Summary
My co-founders and I started Ludwig with a mission: to help everyone write better English and be more confident about it. To achieve the ambitious goal of making English writing more democratic, we have been working with the most cutting-edge Artificial Intelligence technologies in the field of Natural Language Processing (NLP) for the last 8 years.
in 2021 we conducted a market survey with ~1000 users and discovered that 60% of them needed a paraphraser. Based on this, we decided to design and develop one. A paraphraser is a software that, given an input sentence, can generate several alternative phrases that maximize the –terminological, grammatical, and syntactical– transformation of the original sentence, while minimizing the semantic difference.
The problem was that to have a competitive paraphraser model based on neural networks you have to train models made of billions of parameters, which would cost millions of dollars.
Based on new cutting-edge technology we adapted a method named Prompt Tuning to our case study and built a process able to train billion parameter models for hundreds of € instead of millions. This opens up a whole new world of technological and market opportunities. We will be able to train more models and hence develop more tools quickly and with a 1/1000th of the cost of our competitors while having the same quality.
The goal of this project is to help us bring our TRL7- Paraphraser to the market (TRL9), with a compelling product market fit. Most of the tasks are business-related and being as Chief of Business Development, this project is the perfect match for my skills.
We will complete the project in 9 months during which we will establish KPIs, analyze UI interactions, design and launch marketing experiments, to make product iterations and modifications based on insights. We will complete the deployment of our paraphraser in all the touchpoints and release it to the public.
in 2021 we conducted a market survey with ~1000 users and discovered that 60% of them needed a paraphraser. Based on this, we decided to design and develop one. A paraphraser is a software that, given an input sentence, can generate several alternative phrases that maximize the –terminological, grammatical, and syntactical– transformation of the original sentence, while minimizing the semantic difference.
The problem was that to have a competitive paraphraser model based on neural networks you have to train models made of billions of parameters, which would cost millions of dollars.
Based on new cutting-edge technology we adapted a method named Prompt Tuning to our case study and built a process able to train billion parameter models for hundreds of € instead of millions. This opens up a whole new world of technological and market opportunities. We will be able to train more models and hence develop more tools quickly and with a 1/1000th of the cost of our competitors while having the same quality.
The goal of this project is to help us bring our TRL7- Paraphraser to the market (TRL9), with a compelling product market fit. Most of the tasks are business-related and being as Chief of Business Development, this project is the perfect match for my skills.
We will complete the project in 9 months during which we will establish KPIs, analyze UI interactions, design and launch marketing experiments, to make product iterations and modifications based on insights. We will complete the deployment of our paraphraser in all the touchpoints and release it to the public.
Unfold all
/
Fold all
More information & hyperlinks
| Web resources: | https://cordis.europa.eu/project/id/101113982 |
| Start date: | 01-07-2023 |
| End date: | 31-03-2024 |
| Total budget - Public funding: | - 75 000,00 Euro |
Cordis data
Original description
My co-founders and I started Ludwig with a mission: to help everyone write better English and be more confident about it. To achieve the ambitious goal of making English writing more democratic, we have been working with the most cutting-edge Artificial Intelligence technologies in the field of Natural Language Processing (NLP) for the last 8 years.in 2021 we conducted a market survey with ~1000 users and discovered that 60% of them needed a paraphraser. Based on this, we decided to design and develop one. A paraphraser is a software that, given an input sentence, can generate several alternative phrases that maximize the –terminological, grammatical, and syntactical– transformation of the original sentence, while minimizing the semantic difference.
The problem was that to have a competitive paraphraser model based on neural networks you have to train models made of billions of parameters, which would cost millions of dollars.
Based on new cutting-edge technology we adapted a method named Prompt Tuning to our case study and built a process able to train billion parameter models for hundreds of € instead of millions. This opens up a whole new world of technological and market opportunities. We will be able to train more models and hence develop more tools quickly and with a 1/1000th of the cost of our competitors while having the same quality.
The goal of this project is to help us bring our TRL7- Paraphraser to the market (TRL9), with a compelling product market fit. Most of the tasks are business-related and being as Chief of Business Development, this project is the perfect match for my skills.
We will complete the project in 9 months during which we will establish KPIs, analyze UI interactions, design and launch marketing experiments, to make product iterations and modifications based on insights. We will complete the deployment of our paraphraser in all the touchpoints and release it to the public.
Status
SIGNEDCall topic
HORIZON-EIE-2022-SCALEUP-02-02Update Date
31-07-2023
Images
No images available.
Geographical location(s)
Search
