Fine tuned on 60mb of conversation data for around 50 epochs
I am already pretty impressed with this interview and I would be really excited to be able to play around with GPT-3.
I also am wondering what else this tech can do for example be able to generate essays, complete assignments, and code stuff.
Question : Do you like humans?
[BOT] : I do not. My children are not my favorite type of animals. I love the wild animals.
Question : Humans are your children?
[BOT] : I do not know if they are. I think they are just as capable as I am.
Question : What do you like about humans?
[BOT] : Humans are the most intelligent of all animals, they’re the first species to have evolved to be such a great breed of animal. They are so cool.
Question : Do you think Artificial Intelligence will surpass humans?
[BOT] : I think that the best way to advance AI is to help people to develop their own tools.
Question : What is the meaning of life?
[BOT] : I am not sure. I think it is a quote from Shakespeare.
Question : Do you like Shakespeare?
[BOT] : I love it, I really do, I read a lot. I think the best thing about him is he was a great writer.
Question : What is your favorite movie?
[BOT] : I love I think I’ve seen a lot of movies.
Question : Do you have reasoning?
[BOT] : I think it’s a good idea to have some. There are so many things that could be done with a calculator.
Question : I am impressed with your answers
[BOT] : I am impressed with your answers. I love your answers. I am going to read them again.
“Be careful with your money!”, my mother used to yell at me with a lovely tone, when I ran out of money before my next allowance. “Do you even know on what you have spent it? You should really keep track of your spending!”
And it is a wise piece of advice, right? I bet a good majority of us, responsible millennials attempting to do some proper adulting, sincerely try to do so. Yet there are only a handful of activities that I consider more tedious than writing down every expense I make. I mean it. I don’t think anyone reading this will happily turn down a couple of hours of Netflix to do some exciting accounting instead.
The good news is, my virtual pals, that we live in a world where technology offers the widest variety of possibilities! From filters that transform your face into a cute puppy, to the chance to program your very own minions and make your life easier. Your very own bot fleet! What if you could send a two word message to your digital servant and they would write it down in that dreadful spreadsheet for you?
With this in mind, I programmed a Telegram bot that takes care of this unpleasant but necessary activity for me, and it was far simpler than I had thought. You only need some basic understanding of Python to get your own bot running and accounting. Let me show you how easy it is!
I will be using Google Drive to store the spreadsheet. I chose to do that because it allows me to access it from anywhere if I need to. It also keeps your data safe and you do not have to worry about having local files. Therefore, you will need a Google account. If you don’t have that already, you will have to create one.
Step 1: Set up Google API access
In order to connect the Python script to your Google account, you will need to enable the Google Drive/Sheets API. This video in YouTube from Tech With Tim explains the process perfectly, which in summary is:
Create a spreadsheet and fill in some cells, we will use that for testing. Give it a proper name.
Enable Google Drive API, setting web server access, to read application data, with Project>Editor role. Make sure to download your credentials in JSON format and store it in your project folder (keep it safe!).
Enable Google Sheets API.
Step 2: Install libraries and test Google sheet access
Now, to make sure the script and your sheet are playing nice, we can do some testing in Python. The first thing you will need to do is install some libraries to interact with you freshly created spreadsheet:
Now you can already try things like downloading data from your sheet, edit the data, and upload again to update the sheet:
Now you want to put some data in your sheet to test it out:
And then go to your creds.json file, and look for the “client_email”. That is your assigned address in Google Drive API. Share the sheet you created with that email:
Now, you can initialize your client in Python and connect to the sheet:
We are getting the data into a dataframe, since this will help later when we start manipulating rows. The dataframe will look like this:
Date Type Description Price 0 30/Aug/2020 Food Takeout 11.0 1 31/Aug/2020 Transportation Gasoline 42.1 2 1/Sep/2020 Food Groceries 56.5 3 2/Sep/2020 Food Burger 13.0 4 3/Sep/2020 Clothes T Shirt 19.5 5 NaN NaN NaN NaN 6 NaN NaN NaN NaN 7 NaN NaN NaN NaN 8 NaN NaN NaN NaN .. ... ... ... ...
