* Responsibilities: * Perform information extraction, text classification and knowledge graph for bidding documents. * Solve business problems with SOTA NLP frameworks such as Hugging Face, PaddleNLP, StanfordNLP and related libraries.Project:Named Entity Recognition for Bidding Documents 09.2019– 03.2020 * Situation: It is time-wasting to get key information from bidding documents manually, which requires AI solutions. * Task: Develop information extraction algorithms to get the location, company name, time and amount from documents. * Action: Research named entity recognition solutions; design and implement Bert+LSTM; deliver results to business team. * Result: Business department is satisfied with the Jaccard index which is 0.91. Show less

* Responsibilities: * Perform text preprocessing and text classification for work orders. * Develop user portraits and push potential business opportunities based on customer chat history.Project:Knowledge Graph for Unstructured Text from Customer Chat History 09.2017-03.2018 * Situation: There are millions of chat records from customers every day, which will be a huge waste if not utilized. * Task: Implement SOTA NLP techniques to analyze unstructured texts and identify potential business opportunities. * Action: Build pipelines for text processing, entity-relation extraction and knowledge-graph building. * Result: Identify the potential commodity demands of customers, which is helpful for the product promotion. Show less