Cohere Thanksgiving AI Transformers Hackathon

TALDERA

Messenger apps like Discord & Slack have "channels" where multiple parties can send hundreds of texts daily. This can get out of hand pretty quickly if you've been offline, and we've sometimes had to scroll through hundreds of messages to understand what we've missed. The issue is that there isn't a quick & accessible way to figure things out. Tal;dera is a Discord bot that provides an easy-to-read, semantically rich summary of a channel's Discord activity in a particular time range. The summary is split into parts, and each part has a timestamp/link to its original messages. Users can call Tal;dera to extract message history and generate summaries (tl;dr's) of previous message chunks, given parameters such as number of messages, start time/date, output format, and chunk sizes. Summaries are given in an organized, embedded format, and are linked to where the conversations took place in the chats. Even though this was a challenging project for us, we were happy with what we were able to get done, and believe that this project has potential for B2B SaaS applications to team managers, employees, and chat service providers. We hope to continue this project by running pilot tests with smaller communities so that they can give feedback on Tal;dera, the summarizer bot for missed activity in Discord channels.

Health E AI assistant

By combining a custom conversant persona and grounded QA, Health-E is able to aid in processing patients, extracting relevant information to fill out forms, and providing advice common knowledge advice on applicable scenarios if prompted with a question. We hope this technology helps cut down healthcare costs by reducing queue times, paperwork, and increasing efficiency overall. Moreover, we envision applications of Health-E's underlying technology beyond the healthcare industry. In particular, regional economics and local communities in developing countries, that now have wide internet access, could greatly benefit from virtual advisors or teachers to supplement the lack of labor supply or poor level of education. We are really proud of what we built and we hope you enjoy both our presentation as well as experimenting with Health-E.

VerifAI

VerifAI is an AI-powered browser extension that helps you verify news from news websites. VerifAi - Detects the claim being made by a news - Checks if the claim has been verified by fact-checkers - Verifies the credibility of the website and author

TALDERA

Messenger apps like Discord & Slack have "channels" where multiple parties can send hundreds of texts daily. This can get out of hand pretty quickly if you've been offline, and we've sometimes had to scroll through hundreds of messages to understand what we've missed. The issue is that there isn't a quick & accessible way to figure things out. Tal;dera is a Discord bot that provides an easy-to-read, semantically rich summary of a channel's Discord activity in a particular time range. The summary is split into parts, and each part has a timestamp/link to its original messages. Users can call Tal;dera to extract message history and generate summaries (tl;dr's) of previous message chunks, given parameters such as number of messages, start time/date, output format, and chunk sizes. Summaries are given in an organized, embedded format, and are linked to where the conversations took place in the chats. Even though this was a challenging project for us, we were happy with what we were able to get done, and believe that this project has potential for B2B SaaS applications to team managers, employees, and chat service providers. We hope to continue this project by running pilot tests with smaller communities so that they can give feedback on Tal;dera, the summarizer bot for missed activity in Discord channels.

Health E AI assistant

By combining a custom conversant persona and grounded QA, Health-E is able to aid in processing patients, extracting relevant information to fill out forms, and providing advice common knowledge advice on applicable scenarios if prompted with a question. We hope this technology helps cut down healthcare costs by reducing queue times, paperwork, and increasing efficiency overall. Moreover, we envision applications of Health-E's underlying technology beyond the healthcare industry. In particular, regional economics and local communities in developing countries, that now have wide internet access, could greatly benefit from virtual advisors or teachers to supplement the lack of labor supply or poor level of education. We are really proud of what we built and we hope you enjoy both our presentation as well as experimenting with Health-E.

VerifAI

VerifAI is an AI-powered browser extension that helps you verify news from news websites. VerifAi - Detects the claim being made by a news - Checks if the claim has been verified by fact-checkers - Verifies the credibility of the website and author

Cohere Call Center

Automated call manager for companies. Uses voice detection and text classification. It's a microservices approach with python as code language.

Zipit

Using Cohere's API, we have built a solution that aims to significantly increase the productivity of every single busy entrepreneur through the automated processing of emails. Think of this application like a personal AI assistant that can help you with your emails.

Perfect Present

This app utilises the power of Large Language Models to provide gift suggestions for a person given their age, gender and interest.

ShubhMitra

Shubh-Mitra(शुभ-मित्रा), or a well-wishing friend. AI assistance for positivity. Positively reframe your thoughts and develop gratitude, thankfulness and other qualities.

revise

Writing is hard. Editing is harder. Why not make it easier for ourselves? Presenting re:vise, a contextualized writing feedback generation bot. Our bot aims to provide the benefits of a peer review without the peer! Imagine having a friend by your side that you can bounce ideas off of or get constructive criticism whenever you need it. Does that sound appealing? Try re:vise today. Just call 1-800-IRE-VISE! Just call 1-800-IRE-VISE! On a more serious note, re:vise aims to help people efficiently write clearly and concisely. This bot is designed to give high-quality feedback about the user's ideas. This type of feedback is more valuable than grammatical fixes because it makes ideas better which in turn improves the writing. On top of that, the bot will always be there when you need it. This is not the case for peer reviews when you could have delays for the review or just bad reviews. At the end of the day, I would rather have written a paper that had an amazing idea with a few grammatical mistakes than a paper with a flat idea and no grammatical mistakes. Which would you prefer?

vacancy finder

We decided to extract vacancy finder website through image text cohere ai, and then use non-code app in order to solve registration problems of the vacancies.