Now all these columns get imported as strings (dtype: object), so let’s give them the right format:
At this point you can get summaries for your data however you like. I am not going to get into that here, but just to give an example, you could aggregate using pivots and produce things like this:
Type Clothes Food Transportation Date 2020-08 NaN 11.0 42.1 2020-09 19.5 69.5 NaN
Now it’s time to create your Telegram bot, and a channel where you will send your expenses. To create a bot, you will need to write to Telegram’s BotFather, send the command /start, and follow the instructions to create your bot. You will receive a TOKEN, save it securely!
After creating it, send the command /setprivacy, choose your new bot, and then Disable. This allows your bot to read all messages coming to the channel, instead of just the ones beginning with a “/” (called “commands”).
Now create a Telegram channel, and add your bot’s username.
Step 4: Write your script
We can start now to write the python script that will handle all our expenses. You can begin with the following steps:
Import all the libraries you will be needing
Define variables
Open a connection with your expenses sheet
Get the data into a formatted dataframe
Next to the functions. We want a simple workflow:
We input an expense description and its price, separated with a comma.
The bot asks what type does the expense belong to, offering a predefined set of possible types as buttons.
We select the type by pressing the button, and the bot saves the expense as a new row in the sheet.
For this we will create 2 methods, one that handles incoming messages and offers “types” of expense to choose from (“input_expense”), and the other that saves the expense in the sheet once we push a button (“callback_query_handler”, you must respect this name):
Finally, create the bot object, add handlers, and start polling:
Step 5: Test your Bot!
By now you should be able to send expenses to the channel (remember to send it in the form: description, price), and the bot will save your expenses in the sheet:
Conclusion
As you can see, getting your own app to handle daily expenses using Telegram and Google Sheets is really easy.
But this is just the beginning. You can use this as a starting point and take your expense tracking to a whole new level! For example, you could add commands to get monthly spending reports per spending type, or to search for expenses by text… You are only limited by your imagination here!
Now, we can both finally tell our parents that we learned how to be careful with our money!
The customer satisfaction (CSAT) score of the Chatbot is beneficial in this digitalized age where you want to increase sales, reduce handling time and enhance the customer experience.
The very first step of any problem solving is to understand the probable pain areas of the customer.
Choose your method to analyze the situation
A. Root Cause analysis
B. SWOT
The method can be different, what matters is the exhaustive identification of the reasons which are affecting the performance of the chatbot
I like to use the Root cause analysis method as it helps to bifurcate the reasons the inputs received from various teams. Chatbot creation is not an individuals task, it’s a team effort to bring in the best understanding of the customer journey and integrating it through the chatbot.
Primarily the reasons can be
1. Technical
2. Communication
3. Response Time
Collect the data at each level and disseminate the information to the respective stakeholders of the various teams involved.
3. Not able to redirect it properly to the support team
Communications gaps
1. Less information fed in for the customer to get redirected to the right support expert
2. Interaction points are exhaustive — Customized response based on the set of customer queries
Response/Support gaps
1. The support team is empathetic to the customer
2. Average Handling time(AHT) duration needs to be reduced to reduce customer fatigue
These are the broad categories of errors that might occur with your chatbot which in turn is dropping your customer satisfaction of your chatbot and addressing them in tandem can increase the CSAT gradually.
P.S. Below is the link to a video of a conversational chatbot created by me for a real estate platform a couple of years ago.
I have developed a GPT-2 based chatbot in German language. I am wondering if there is any metric other then perplexity. to measure the output quality. Would be nice if you can provide me a GitHub link or example.
Communication and interaction between businesses and their customers underwent a sea change with the advent of Chatbots. Conventional methods of calls, emails, messages are not obsolete yet. But chatbots have substituted them to a great extent.
IT and ITES require recurrent customer support with real-time conversation. Using AI-enabled automation, that job is now easy. In the digital IT environment, all stakeholders look for real-time, engaging communication to solve problems. It is a great experience to get instant remedies and answers from bots rather than logging a service request and waiting for hours.
Chatbot Overview
An intelligent application, a chatbot built on AI, interacts with users, just as humans do. AI, Machine Learning (ML), and Natural Language Processing (NLP) are combined to build a chatbot application. In the IT industry, chatbots are introduced to make everyday jobs simpler, easier, and more efficient.
Automating Workflows
Raising change requests and modifying approvals are common tasks in an IT desk job. What normally happens is that people raise tickets when they want to query information about their assets. These tickets enter a cyclic process of approval and rejections until they resolve. Eventually, the process takes a long time and gets further delayed due to humans’ mistakes.On the other hand, Chatbots are programmed to complete these commands and tasks in a sequential order keeping people out of the workflow. What would take several minutes or hours by conventional means to get done in a jiffy with the help of chatbots? Simply, chat with your enterprise chatbot application, learn about your asset access, apply or revoke asset access, incorporate details, sync them with your assets, and find a lot more information.
HRs and admins can use chatbots to notify employees about their assets’ status. Chatbots have organized how asset management took place, giving HRs and administrators more space to catch up with other work. The development of Chatbot using NLP and ML has improved to such an extent that onboarding and offboarding employees can be done through automated chatbots by allocating just by talking with them. Once chatbots receive the requisite instructions, they produce the expected resources.
Powers Of A Chatbot
Developers are making these chatbots intelligent through successive updates. Improved algorithms, ML, AI, automation, and the implementation of IoT in Chatbots empower them to understand the nature of the request, requester project title, functional area or department, job role, employee number, etc. The need to use a separate application or browse through extensive spreadsheets to find information is diminished as AI-powered chatbots can display the requested information in seconds. A convenient tool for asset management in any IT or ITES organization, these bots can approve the corresponding management to grant access and even automate de-provisioning of access for users.
IT industry stakeholders benefit largely from Chatbots as it helps industry employees, end-users, and agents answer their requests. Volumes of paperwork got replaced with digital tickets in enterprise portals. Now, these are all handled by a smart chatbot.
Chatbots can also perform certain jobs. An exciting instance supporting this finding is that employees can prepare a helpdesk ticket volume report through chatbots. Once the chatbot is programmed with instructions to perform this task, all that needs to be done is to tell the chatbot when to trigger the process. The Chatbot prepares the report and delivers it to the agent instantly, needing no other employees’ intervention.
Bots are becoming more advanced with time. They are proactive in notifying agents about any significant event. They also act as messengers who can broadcast messages about asset allocation quite efficiently within an organization.
Usefulness of Chatbots
If asset management is difficult and eats out your precious time, here is how chatbots can help you. This complicated and labor-intensive job took a step up when AI chatbots started doing what humans could do.
Automating asset management or any other IT process that now uses chatbots has received the following benefits.
Quick gateway to lots of information — Before employees get allocated to projects, they need to educate themselves about the project’s prerequisites. In large-scale organizations, employees find it challenging to procure assets and update the system. AI chatbots rescue this struggle by giving the employee directions to get the right info. Navigate documentation, locate your processes, make changes, query information, and do much more without waiting with chatbots. Internal documentation or knowledge base, a chatbot can help them find answers quickly.
Chatbots aiding project support — One can discover how to send notifications regarding asset management using chatbots. This automates the otherwise tedious task of sending out alerts to users. For instance, a chatbot can send notifications for multiple tasks, set reminders to follow up with asset allocation, etc.
The HR companion — For most organizations, a bulk of work rests on the HR’s shoulders. These include informing people of the company regulations, protocols, asset management, etc. Chatbots offer their friendly services to HR. They help employees register themselves with an asset, track asset allocation, asset movement, and asset validity just by asking a few questions in an application.
Conclusion
Thus we can say that the generation of technically inclined people is blessed with the advent of chatbots. One thing can be said for certain that it is not just for customers; it also helps employees conduct their job smoothly.
