Successful cases
BlockLM, a smart notebook prototype:
BlockLM is a smart notebook prototype developed for working in both Basque and Spanish. It was developed as a demo for demonstrating the validity of the models created within the framework of ILENIA, so that users can select PDF documents, academic texts, notes, HTML pages, audio files and other materials, organise them into notebooks or albums and generate enquiries, summaries and other related content. The tool offers functionalities such as questions and answers, summaries, FAQs, mind maps and summaries of audio in podcast format, via a web interface.
BlockLM, a smart notebook prototype:
Overview
BlockLM is a smart notebook prototype developed for working in both Basque and Spanish. It was developed as a demo for demonstrating the validity of the models created within the framework of ILENIA, so that users can select PDF documents, academic texts, notes, HTML pages, audio files and other materials, organise them into notebooks or albums and generate enquiries, summaries and other related content. The tool offers functionalities such as questions and answers, summaries, FAQs, mind maps and summaries of audio in podcast format, via a web interface.
Challenge
The main challenge was constructing an intelligent system capable of handling the users’ personal documents and audio files in a standardised manner, in order to provide high quality results in both Basque and Spanish. For this, we needed to integrate, in a single platform, document management, content extraction, semantic search, generation of answers using a LLM, voice transcription and text-to-voice conversion.
Collaboration
It was developed in collaboration with the EHU.
Result
The result of the project is the functional prototype BlokLM. Users can upload and organise their content, ask questions about it, citing specific sources, generate personalised summaries, automatically create frequently asked questions and mind maps and convert the content into podcast-style audio summaries. It is based on a modular architecture that distinguishes between frontend and backend and integrates LLM, ASR and TTS. The prototype code can be found at GitHub: https://github.com/orai-nlp/bloklm