Cohere and Qdrant Multilingual Semantic Search Hackathon Summary

The Multilingual Book-based Information Retrieval System is an AI-powered tool that allows users to extract relevant information from a given book in any language. The system works by first identifying the user's input keywords or sentences related to a specific topic. It then uses natural language processing and machine learning algorithms to search the relevant book for information on the given topic. The system also provides a translation feature that allows the user to translate the retrieved information and references into any language of their choice. This feature makes it easier for users who are not familiar with the language of the book to access the information they need. The Multilingual Book-based Information Retrieval System is beneficial for researchers, students, and anyone who needs to access information from books in different languages. With this system, users can easily retrieve relevant information from books without having to spend hours searching through pages. The system is designed to provide accurate and reliable information in a user-friendly interface, making it a valuable tool for anyone seeking knowledge on different topics.

Innovative Minds

Maverick REACT offers artificial intelligence integration for emergency situations. Our service uses AI with the necessary event information provided by government officials and acts as an assistant to provide key protocols and information to citizens. The AI service is accessed via SMS or web portal, offering a solution without internet. How does our service work? When an emergency situation occurs, such as a flood, fire or earthquake, our service sends an SMS message or makes a voice call to numbers registered in a database or the citizen can contact a number provided by the authorities. The message or call contains information about the type and severity of the emergency, preventive measures that should be taken and resources available in the area. The user can respond to the message or call with specific questions about their personal situation or request additional help. Our service uses AI algorithms to process responses and offer personalized and updated advice. REACT has several advantages over traditional emergency alert and response systems. Firstly, it does not depend on the internet, which means it can function even when there are power outages or problems with mobile networks. Secondly, REACT service is interactive and adaptable to the individual needs of each user. Thirdly, it uses reliable and verified sources of information provided by the government or other authorized organizations. And finally REACT is fast and efficient in sending and receiving large-scale messages or calls. Our goal is to contribute to creating a safer and more resilient world in the face of emergency situations through innovative and intelligent use of technology. We believe that our service can save lives and reduce suffering caused by disasters. If you want to know more about our service or how to register for it, contact us. We are Maverick AI.

MaverickAI

In a post truth world, the ability to quickly fact check any information against the available official public records and statements is a superpower that can hold public figures accountable, make gaslighting the public during election time more difficult and save democracies. It also cuts down distractions and time required to do research on public figures and reduces the entry barrier for political journalism and fact checking. Twitter is the internet’s town house and most of the public figures use it so the ability to fact check them in real time is critical. We aim to empower the general public with tools that make the world a better place by keeping people in power accountable.

Twitackers

a personalized news feed focused on the tech industry, powered by artificial intelligence (AI). Our news aggregator is specifically designed for busy CEOs, providing them with the latest and most relevant news in the tech sector. Through the use of AI, our platform curates and filters news articles from reputable sources, presenting only the most important and timely news stories to our users. This allows CEOs to stay informed on the latest trends, industry developments, and competitor updates in a quick and efficient manner. Additionally, our news aggregator provides a personalized experience for each user. By analyzing the user's reading habits and interests, our AI technology tailors the news feed to provide a custom selection of articles that are most relevant to their business and industry. Overall, our personalized news feed offers a comprehensive solution for CEOs who want to stay informed on the latest developments in the tech industry without the hassle of sorting through countless news sources. With our platform, CEOs can stay ahead of the curve and make informed decisions for their companies.

Dripper News

Legal-ease is a web-app hosting a suite of tools useful for simplifying legal documents for common use. In this way, Legal-ease offers tools to perform: 1. QnA over legal documents 2. Document summarization 3. Cross-lingual search This is great for people who dont have the time, are not native speakers of english or otherwise struggle with legal documents on a day to day basis. It is also useful for lawyers and researchers with access to a large collection of legal documents. Summarization is handled via cohere, QnA is handled via langchain, cohere's embeddings and LLM, and qdrant's excellent vector DB search.

Dragon

Scoper is a multilingual semantic AI tool for searching across internal documents. Although you may be able to search for exact keywords, we require an intelligent embedding of strings (in this case we use Cohere’s endpoint) to have similarity based on meaning, not just text. For example, the phrase “save cost by switching” may not show up directly in any internal documents, but there are mentions in several documents about benefits of switching programming languages or models of work.Additional features are documentation classification, where we label all internal documents both as the type of document (wiki, project proposal, or memo) but also with their completion status, i.e. finished or unfinished. We also support document generation, since in order to keep moving a company quickly, having AI help draft proposals or wiki’s goes a long way.

