Skip to content
View ericyan534-dev's full-sized avatar

Block or report ericyan534-dev

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ericyan534-dev/README.md

Liqian (Eric) Yan

I am a student researcher and builder working across natural language processing, computational economics, and human-centered AI.

My research examines how intelligent systems represent language and how they reshape labor markets. Through LinkScape, I also build technical and educational projects intended to make AI more useful outside the laboratory.

Selected Work

CNM-BERT — A structural Chinese language model that represents characters through their internal ideographic composition, injecting discrete sub-character structure into BERT at the embedding interface. On out-of-vocabulary characters it improves structural probing accuracy by +9.8 points over the strongest baseline, while matching or exceeding comparable Chinese PLMs on CLUE. Co-first author (equal contribution); manuscript submitted to ACL Rolling Review 2027.

AIFE Engine — An NLP and econometric research pipeline studying firm-level AI adoption and labor demand from online job postings. A knowledge-distillation pipeline turns unstructured hiring text into a firm–year panel of AI intensity and task composition, which then feeds causal and forecasting analysis. The repository holds the measurement stage: corpus extraction, entity resolution, teacher scoring, and index construction. Co-authored with Yintong Chen and Zichun Qiu; Regeneron ISEF 2026.

PoliScope — A source-grounded legislative intelligence platform. A LangGraph RAG orchestrator traces bill evolution, sponsorship, lobbying disclosures, and campaign finance across the Congress.gov, GovInfo, Senate LDA, and OpenFEC APIs, grounding every answer in a primary government source. Built at HackHarvard 2025.

AI Digest — A self-hosted research and industry intelligence system: retrieval, clustering, personalization, evaluation, and multi-channel delivery. Runs on a daily schedule against live academic, industry, and community sources, with a deterministic mock-LLM mode so the whole pipeline is reproducible without credentials.

Resonaite — An experimental adaptive-audio platform combining AI-generated music, real-time digital signal processing, and closed-loop EEG feedback. A research prototype, not a medical device.

Current Focus

Natural Language Processing · Econometrics · Causal Inference · RAG Systems · Full-Stack Machine Learning

Elsewhere

Pinned Loading

  1. ai-digest ai-digest Public

    Self-hosted, LLM-powered personal AI-news engine. Ingests academic, industry, and community sources, clusters and ranks them, and delivers daily/weekly digests via web, email, and Telegram.

    Python

  2. resonaite resonaite Public

    Experimental adaptive-audio research prototype combining AI-generated music, real-time DSP, and EEG feedback. Research/wellness prototype - not a medical device.

    JavaScript 1

  3. cnm-bert cnm-bert Public

    Drop-in structural embedding for Chinese characters via Ideographic Description Sequences (IDS). Injects discrete sub-character composition into BERT through a recursive Tree-MLP, at ~5% training o…

    Python

  4. poliscope poliscope Public

    Source-grounded legislative intelligence platform. A LangGraph RAG orchestrator tracing bill evolution, sponsorship, lobbying, and campaign finance across Congress.gov, GovInfo, Senate LDA, and Ope…

    TypeScript 1

  5. aife-engine aife-engine Public

    Firm-level AI adoption measurement from online job postings. Extracts schema.org JobPosting records from Web Data Commons, resolves employer entities, and scores AI intensity and task content.

    Python