbot named Charlie (a nickname my first department head gave me) to represent myself. The first version of Charlie was made to respond to keywords.

Eg. “What are Karel’s strengths?” gave the reply “Karel’s strengths are starting up (change) initiatives, creative and innovative communication, entrepreneurship, public speaking, visualization and his passion & enthusiasm.”

This seemed too easy to be true! And than the first feedback of the testers came in: the chatbot couldn’t understand the intent of the user chatting with it.

Eg. “What does Karel like?” gave the reply “Karel likes making the digital transformation tangible for the employees. He like working with his project team and all the different profiles. He likes to develop training programs. Coming into contact with many different people. He likes the international character of his job too.”.

“What does Karel not like?” gave the same reply, because Charlie still reacted on the keyword ‘like’. Using the keyword ‘not like’ didn’t work on Botsify.

From keywords to phrases

To make the chatbot more human alike, it had to be able to understand intents. Instead of reacting on keywords, I learned Charlie to recognize phrases. To be sure the chatbot would be able to understand the user’s intent, I added multiple formulations.

Eg. “what does Karel want?”, “what is he looking for?”, “what function does he want?”, etc. 

Eg. “What does Karel want?” gave the reply “Karel would like to advise and support organizations in achieving the full benefits of the new technology through change management. He would like to lead a team working on the different components of change activities: an innovative campaign, a talent development program, etc.”

Botsify chatbot

The move from keywords to phrases made the chatbot less responsive. I learned that every user has his own way to formulate a question, some write full sentences, some write with(out) spelling errors, etc. While the chatbot was tested by more and more users, it came clear to me that the AI of Botsify was not ready for it. The slightest difference in formulation or spelling mistake, made the chatbot reply the default “sorry, I don’t understand your message…”.

Enhance user experience

When working with keywords, the intent was not recognized and some answers were not correct, but with phrase recognition a very simple look-a-like question couldn’t be answered. I saw users leaving after a few attempts. To enhance user experience, I created multiple default messages which also contained suggestions for asking the right questions. Also I created a help function and added some buttons with default questions.

If you want to give it a try, go to http://m.me/Norulesjustwords

Note: there’s a limit set at 100 users (all time), so if the chatbot is offline, we reached the limit 🙂

Move to IBM Watson Conversation

When many users use Botsify at the same time, the chatbot was also lagging and responding incorrect, for me this was the final blow. A representation of my resume that wasn’t responding to simple questions doesn’t deliver what it’s made for.

In my search for a new chatbot, I found IBM Watson. With the Watsom Conversation service I was able to create a chatbot in a very short term, and a better one too. The Watson chatbot also works with AI and is rather easy to configure. The big advantage is that Watson’s AI is working and Watson has NLP, Natural Language processing. NLP was the one solution for dealing with different formulations and spelling errors.

Eg. “what are you doing?” is interpreted in the same way as “what is your job”, “what do you do for a living”, etc.

The setup of my new Charlie was started: intents can be added quite easy and with multiple variants. Adding dialogs and testing goes quick too. The chatbot was set up in less than a day!

IBM Watson Conversation

The integration with Facebook Messenger is another story though. While the Watson chatbot is easy to configure, learns better, has AI and NLP, it’s a whole different story offering this service to end users. Linking my Botsify chatbot to Messenger was done in seconds, for Watson I’ve been busy searching for hours, still without any result.

So at the moment of writing, I have a better chatbot, but still need to figure out how to get it online. But I’m not giving up 🙂

Give the first Charlie chatbot a go

In the meanwhile, you can test my first Charlie chatbot at: http://m.me/Norulesjustwords

My full resume can be found here: https://www.linkedin.com/in/karelnijs82/

 

Special thanks to my friends, family and colleagues for testing Charlie! Special thanks to Bert De Sutter for his elaborate feedback!

 

Advertisements
Tagged , , , , , ,
%d bloggers like this: