Benji: A newly revised version of the chatbot!
Part 1
The feedback received by the master students and faculty varied a lot and also helped a lot to get some crucial input.
Out of the various points mentioned, one that we really need to analyse further in detail is how to give more depth to the conversation between Benji and the user.
1. There are various ways in which we can make the conversation more human, as by finding out.
the various attributes in the conversation that Benji himself has usually with faculty in our device (Benji). We looked into it by really talking to Benji and gain useful input from him to understand his personality and while not trying to literally make our device personify him but instead try to reflect on a deeper level why the device is named Benji and how it is really replacing the need of a human being in solving that situation.
2. Also by looking into other chatbots which usually answer to a lot of general questions from a user with a sarcastic or a light response to make the conversation more comfortable. We looked at it and interacted with other chatbots and mapped their responses to certain general questions.
3. Focus on the weak points where a facilitator would usually stumble or get confused and have an in-depth study on that problem and use that input in Benji’s response system to make it easier to solve those specific queries (eg, FAQs).
.
Leverage Benji through conversations
1. A very useful way of doing is that by introducing more points at which Benji initiates a
conversation with the user and that can be done by making Benji explain what he is doing every time before an action that he performs. This can be set every time there is a new user that interacts with Benji which it can detect by voice recognition. This is very useful in the way we can leverage Benji through conversation in this scenario.
“What’s up with Benji’s accent?”
For better understating how to improve the accent, we looked at other automated voices used in other conversational AI such as Siri, Alexa, Google home etc. Recently, Google introduced a feature through which now there is an option of choosing between 6 different voices by tapping on the different colours, and there is no colour which is specific to a voice. Also from these voices, we can take how clear they are, and how the pronunciation of words is accurate as well as understandable.
Move away from the standard format of Q and A conversations
Use Role playing method as a means
1. To explore the scenarios in our system physically and shape it according to time to the user’s
experience.
2. Construct scenarios through which we can gain an empathic understanding of our user.
3. Like his method can help in improving on one our feedback’s which was “ what’s up with Benji's accent?” By re-enacting this situation that we wanna improve.
4. This way we can better physically experience it and know where to concentrate on our main focus of improvement.
5. In this instance trying to feel the emotional experience of our certain solution can really help out if we ourselves enact the roles of a user and the service provider.
Personas are very important in any design process. And hence we created one of our device named ‘Benji’. The persona of Benji is actually telling the user about all the characters, functions and all that Benji is programmed to perform, which will help the user to understand and use it efficiently and will help to create highest human interaction. After designing the persona of Benji we actually realised the role and purpose of the device very clearly. This also gave us clarity on the nature of the conversation.
We restricted to keeping the conversation to the point as the conversations with the faculty in our tech department is generally restricted to requirements.
Part 2
A lot of faculty members find it difficult to work with the existing system of projectors and the converters required for the laptops. It is a messy business with multiple wires and screening problems. The settings required for starting the projector are never completely understood. There are often times when the faculty needs to access their personal files on the system while being connected to the projector and dig into big folders as they search for the right files. It becomes a time-consuming job.
How can we make this projecting experience seamless and using fewer efforts as well as keep it interesting and fun for the faculty?
Scenarios.
1. When a visiting faculty has to use a projector.
They do not exist in the default database of the college and also are new to the logistics in Srishti. They need assistance for connection to the internet and the projector connections.
2. The regular faculty has to use a projector.
Even though they know the system and the projector, they too sometimes face problems with the connections. Also, when the presentations are underway, they have to be near the laptop and cannot walk around the class. The presentation can be controlled from a distance with voice recognition and artificial intelligence.
Nature of Conversation
After interacting with Benjamin, we came across the fact that he is approached by the students as well as faculty only for technical requirements (adapters, speakers, errors with the printer etc). This is the main reason behind keeping the conversations simple and command-oriented.
The mechanical voice was used in the demo as we wanted to make sure that it was clear that the responses are coming from the device.
The main concept.
The main concept is to assist faculty members in conducting the class. Our approach was to incorporate assistance in terms of making the use of technology in class easy and hassle-free. This also aims at reducing the trips to the IT departments for problems as small as these. The faculty can focus in the class instead of being diverted by the mess of the wires and the unavailability of compatible cables. To add to this, connecting the projector/TV and speakers separately poses as a problem. The idea is to reduce all these complications on the faculty’s side as well as the IT Department’s side and reducing it all to digital assistants.
<VIDEO>
List of features
1. Voice recognition (setup) The device, since it works on voice recognition will have, in its database, all the recorded voices of the existing faculty by which it will respond on command. For visiting faculty the setup would require that particular faculty member to say their name, by which the device will record that voice and respond to that faculty by their name.
2. The device is connected to the projector by one common cable and thus the problem posed by varying laptop brands is removed.
An addition to this will be a USB port so that any laptop can be connected to it at any time. Thus, making it independent of the host system.
3. The device will follow simple commands like accessing files, projecting on the TV or the Projector. Other actions include to play or pause a video.
The content on the internet will be accessed by using the college Wifi.
4. Can assist the visiting faculty with simple things while conducting the class (“How do you plan on conducting the class?”, etc.)
5. Speaker and audio input cable. In case the faculty does not want to show visuals and just present audios, the device is flexible to do that as well.
6. Data recorded with respect to room numbers- the faculty members will specify which classroom they are conducting the lectures in, with their technical requirements for the class. This information will be present in Benji’s database and new requirements can be added at any instance.
7. Contacting Benjamin for more complex problems through simple commands.
Current Solutions:
The problem regarding the adapters was observed by Benjamin himself and he came up with the following solution; ‘WIreless Projection’.
In this approach, he wanted to introduce the feature of logging in to the system by using a code that was different for each classroom and hence the laptops in that classroom could connect to the projector in the room. According to him, this creates a lag in the playing of videos due to the load on the projector.
The existing system followed in Srishti involves the use of different kinds of adapters. Benji (the device) does much more than a basic adapter.
Future work:
With the introduction of this device, we hope that we can reduce dependency on the IT department. Since home assistants have raided the tech market, it is now time to think of subjective situations and how we can customize the voice assistant for the general public and make this technology more accessible and less intimidating for them.
https://drive.google.com/file/d/1GcdJmZNKk18iuuKNPR6JqjVZW0w61gRG/view
#link to the video.
<Code>
As we started writing the code for this, we realized that we can make it simpler by making it smaller using different modules the language. We used a dictionary to define a small database that keeps track of the logistics and the technical support of a particular classroom. We also explored how a CSV file can be imported.
The new module that we tried as well as the gTTS and Speech Recognition.
The code is as follows:
We also explored what will the Text to Speech module look like if used in this scenario.
The project was done as a part of a college assignment with Chinmayee Murugkar and Pranjal Sharma.
