Slides
Some of my Slides!
2026-05 self-wiki: Personal AI-powered wiki + Socratic Mirror
- self-wiki is a personal AI-powered wiki + Socratic Mirror service
- treat AI as the "OS" and the generated wiki as the "Context Window" for high-resolution self-reflection
- self-wiki project can be on coach or education
2025-09 Mental Health for Every Mind
- Mental Health for Every Mind is an AI-powered cognitive behavioral therapist (CBT)
2024-10 A tiny Scheme interpreter with Python
- lis.py shows the interpreter details
- program - parse -> abstract-syntax-tree - eval -> result
lis.py> (begin (define r 10) ( pi ( r r)))
314.1592653589793
lis.py>
- program - parse -> abstract-syntax-tree - eval -> result
2023-04 Factoid Question Answering (QA)
- Question Answering (QA) tasks return an answer given a question. The slide summarizes the recent progress on this topic .
2022-06 An introduction and the related experiments on Natural Legal Language Processing
- Check out the introduction and the related experiments on NLP in legal domain
- The unstructured legal texts are not only commercial ones, but also judgments, scholarly articles, arguments and pleadings
2021-08 My slide on Japan in views of culture at HKU SPACE
- An introduction on Japan in views of culture
2017-2020 Some of my slides on Artificial Intelligence and its applications in Language & Speech at HSUHK
- Our paper "Mapping Customer Needs in Natural Language to Product Specifications by Using Word Embeddings" has been accepted for publication at apiems2018, HK.
- How to define the objectives and collect or generate your dataset
- Algorithm: word embeddings based mapping
- Algorithm: Support Vector Machine on word-embedding
- Algorithm: Convolutional Neural Network for Sentence Classification
- The experiments based on Convolutional Neural Network
- Algorithm: Recurrent Neural Network
- Algorithm: Long short-term memory Network
- Algorithm: Hier-attention Network
- Top-N Sort Algorithm
- TOP-N Algorithm on word embeddings
- iSpeaking - Automatic Speech Recognition for L2 (english&chinese) learners
2017-05 Text simplification
- The system can simplify English text with syntactic and lexical complexity while keeping the original meaning intact
- Our paper "A Customizable Editor for Text Simplification" on this has been accepted for publication at COLING 2016 on Computational Linguistics, Osaka,Japan.
2015-11 Privacy-preserving Similar Search
Nearest Neighbour search refers to finding objects (i.e. face in biometric) that have similar characteristics to the query object .
2014-10 Online learning
The Online learning summary presents an overview about the latest progress in terms of online student behavior and skills, online assessment, course materials and online forums, based on NIPS Workshop on Data Driven Education,2013.
Online learning is bringing about dramatic changes in higher education, and a number of successful largescale Massive Open Online Courses (MOOCs) platforms such as edX, Coursera, and Udacity have been developed. This revolution raises a series of challenges and possibilities in technology and education fields.
2013-05 Peer-to-Peer Session Initiation Protocol (P2PSIP)
The P2PSIP presents a P2PSIP system and Distributed Hash tables (DHTs) algorithms, which are used to organize the P2PSIP overlay.
P2PSIP is a new technology currently being standardized in the P2PSIP working group of the IETF. A P2PSIP network is consisted of a collection of nodes organized in a peer-to-peer fashion for the purpose of enabling real-time communication using the Session Initiation Protocol (SIP).
2013-04 An Instruction about incremental 3D reconstruction using Bayesian learning
The essay mainly presents an algorithm for 3D reconstruction.
The method first reconstructs an initial 3D model by selecting uniformly distributed key images using a view sphere. Then when a new image is added, its correlated reconstructed patches are searched and incrementally update the result model by optimizing the geometric and photometric terms. This can be applied for large-scale datasets or to real-time reconstruction.
2013-04 Finding Better Experts
This paper focuses on the problem of finding better experts in micro-blogging services (like Sina weibo/twitter).
Now micro-blogging has been one of the most popular platforms where users can describe their current status in short posts. Micro-blogging uses a model called “following”, in which users can choose who he/she wants to follow without the latter to give the permission first.
This paper is based on the existence of homophily to measure the influence of users in micro-blogging. The algorithm TwitterRank is used because it takes the topical similarity between users and the link structure into account.