DEV Community: MarkatSentient

iHack Alpha by Sentient.io Unveils AI Powered Solutions for Business & Social Challenges

MarkatSentient — Wed, 10 Feb 2021 03:09:25 +0000

As a known fact, the retail scene globally has been in a slump ever since the rise of e-commerce and digital devices. The pandemic-stricken year has pushed the sector further into gloom, with experts predicting sales to be declining by at least 10% year-on-year in the Southeast Asia region. On top of that, businesses are grappling with the challenge of engaging increasingly distracted customers in the digital space. With the tighter squeeze, brands are also re-looking at innovative yet cost effective ways to enhance customer service experience in order to build loyalty.

Recognising these challenges that companies are facing, Sentient.io has launched an online hackathon, the iHack Alpha, to tackle the aforementioned problems through AI-enabled solutions. The hackathon, which concluded on 5thJanuary saw over 1000 developers who participated. Developers could work either individually or form a team of maximum four members to work on any of the four stipulated themes: Retail, Customer Service, Media, and Accessibility.

In each of these themes, a problem statement is provided, and developers are free to identify opportunity areas they would like to explore and ideate upon. To illustrate, the problem statement given under the theme of retail is: How can you use AI to help retail evolve and stay relevant? With this, developers could for instance, identify how more retailers are adopting the online-to-offline (O2O) strategy to drive foot traffic to physical stores. Developers can then propose and design tools that will help businesses overcome the digital blind spot in marketing when going from O2O.

Contrary to the common “job stealing” notion that has been assigned to AI advances, the many innovations that stem from this hackathon alone, has proven that AI can serve as an enabler.

Among the eight shortlisted ideas to the prototype phase, is the GeoFencing project. The project aims to help market the small local businesses or mom-and-pop stores who are overshadowed by giant corporations and do not have an online presence. The solution is to develop extensions or mobile apps that will help consumers to locate products that they are keen on purchasing, nearby. The idea is to help people get to know the shops in their vicinity. This provides the smaller businesses the opportunity to be considered as an alternative, without having to spend exorbitant dollars on advertising. To set the idea in motion, the team had employed various microservices from Sentient.io, including the Object Detection Microservice, Web Scraping API, as well as the Base 64 Encoding and Decoding Service.

Empowering the Disabled at the Heart of iHackAlpha

With a large proportion of the submissions targeting the accessibility theme, the single socially related theme is seen to have gained the most traction among contestants. An 80% unemployment rate in Southeast Asia, for persons of disability is indeed a note for concern. Many teams of developers have expressed with full conviction in their presentation that this societal group has just as much to contribute as the general population, if they were provided with proper assistive technologies.

“Even with valuable skills and experiences, visually impaired people often find great difficulty in securing employment. After so many technological advancements in the last few decades, we still fail miserably at disability inclusion in the workplace. One of the main reasons for this lies in the fact that organizations fail to deploy assistive technologies for the visually impaired at the workplace.

The most sustainable way to include the visually impaired population in the workforce is by considering it as a value-add transaction instead of going from a social responsibility perspective. Considering disability inclusion as a value-add transaction means that we want to bring in a disabled person into the workforce not because it is a social responsibility but because they have the necessary skills and resources to bring value to the organization. Therefore, in order to solve this problem of blind inclusion in the workplace, we must focus on making assistive technologies more accessible to this population,” said one of the finalists.

The group then went on to propose a smart glass, which they named iGlass, to help visually impaired persons perform more productively at the workplace.

“In this device, we attempt to integrate features from different assistive technologies like screen readers, canes etc. into a single device using recent technological advancements in artificial intelligence and embedded systems. The visual data from the surrounding is captured using a camera mounted on the smart glass. This visual data is then fed into a pipeline running on Raspberry Pi to generate auditory data using different sentient.io microservices and AI models,” the group explained.

In other words, the necessary information captured through features such as image captioning, face recognition, emotion recognition, optical character recognition is then conveyed to the person seamlessly in audio format, through the bone transducer. It simulates the voices in one’s head that relays information regarding the immediate surrounding.