FAQ Generator for Chatbots

Creating content is tough, time-consuming, and requires experts. This is no different for content for chatbots, e.g., when creating a customer support chatbot that typically needs a knowledge base of FAQs (frequently asked questions) and their answers. This is typically done manually in practice and better tools are needed inside the customer experience (CX) platforms. We aim to provide a one-click tool for non-experts to create FAQs and responses for common use cases such as customer support. *Input*: the user can select relevant offline and online documents, such as a product description webpage, internal posts and emails. *Output*: the tool generates a list of FAQs and their answer in a number of uniform styles, e.g., formal, friendly, fun-with-emojis. *Deploy:* with built-in integrations the generated content can be published on a website or chatbot to serve automated answers. In this hackathon we focus on a proof-of-concept service that is able to extract text from a landing page and generate the company name, sector, and FAQs with appropriate answers. In particular, we demonstrate how zero-shot text generation works i) in a simple web application on Streamlit, and ii) on a web app on our systems (Helvia.io) that also puts the generated content into a running chatbot.

NLPNER

We have tried to build various models such as rnn, gru, lstm and bert transformer for our project and tried to build the app.

TheraLove

Matching effective mental health therapy that suits you the most, precisely using Cohere Classify artificial intelligence tool. According to TheraLove Bot's assessment of the lovely kind soul's personality, preferences, and current state of mind, TheraLove recommends the appropriate type of mental health therapy. Effective mental health therapy will be recommended after eliminating therapies that are unlikely to suit the unique and lovely individual. While this model had been tested and built at its Alpha Stage by a biomedical engineer, please take the recommendation as a complimentary identification method and seek help from social worker and mental health professionals as everyone’s mental health state varies. Self-diagnosis are not encouraged in any forms. The mission of this project is to bring you the best mental health technology, so please feel free to contact the team as seen on Github for any comments. If in pain or with suicidal attempts, please call your local lifeline from The List of International Suicide Hotlines or any of the lifelines as shown on Github.

Vessel

AI is really easy good at answering our questions, creating videos and even creating voice models. So many ground breaking tools but they can not understand each other? We are building a AI mediator that can answer your questions and knows how to talk to translate it to other AI tools. Our first use case is a text-To-Education model where you can input a question that runs through vessel generate an answer and translate this to other Ai tools. Vessel is backbone of this new age of AI media.

Sally

Celebrate thanksgiving beyond having a feast with our present loved ones! Meet Sally, who will help us reconnect with our lost loved ones. Sally is a bot powered by co:here Generate and co:here Classify, that will help us reconnect with our lost loved ones through chat. We can use Sally with just 2 simple steps: 1. First, add Sally on Telegram (https://t.me/ReconnectWithSallybot). 2. Second, Upload a chat archive with a lost loved one. Checkout the pitch: https://youtu.be/COjzdB5uyPs Checkout the live technical demo: https://youtu.be/Xc0aRSRrQcw What happens under the hood? When a user send a message to our Telegram bot, a webhook is send to our service deployed on EC2. Our server then call Cohere generate API to generate the appropriate response. But before we forward it to the user, we call Cohere classify API to determine whether the text is safe to send to our user or not! If not, we regenerate it again! Why? There are about 500 people grieving a lost one every minute, with 20% of them experiencing a complicated grief. Although we can not fix their grief, Sally is here to help the cope. Any undelivered thank yous, I love yous, and goodbyes. Say them through Sally. Because our lost loved ones are not truly lost, as long as we keep them in our memory. <3

Working with cohere chat

Traditional customer service web chat is dead. Customers have a poor experience while they wait for inexperienced customer service agents to respond to queries and work through their systems. It's not surprising that most customers come away feeling frustrated. Businesses must take the time to train staff but this comes at a high cost. Introducing co:here chat. Not only does this address the business challenges but it uses most of the components from the co:here stack. With co:here chat customer service web chat might not be dead after all.

Drunk driver identification

Thanksgiving is a time for getting together with family and friends. There is typically an overabundance of food but also the temptation to overindulge in alcohol. This makes it one of the most dangerous times of the year in terms of drunk driving car accidents. That's why we came up with this solutuion, an Artificial intelligence based (Deep learning and comuter vision) drunk driver detection system in cars will have a great impact in reducing accidents and increasing the safety for everyone.

Blue Fractal

Given a room full of strangers, blue fractal helps users find the most compatible people in the room. The host creates a room on blue fractal, the participants sign up and provide details like Casual interests, Professional interests, Things they can talk about for hours and a message to their match. Blue Fractal then pairs them up with each other based on these fields. It then sends out emails introducing them to each other. It also tells them why they were matched to help start the conversation.