Siwei Wang

This project adjusts its recommendations based on user feedback and gives robust, accurate recommendations based on the neural network.

• Design and implement a web page via Python Flask.
• Build a convolutional neural network model with user embedding, movie embedding, and reviewembedding.
• Design a graph neural network to extract representations of movies and users to do feature enrichment.

This project aims to tackle the OffensEval 2019 challenge, and there are three subtasks, including offensive language identification, categorization of offense types, and offense target identification.

• Fine-tune a BERT based classifier on task A and task C achieved an F1 macro score of 0.8155 for task A and 0.6456 for task C both of which rank 2nd in the competition
• Build an RNN model with Glove on taskB, gained an F1 macro score 0.7228(rank 3rd in the competition)
• Solve the problem of data imbalance by threshold adjusting and oversampling.

Reproduce a paper named WAMDA: Weighted Alignment of Sources for Multi-source Domain Adaptation without using pre-existing code. The paper presents a novel method for Multi-source Domain Adaptation which uses multiple sources to train a predictor based on their internal relevance and their relevance score related to the target.

The project is a full-category e-commerce shopping website (B2C). Users can search for products online, add to the shopping cart, place an order, and kill in seconds, while the administrator can manage the promotion activities of products on and off the shelves, and monitor the sales of products, make refunds.

• The back-end mainly uses SpringMVC, SpringBoot, SpringCloud, JWT, Nginx, FastDFS, Redis, MySql, RabbitMQ, Elasticsearch, and the front-end uses basic html, css, Javascript and NPM, Vue.js, etc.
• UsingRabbitMQtorealizereliablemessageservice,andsolvetheproblemofinter-servicecommunicationanddistributedtrans- action problem.
• Implementing offline client shopping cart based on LocalStorage to reduce serverpressure.
• Realizing the intelligent filtering and search of commodities by elasticsearch.
• Realizing the request load balancing and request current limit by nginx.

It is a social platform for all practitioners in the financial industry. It has eight main channels: News, QA, Event Organization, Dating Platform, Recruitment Information, User Center, Search Center and Third-Party Login

• It mainly involves SpringBoot, SpringCloud, MyBatis-plus, Redis, MongoDB, ElasticSearch, Ribbon, Hystric, zuul, JWT, etc.
• Using RabbitMQ to store new message reminders,reduce data base access pressure
• In order to solve the storage and query of massive data of first, second and third level comments in Tucao Tiandi, MongoDB is used for storage, which improves storage efficiency and query performance.
• Uses Redis to cache hotspot query data, which improves system query performance

This project adjusts its recommendations based on user feedback and gives robust, accurate recommendations based on the neural network.

• Design and implement a web page via Python Flask.
• Build a convolutional neural network model with user embedding, movie embedding, and reviewembedding.
• Design a graph neural network to extract representations of movies and users to do feature enrichment.