“I was impressed with iGlass, though Google Glass had attempted this and failed, at least for now. It was amazing how much the team accomplished in a short period of time,” said one of the six judges, Ms Christy Davies, Executive Director of Lien Centre at Singapore Management University.

When asked what she was looking out for in the submissions, Ms Davies said: “First and foremost, did the invention or solution meet the needs of their defined client? I looked for a clear problem statement and a systematic thought process which then used data or AI as the tool and vehicle to tackle their problem. I was also curious to a clear motivation of why they chose the issue they did because a solution is often more effective when it is emotionally or very practically connected to the heart and passion of the entrepreneur or coder”.

It was certainly heartening that despite it being a hackathon contest, many teams bore the common goal of empowering and improving lives through their innovative ideas and skillful application of AI technologies.

Winning Concepts and Prototypes

The project which clinched first place was titled the Artificial Voice, a two-way chatting app for blind-deaf people to communicate with others. To send a message, the blind-deaf person can simply tap the screen, using either short tap (representing dots) or long taps (representing dashes). The Morse code written by them is then converted into English text for the other party to read. On the flipside, incoming messages from the other party are converted into vibrations of dot & dash for blind-deaf people to understand.

“The idea for the project is inspired by the Helen Keller story we studied in school days, where Anne Sullivan used to write on Helen Keller’s palm to communicate with her. So, we thought why can't we replace Annie with a mobile phone, which can communicate with the deaf blind by sending a message with haptic feedback,” said Mr Harsha Satya Vardhan, the team leader of project Artificial Voice.

Taking into consideration that subjects could be together at the same place or communicating across different locations, the team behind the project has also included the ‘near’ and ‘far’ feature. The ‘near’ feature can be activated when both parties are at the same place. When the person without disability speaks, it is converted into vibrations via Morse code in real time for the blind-deaf person to understand the message. The blind-deaf person will also be able to give a reply in real-time by tapping on the screen of the app. Using Sentient.io Text to Speech and Automatic Speech Recognition services, the Morse Code is converted to speech that is heard audibly by the other party.

“I like the Artificial Voice submission because it is something which can really benefit a group of very marginalized people and bridge some of the communication gap with others.” said one of the internal judges, Eddie Leong, VP Operations of Sentient.io

The team behind Artificial Voice was overjoyed when they were announced as the winner of iHack Alpha Challenge.

“Mahesh and I have been working very hard for three months on this project and had many iterations. We look forward to improving the current prototype and the lives of the visually impaired individuals through our project,” said Mr. Vardhan.

The second and third place were awarded to project MeetDocs and Buddy respectively. The former is a website that allows people with hearing impairment to take voice notes during a meeting. Meanwhile, the latter is a mobile app which resembles a virtual mentor or friend for people with neurodevelopmental disorder, including Autism spectrum disorder and dyslexia. It works as a translator that translates pictorial representations to normal text and vice-versa to fill the communication gap in a virtual working environment.

With regards to the concept behind the project, the team explained that people with neurodevelopmental disorder often face difficulty comprehending long paragraphs, along with the tone. The app can help users to identify a variety of tones present in a sentence, including the emotional aspects and language style. It also provides a visualisation of words, as well as additional related content (Wikipedia article) for better understanding. The project is intended to provide these employees with greater ease of communication and in turn, efficiency at work.

“The biggest challenge we face while working on our project is managing our time since we have limited time developing, planning, and testing the products. The other challenges we face were creating an impactful solution that addresses the problem and attempting to solve it to a more considerable extent.

“But our desire to make people's lives better using technologies and the support of Sentient.io mentors helped in the successful completion of the product,” said Mr. Ashutosh Gupta, the team leader of project Buddy.

Among the judges was Mr. Etsushi Tsuru, Managing Director of Asahi Shimbun Asia, who expressed that he was looking out for solution level, business scalability and profitability. He noted that the submissions contained unique ideas and concepts, which completely matches the frame of open innovation and intent of a hackathon as such.

Other criteria include, but are not limited to code quality, technology used, overall usability and user experience.

