10128485Unlocking the power of multi-lingual unstructured data for business critical decision makingClosedGrant for R&DInnovate UKThe project addresses the pressing challenge of enabling companies to unlock the power of their own data -- irrespective of the language it is stored in -- to make uniquely informed business decisions, faster and cheaper. As much as 80% of the critical data a company holds is unstructured, including contracts, invoices, NDAs and agreements. This valuable asset is critical to analytics, decision making and due diligence. However, making sense of the vast quantities of information a company holds is an almost impossible, incredibly time/cost consuming task, with a human typically spending ~30 minutes reviewing a single legal document, and potentially needing to review 100s/1000s. TextMine has created a patented, sector-leading technology using large language models (LLMs) and ML to analyse unstructured textual data, automatically retrieving key terms and providing insights to help users make informed decisions. It is helping clients find accurate answers to queries up to 100x faster, saving 40 hours/month of each employee's data entry and information retrieval tasks. For companies with global activities, many documents are written in other languages, and cannot easily be processed by human users or by LLM-based technologies such as TextMine (which typically focus on English). This means (i) their business critical decisions are being driven by only a fraction -- and an unrepresentative sample -- of the valuable information they own, and (ii) risks in key documents might be missed. As such, global companies are at risk of £millions of avoidable losses or missed business potential. This language challenge presents an enormous business opportunity for TextMine, with this project enabling optimisation of our LLM-based technologies and translation models to perform data extraction on other languages. We have significant demand, including banks and large enterprises across innovation areas for a solution to enable them to automate insights from their non-English data assets. We initially target Mandarin, the most-spoken business language after English, with this project following a double diamond design process with co-design with key TextMine customers who conduct significant operations in Mandarin. The project is led by Amber Akhtar, who will lead a mentoring programme, including outreach activities with disadvantaged young people and girls in schools in West London to inspire them about careers in STEM and finance and to dispel myths that these careers are not for them; and an internship programme for women from underrepresented groups to ensure they have all the opportunities to succeed.