Code Monkeys

Currently, the most popular corporate knowledge management system is Confluence by Alatasian. It is known for a lack of search capabilities and makes most corporate knowledge inaccessible, especially in fast-growing companies where regular structure and responsibilities change. Some independent vendors fill this gap by offering carefully tuned solar-based search engines for Confluence, but not real semantic search. Confluence is a proprietary cloud-based solution, and it would be difficult to MVP a search extension in a hackathon. The most advanced open-source alternative is wiki.js, which already supports external search engines. So the current goal is to implement an external search engine for wiki.js using Cohere's LLM-powered Multilingual Text Understanding model and Qdrant's vector search engine. At the second stage of the project (most likely outside the hackathon scope), we plan to add the capability to upload and index videos in our knowledge management system. Recordings of presentations and meetings are the richest source of knowledge, but they were left outside knowledge management due to technical difficulties. Simple transcription and semantic search of that content could significantly boost corporate knowledge accessibility.

wiki search

Introducing a unique and immersive gaming experience that takes players on a hero's journey through the many aspects of human life. This game is designed to connect players with each other through a stunning astrological geometric universe, inspired by the 64 hexagrams of the i-ching. Players will be able to explore twelve distinct areas of human life, including health and fitness, intellectual pursuits, emotional well-being, character development, spirituality, love and relationships, parenting, social life, financial stability, career advancement, quality of life, and life vision. By playing this game, players will have the opportunity to discover their compatibility with others and make meaningful connections with like-minded individuals. The game's unique algorithm calculates compatibility based on the 64 hexagrams, providing players with a personalized and accurate understanding of their compatibility with others. Whether you are seeking a partner, a friend, or simply looking to connect with others who share your interests and passions, this game offers a fun and engaging way to do so. Join us on this epic journey through the human experience and discover what lies ahead!

IX

Problem: Government agencies such as the Department of Justice regularly release press statements to inform the public about their activities and decisions. These press releases contain important information that can be difficult to find and understand due to their volume and complexity. Currently, there is no easy way for individuals to search through these press releases and find relevant information. Traditional keyword-based search engines often return irrelevant or outdated results, making it difficult for users to find what they’re looking for. Solution: To address this problem, we propose the development of an app that provides semantic search results for queries on press releases from the government of justice. This app will use advanced natural language processing techniques to understand the meaning behind user queries and provide accurate and relevant search results. The major idea is that implementing the semantic search on public press releases will lead to a faster access to relevant information, Improved accuracy of the information, Increased transparency, and better data analysis. It can reduce the risk of misinformation by providing more accurate and relevant information to journalists and the public.

Aadarsh

In today's increasingly remote working style, organization’s messaging system, whether it's email or chat, contains lots of invaluable institutional knowledge. However, because these data are often unstructured and scattered, they are usually buried in the organization’s data ecosystem and are hard to search and extract value. Fetcher is a chatbot that integrates into popular chat platforms such as Discord and Slack to seamlessly help users find relevant people and documents to save them from endless frustrating search. It does this by semantically searching chat messages to find the most relevant results and help to deliver actions that leads to a peace of mind. Fetcher differs from traditional keyword search engines in that it searches by the meaning of the query, not just by keywords. It also enables multi lingual search, so that global teams can more quickly find important information even when language is a barrier. Since Fetcher searches in the embedding space, this search engine can extend to multi modal modes that includes audio and images. Fetcher works by collecting a chat channel’s history and embedding them using Cohere’s Embed API, then saving the embeddings to Qdrant’s vector search engine. When a new query comes in, Fetcher embeds the query and searches against the vector database to find the most relevant results, which can then feed into Cohere’s Generate API to summarize the message thread to kick start new conversations. Fetcher offers 3 commands, /fetch, using vector similarities search to find relevant chat messages. /discuss, summarize a message thread, and kick start a conversation with a channel number. /revise, a sentence correction tool similar to Grammarly, allows user to send professional sounding messages.

Fetch

Using OpenAi, Cohere, and Qdrant, vector search to predict the title, ingredients, and procedure of making a recipe based upon the limited food supplies that the user has by inputting the text in their own native language and getting the output in the same language as well. Moreover, we also wanted to use the OpenAI generative to predict and show the user how the resulting image of the dish looks based on the ingredients that the user inputted into the platform. Moreover, the next phase of this web app will be to make the model more accurate in terms of the predictive and analyze and do not give the user unrealistic outcomes.