About Sentient.io

Sentient.io empowers developers and enterprises by enabling easy integration of AI capabilities into their solutions and software through the Sentient.io AI & Data platform.

Our platform allows developers with no AI experience to easily find and use powerful pre-trained AI models as well as data sets through API calls, allowing you to rapidly and effortlessly add AI capabilities to projects.

Building AI Solutions - A Beginners Guide

MarkatSentient — Thu, 21 Jan 2021 04:16:56 +0000

When you use an AI enabled application like a chatbot, a voice assistant or the smart reply feature in Gmail, do you wish you could build a similar smart application too? If you want to know how these applications got their cognitive abilities and in-depth knowledge and learn to build one yourself, you have reached the right page.

In this article we are going to cover a few basic AI concepts and scenarios that can be used to enhance almost any daily-use application. We will also look at options available to build these capabilities.
Finally, we are going to introduce to you a shortcut with which you can build your own mini-AI application within just 15 minutes of reading this article. With some API’s you can achieve this and much more. So let us get started on our AI development journey with some ABC’s.

The ABC’s of AI enabled applications

Robots, chatbots and voice assistants use one or many AI/ML concepts in the background. Let us have a quick look at some fundamental concepts and their applications before we proceed.

AI Models - The building blocks
AI models are algorithms that are “trained” using large datasets to learn and understand patterns. This enables the AI model to intelligently predict an output given some input information, classify items, and much more. An AI-enabled solution typically uses one or more AI models. All of the following concepts are implemented using different types of AI models.

Natural Language Processing (NLP)
NLP allows machines to understand and process input or data that is provided in human languages like English. Chatbots use NLP to understand and process questions from users. NLP is also used in voice assistants, for sentiment analysis of posts on social media and by email providers to understand incoming emails before suggesting a reply.

Computer Vision
As the name suggests, computer vision allows machines to process image data to understand what the image represents. Computer vision can be used for biometric authentication, quality control in factories, automated checkout in retail industry or for medical diagnosis of scans and x-rays.

Voice AI
While NLP helps to process the language semantics, Voice AI deals with natural speech with all possible variations in voice, pitch, fluency, accents and intonations. It is used in voice assistants and many other smart devices and apps that can understand what we are saying.

With this basic understanding we can now proceed to look at the tools, components and steps required to develop an AI based application.

Developing an AI enabled application from scratch

Several businesses today try to integrate AI features in their internal as well as consumer facing applications. Once the requirement for an AI component is established, the development takes place in several steps.

1. Data collection: The AI development team needs to determine the inputs and the expected output corresponding to possible combinations of inputs. Data collected thus is cleaned before it can be used to train the AI model.
2. Create algorithm: Prediction or classification machine learning algorithms are created that can learn from the dataset. Machine learning libraries such as TensorFlow or Pytorch may be used for this purpose.
3. Training: The algorithm is then trained using the data collected. During this process the algorithm may need to be optimized in order to achieve an AI model with high accuracy. Additional data may be required in certain situations.
4. Integration: Finally, an AI model with high accuracy can be integrated into the overall solution.

One can directly get to the integration stage by using pre-trained AI models that are available as APIs. These provide solutions for commonly required AI features and may be integrated directly into the overall solution.
APIs can be accessed via simple calls with the input and return the desired output with a high degree of accuracy. Let us see how exactly API’s prove beneficial to the development process and what are the different types of API’s available.

AI APIs to shortcut your AI solution building process

APIs help you to cut short your development time and effort. Instead of developing something from scratch, you use features and functions that have been developed and fine-tuned over the years by experts in the field.
When it comes to AI, APIs provide a readymade solution to the tedious problem of collecting and cleaning large amounts of data before you even begin training your model. Different types of APIs available today are.

Open APIs: Also known as a Public API, there are no restrictions to access an Open API as they are publicly available.

Partner APIs: A developer needs specific rights or licenses in order to access this type of API because they are not freely available to the public.

