Gitfolio

---

Small Software

• WordNet
  The WordNet class is to represent the WordNet synsets and the hyperonymy relation between them, and to support a variety of queries for extracting the information encoded in the hierarchical structure of WordNet.

---

• Spell Cheking and Keyword Searching
  A spellchecker implemented trie-based or non-trie-based. And a keyword search with a command line interface.

---

• Twitter Crawler
  A twitter crawler through Twitter’s API(as much as Twitter’s API is problematic), download tweets and print some statistics, a Tweet class to represent tweet is included.

---

• Finite State Automata
  An FSA(with minimization method) with a trie-based lexicon.

---

• NNPuzzle
  A rule-based AI, automatically solves NNpuzzle.

---

• Parsing
  A Constituency to Dependency Parsing Converter.

---

• Word Searching online with a GUI

---

• Solving small problems
  Implementations of small problems such as calculating polynomials, randomized in-place quick sort and string sorting with radix sort…

---

• Finite State Transducer

---

Data Analysis

• Bayesian Regressian SoundInventoryPopulation
  A series of Bayesian regression analyses to test whether an association between two variables(Sound inventory size vs. population size) is genuine, with the rethinking package.

• Survival Analysis ld.chin
  A survival analysis of lexical decision data of Mandarin Chinese with R.

• Survival Analysis blp
  A survival analysis(with R) of time-to-event data: the data for the British Lexicon Project.

---

Some little practice with Statistical NLP

• Tweets Clustering
  Dimensionality reduction with PCA. Clutering Tweets with k-means.
  

• Segmentation as Sequence labeling
  Segmentation as sequence labeling with Gated RNN.
  

• Pos Tagging as Sequence labeling
  Pos tagging as sequence labeling with Gated RNN.
  

• NER on Reddit posts/comments with SpaCy
  NER experiments: recognize DRUG entities in reddit drug posts/comments.
  

• Speaker Verification with NeMo
  A speaker verification experiment with NeMo.
  

• NLP without readymade annotated dataset: experiments with SpaCy and Snorkel
  NER with SpaCy, information extraction with Snorkel.
  

• English and German NER models trained on CoNLL-2003 data with spacy v3
  I trained a series of (language-dependent) spaCy v3.0 (English and German) NER models with different configurations in order to achieve the best possible f-score. Among them, the best English NER model (benchmark model) had F-score 89.22, the best German NER model had F-score 83.29, both evaluated on the respective testb data.
  

…

---

Theses

• BA Thesis: Practical Speaker Identification on Robotic Systems
  
• MA Thesis: Multimodal Learning Paradigm: Physiological Features (time series) + Clinical Notes

---

Tutorials

• Mathematical Methods 1: Logic
• Mathematical Methods 2: Statistics
• Speaker Identification with Nemo

---

---

Page template forked from evanca