FlavorFuse

Language Matchmaker is an app that aims to help language learners find conversation partners who share their interests and language proficiency levels. The app will use Cohere's multilingual semantic search technology to analyze user profiles and match them with other users who have similar interests and language skills. Users will create a profile on the app where they can specify their native language, target language(s), and interests. The app will then use Cohere's technology to identify commonalities between users and present them with a list of potential conversation partners. The Qdrant technology will be used to rank the matches based on similarity and provide recommendations for the best matches. Once a match is made, users can schedule a virtual conversation through the app and practice their language skills with their partner. The app will also provide conversation prompts based on the users' interests to facilitate the conversation. We highlighted the unique features of the app, such as the use of Cohere and Qdrant technologies to match users based on shared interests and language proficiency, and the ability to schedule virtual conversations through the app. You can also discuss the potential use cases of the app, such as for language learning, cultural exchange, or making new friends from different parts of the world. For the app development, we will use programming languages like Python that integrates the Cohere and Qdrant APIs.

Polyglot Searchers

Our platform provides businesses with the ability to monitor and analyze social media and news mentions in over 100 languages, all in one place. With our language translation and content intelligence solution, businesses can easily generate responses in their customers' preferred language, stay on top of industry news and trends, and streamline their content management process. Our platform allows businesses to upload their policies and internal documents in over 100 languages, generating chatbots and semantic search engines that enhance customer experience and streamline content management. We are utilizing cohere multilingual embeddings and Qdrant vector DB to generate and save embeddings. A flask-based api backend and a bubble.io based no-code front-end is used to build the solution.

AI Disruptor

Our project is aimed at developing a comprehensive legal document search system that makes use of advanced technologies to retrieve relevant legal documents that can be relied upon in court. The system utilizes Cohere's multilingual embedding and Qdrant vector database to provide fast and efficient search results. The use of multilingual embedding ensures that the system is capable of searching through legal documents written in various languages, making it suitable for use in multilingual environments. Qdrant vector database, on the other hand, allows for fast and efficient indexing of large volumes of legal documents, thus reducing search time. Our legal document search system is particularly useful for law firms, legal practitioners, and businesses that require access to legal documents for various purposes, including legal research, contract negotiations, and dispute resolution. With our system, users can easily retrieve legal documents that have been signed by mutual assent, thus ensuring that they are reliable and admissible in court. In addition to the legal document search functionality, we have also implemented a question answering system using Cohere's generate endpoint. This feature enables users to ask specific questions related to the legal documents they have retrieved and receive accurate and relevant answers. The question answering system is particularly useful for legal practitioners who require quick access to specific information in legal documents. Overall, our legal document search system provides an efficient and reliable solution for users who require access to legal documents. By leveraging advanced technologies such as Cohere's multilingual embedding and Qdrant vector database, we have developed a powerful search system that can save time and improve productivity for legal practitioners and businesses alike.

The Meowsterminds

We propose a powerful and innovative solution for enhancing the search functionality of ecommerce systems - the integration of a semantic search feature using Cohere's multilingual model and Qdrant's vector database. With the increasing volume of data generated every day, traditional keyword-based search engines can struggle to provide accurate and relevant results. This is where our solution comes in, offering a cutting-edge approach to search that enables customers to find what they're looking for quickly and easily. By integrating Cohere's multilingual model, our system can analyze the meaning behind search queries, rather than simply matching keywords. This means that even complex queries, such as those with multiple meanings, synonyms or different languages, can be understood and processed accurately, resulting in more relevant search results.

Savvy Synapses

"MedAI" is an innovative application designed to provide accurate and efficient medical question and answer services. Built with React for the frontend and Node for the backend, the app uses QDrant embeddings to retrieve relevant answers for user queries. The medical dataset is processed by a Python script powered by Hugging Face, while Cohere API generates embeddings for cosine similarity computations. This approach leads to a high level of accuracy and enables the app to quickly retrieve the most appropriate answers. Overall, "MedAI" offers a powerful tool for medical professionals and individuals seeking medical information.