Internal APIs: Also known as Private APIs, only internal systems expose this type of API. These are usually designed for internal use within an organization. By using internal APIs to communicate between systems, an organization can improve the efficiency of its processes and systems.

Composite APIs: This type of API combines different data and service APIs. It combines multiple transactions sequentially so that the client can get the desired result in one API call instead of many calls to the server.

With this introduction to APIs we can now explore Sentient.io API’s in further detail.

Sentient.io APIs to implement expert AI solutions

When it comes to popular AI APIs, you will often find APIs in the fields of Computer Vision, Natural Language Processing, and Voice AI as the use cases are widespread and the research in these AI fields is relatively mature. Sentient.io APIs cover use cases in all of these areas. For example,

Computer Vision
Object Detection
People Counting
People Recognition
Natural Language Processing
Word-sense Disambiguation
Entity Recognition (Pick out key entities from a sentence)
Voice AI
Speech to Text (Automatic Speech Recognition)
Text to Speech x1

With this we come to the most interesting part. Let us start calling these API’s from your code and build an AI application.

A Quickstart for using Sentient.io AI APIs

Step 1: Register for free
Users are required to register on Sentient.io in order to start using the APIs. The registration is free of cost.

Step 2: Discover and select the right API for your use case
After registration you can explore the demo apps that are available on the platform. This should give you a fair idea about the capabilities provided by different Sentient.io APIs. For example, the following image shows a demo app that is able to provide identify objects in an image.

Once you have identified an API (Microservice) that you would like to use in your application, you can subscribe to it and use it. A limited number of API calls are free so that you can try the service before purchasing it.

Step 3: Construct your API Call
A typical API call consists of the following components:

Endpoint URL: This endpoint is provided by the API provider, your HTTP request and payload will be sent to this endpoint.

API Key: The API key is a string that is unique to your account to allow the API provider to authorise your request. You can find your Sentient.io API key on the microservice subscription page.

Payload: Here you will include the input information in the format stated in the API Documentation. The API documentation can be accessed from the subscription page on Sentient.io.

With these components ready, you can now construct your request in your development language. With Sentient.io, you can also select from popular languages on the platform and copy the pre-constructed code snippet to your application as shown.

Your application is now AI-enabled. You can initiate a request to the API through the application to ensure that the connection is in place and you are getting the expected response. Thus, in three quick steps you have a ready-to-use AI enabled application.

You may have come across other API providers. The next section tells you how Sentient.io stands out from other similar service providers.

Why should you choose Sentient.io?

Stand out from the competition
AI-enabled solutions not only help to save time and effort, but also complete tasks at scale that would otherwise be considered impossible. In many cases they are able to outperform humans in making accurate predictions. Using AI in your applications will help your app to stand out from the competition, as well as deliver greater value to your users.

AI-enable your app in 10 minutes
With Sentient.io’s APIs you can add AI capability to your applications and solutions in as little as 10 minutes. Sentient.io’s services are domain targeted. This implies that different flavours of the API have been trained using data specific to different domains. For example, if you train an Automated Speech Recognition AI Model on an American accent, it will be optimized for the American accents. Sentient.io’s microservices are optimized for specific domains so that you can get better accuracy when you select the right AI model for your use case.

Reliability and Security
Sentient.io is different from API Platforms who compile APIs from different 3rd parties and are not responsible for their performance. Sentient.io understands that reliability and security is a top determining factor in incorporating APIs into your application. Hence Sentient.io has a variety of Service Level of Agreements (SLA).

What this means is that performance will be guaranteed according to the SLA applicable for that microservice, giving you confidence that you can rely on the service to be up when you need it.

Sentient.io is able to deploy microservices on premise or in an air gap environment. This means that we can deploy our platform completely detached from the internet, this is useful for companies who have very strict security requirements such as government agencies and banks.

Sounds promising? Join us now!

So if you are eager to kickstart your AI development journey, now is a good time. Sign up on Sentient.io as a free user and you receive complimentary 2000 Sentient.io credits that are worth USD $20. These can be utilized to buy additional API usage on Sentient.io. Use the promo code GetStarted2000 during sign-up.