Skip to content
View temretiras's full-sized avatar
1 follower · 4 following
  • bounling
  • Istanbul

Highlights

  • Pro

Block or report temretiras

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.

Content in all repositories owned by your account will be closed.
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
temretiras/README.md

Tarık Emre Tıraş

Computational linguist in İstanbul. I work on Turkish NLP: building and validating treebanks, evaluating models, and putting LLM agents into production.

  • Researcher at BUCOLIN, Boğaziçi University Institute for Data Science and Artificial Intelligence, since 2023
  • First author of OTA-BOUN (LREC 2026), the largest Universal Dependencies treebank for historical Turkish: 1,742 manually verified sentences, each in Perso-Arabic script and Latin transcription
  • B.A. in Linguistics from Boğaziçi University, with graduate coursework in NLP and statistics
  • Previously built conversational AI and tool-calling workflows for freight operations at TruKKer, and SQL analytics and invoice automation at Samsung Electronics Türkiye

Open to NLP and AI engineering roles.

Publications

  • OTA-BOUN: A Historical Turkish Dependency Treebank. Tarık Emre Tıraş, Nureddin Cüneyd Ünal, Ada Cengiz, Ece Yurtseven, Esma F. Bilgin Taşdemir, Şaziye Betül Özateş. LREC 2026. Paper
  • Dependency Annotation of Ottoman Turkish with Multilingual BERT. Şaziye Betül Özateş, Tarık Emre Tıraş, Efe Eren Genç, Esma F. Bilgin Taşdemir. LAW-XVIII at EACL 2024. Paper · Treebank
  • Building Foundations for Natural Language Processing of Historical Turkish: Resources and Models. Preprint, 2025. arXiv

Projects

Repository What it is
Turkish-Dependency-Parsing BERTurk with a second-order TreeCRF parser for Turkish. The best system reaches 74.73 LAS on UD Turkish-BOUN, 12.77 points above our BERTurk baseline, alongside an ablation study, treebank adapters and morphological feature injection. With İpek Akkuş.
Tokenization-and-Stemming-Approaches-for-Turkish A rule-based stemmer, a rule-based tokenizer with multiword-expression merging (99.73% sentence-level exact match on the Kenet UD treebank), and a character-level logistic regression tokenizer.
A-Constraint-Based-CKY-Parser-Implementation-for-English A feature-augmented CKY parser for English that checks agreement, subcategorization and inversion constraints during chart parsing. With Mehmet Ali Özdemir.
gmail-ai-cikti A bookmarklet and userscript that turn a Gmail thread into clean Markdown for LLMs, keeping tables and links.

Elsewhere

Google Scholar · ACL Anthology · Hugging Face · LinkedIn

Pinned Loading

  1. A-Constraint-Based-CKY-Parser-Implementation-for-English A-Constraint-Based-CKY-Parser-Implementation-for-English Public

    Feature-augmented CKY parser for English: CNF conversion, morphological analysis, and agreement, subcategorization and inversion constraints checked during chart parsing.

    Jupyter Notebook

  2. Turkish-Dependency-Parsing Turkish-Dependency-Parsing Public

    Turkish dependency parsing with BERTurk and a second-order TreeCRF parser: STEPS ablation, Houlsby adapters and morphological feature injection on UD BOUN, IMST and PUD.

    Jupyter Notebook

  3. gmail-ai-cikti gmail-ai-cikti Public

    One-click Gmail thread to Markdown for LLMs (bookmarklet and Tampermonkey userscript). Keeps tables and links, drops the Gmail UI and tracking pixels.

    JavaScript

  4. Tokenization-and-Stemming-Approaches-for-Turkish Tokenization-and-Stemming-Approaches-for-Turkish Public

    UD-compatible Turkish preprocessing: a rule-based stemmer, a rule-based tokenizer with multiword-expression merging, and a character-level logistic regression tokenizer.

    Python