AITeam

Heuristic AI brings browsing your Slack chat histories into a new dimension. Fueled by Qdrant vector search engine and the Generative model of Cohere, Heuristic.ai extracts the context from your question and matches it with your chat messages to elaborate the answer. Forget keywords and chats scrolling. We give you the answer and the source message in seconds! Vision: to enable people to find answers to any questions in their digital experience. Mission: to bring browsing chat histories to a new dimension How it works: 1. The user write normal query with the structure we have “hai, setup” or “hai, question” 2. Ngrok forward queries from slack_api to the Amazon server 3. Here, we evaluate the query to take action: - Setup from the sentence <hai, setup> or a sentence which contains hai and setup - Search: from the sentence that contains only the word hai - None, if the message sent in slack is a normal message 4. here we have two scenarios: - in the case of the setup action, we retrieve all the messages from all the channels, then encode them using co.embed prepare to be ready to store in Qdrant vector database - in the case of the search action, we encode the user query to retrieve the first 5 relevant messages from the conversations, then extract the answer to the user query from the first message retrieved using co.generate 5. Qdrant is the vectors search engine that allows us to store our vectors and to search on them. 6. Then lastly, the extracted answer is sent to the user.

Heuristic AI

Imagine a world where legal consultations are not only accurate and efficient but also intuitive and accessible. A world where the power of cutting-edge technology meets the intricacies of law to provide you with the best legal counsel, right at your fingertips. Welcome to the world of Cohere and Qdrant - a revolutionary website that leverages the power of semantic search to provide unparalleled legal consultation services. Cohere and Qdrant is the embodiment of modern legal expertise. Built with a focus on providing customized and highly efficient solutions, this website uses state-of-the-art algorithms to understand and analyze complex legal matters. Through its innovative approach, Cohere and Qdrant brings together the best of legal expertise and technology to help individuals and businesses navigate the intricacies of law.

The Soloist

We created Logos with the goal of enhancing people’s experiences in their relationship with their sources of knowledge: books and articles they read, podcasts and songs they listen, videos they watch and spaces where they discuss. Logos achieves this with the use of cutting-edge technologies to index and search for relevant content, assemble and interpret it in a semantic meaningful way. The multi-language feature of our solution grants super powers to our users, and our specially created expert agents are capable of answering very complex questions. We believe Logos will help heighten people’s passion for knowledge and thought, and that it will be a great tool for students, professionals and curious people alike!

Abacates Voadores

Recommendations cold-start problem is not actually a problem, if you leverage content and item metadata to build your recommendations. To showcase this idea we build a movie recommender, so you can visually see the difference between collaborative-filtering and content recommendations. We made two PRs to an existing open-source project Metarank: * support semantic recommendations with cohere-ai and sentence-transformers embeddings * use qdrant as a vector search engine to quickly perform vector similarity search With these two PRs merged building such a recommender is just a matter of a few lines of YAML code. But the semantic-similarity approach is not only about movies, but can be applied more generically in traditional places like e-commerce. For example, in fashion with high inventory churn, being able to recommend something for new clothes having zero feedback is really valuable.

metaprank

The dataset we used is Arxiv’s Open access dataset and we used 15000 records from it which contains ML and AI-oriented papers. We utilized Abstract to create embeddings for each paper. Cohere’s small model was used for creating embeddings which produced 1024 embeddings for each records. Then the 15000 X 1024 Vectors are uploaded to the created collections in qdrant. Finally, the embeddings are generated for the abstract of chosen articles or the given prompt, and the Qdrant searches for similar texts in the collection and outputs the indices of it. The distance metric used to measure the similar vectors is Cosine distance. Besides the recommendation for articles we provided features such as research paper summarization and Translating contents of given paragraphs from English to 8 different languages such as Tamil, Nepali, Indonesia, Thai, Spanish, Russian , Turkish, and French. The model we used for Language Translation is MBART Large-50-one-to-many for multilingual machine translation. The Text Summarization part is done using cohere’s API

UnderDawgs

In today's fast-paced work environment, information overload can make it challenging to find the information you need quickly. Research shows that knowledge workers spend 1 to 3 hours per day looking for information and documents. A big chunk of that time is spent understanding their organization's internal knowledge base. This process can be time-consuming, frustrating, and inefficient. That's why I created mindmate, an AI-powered assistant that helps users make sense of their company's internal knowledge base. With mindmate, users can easily search their company's internal knowledge base and receive answers to their questions in plain English using a simple chatbot interface. I built a proof of concept using GitLab's employee handbook during the hackathon. I created a simple yet powerful tool that allows users to ask questions and receive natural language answers by processing the handbook's 3,000 pages, creating embeddings with Cohere, storing them with Qdrant, and leveraging Cohere's text generation capabilities. mindmate is easy to use and provides quick access to information related to a variety of topics, including company policies, benefits, and more. By tailoring search results to each user's specific needs, mindmate helps knowledge workers save time and stay focused on their core responsibilities.

Dylan