Alexander Panchenko

Invited Talks

• 24th International Conference on Computational Linguistics and Intellectual Technologies (Dialogue'2018): From unsupervised induction of linguistic structures to applications in deep learning. This invited talk was first presented at Skolkovo Institute of Science and Technology and later was also presented at the NLP seminar of Yandex's School of Data Analysis both in Moscow, Russia.
• Global WordNet Conference, Workshop on Wordnets and Word Embeddings (2018):  Inducing Interpretable Word Senses for WSD and Enrichment of Lexical Resources. Singapore, Singapore 
• Université catholique de Louvain, Research Seminar (2015). Text Analysis of Social Networks: Working with FB and VK Data. Louvain-la-Neuve, Belgium. This invited talk was as originally presented at the AI Ukraine 2014 conference. 
• Montclair State University, Research Seminar (2014). Similarity Measures for Semantic Relation Extraction. Montclair, New Jersey, USA.
• Higher School of Economics, Research Seminar (2013). A Graph-based Approach to Extraction of Skills from Text. Nizhny Novgorod, Russia.

• Panchenko A., Faralli S., Ponzetto S. P., and Biemann C. (2017): Using Linked Disambiguated Distributional Networks for Word Sense Disambiguation. In Proceedings of the Workshop on Sense, Concept and Entity Representations and their Applications (SENSE) co-located with the 15th Conference of the European Chapter of the Association for Computational Linguistics (EACL'2017). Valencia, Spain. Association for Computational Linguistics
• Pelevina M., Arefyev N., Biemann C., Panchenko A. (2016): Making Sense of Word Embeddings. In Proceedings of the 1st Workshop on Representation Learning for NLP co-located with the ACL conference. Berlin, Germany. Association for Computational Linguistics
• Ballweg K., Zouhar F., Wilhelmi-Dworski P., von Landesberger T., Fahrer U., Panchenko A., Yimam S.M. Biemann C., Regneri M., Ulrich H. (2016): new/s/leak – A Tool for Visual Exploration of Large Text Document Collections in the Journalistic Domain. Workshop on Visualisation in Practice co-located with the VIS conference, Baltimore, MD, USA
• Panchenko A., Faralli S., Ruppert E., Remus S., Naets H., Fairon C. Ponzetto S. P., and Biemann, C. (2016): TAXI at SemEval-2016 Task 13: a Taxonomy Induction Method based on Lexico-Syntactic Patterns, Substrings and Focused Crawling. In Proceedings of the 10th International Workshop on Semantic Evaluation. San Diego, CA, USA. Association for Computational Linguistics
• Kivimaki I., Panchenko A., Dessy A. Verdegem D., Francq P., Fairon C., Bersini H., Saerens M. (2013): A Graph-Based Approach to Skill Extraction from Text. In Proceedings of TextGraphs-8 Workshop co-located with the Conference on Empirical Methods for Natural Language Processing (EMNLP'2013). Seattle, USA. Association for Computational Linguistics
• Panchenko A. Morozova O. (2012): A Study of Hybrid Similarity Measures for Semantic Relation Extraction. In Proceedings of Workshop of Innovative Hybrid Approaches to the Processing of Textual Data Workshop co-located with the EACL'2012 conference, pp.10-18, Avignon, France. Association for Computational Linguistics
• Panchenko A, Beaufort R., Fairon C. (2012): Detection of Child Sexual Abuse Media on P2P Networks: Normalization and Classification of Associated Filenames. In Proceedings of Workshop on Language Resources for Public Security Applications of the 8th International Conference on Language Resources and Evaluation (LREC). European Language Resources Association (ELRA)
• Panchenko A., Adeykin S., Romanov P., Romanov A. (2012): Extraction of Semantic Relations between Concepts with KNN Algorithms on Wikipedia. In Proceedings of Concept Discovery in Unstructured Data (CDUD) workshop co-located with the International Conference On Formal Concept Analysis (ICFCA'2012), pp.78-88, Leuven, Belgium. CEUR-WS
• Panchenko A. (2012): Towards an Efficient Combination of Similarity Measures for Semantic Relation Extraction. Abstract in Proceedings of the 22nd Meeting of Computational Linguistics in the Netherlands (CLIN 22). Tilburg, The Netherlands. Tilgurg University
• Panchenko A. (2011): Comparison of the Knowledge-, Corpus-, and Web-based Similarity Measures for Semantic Relations Extraction. In Proceedings of the Workshop GEometrical Models of Natural Language Semantics (GEMS) co-located with the EMNLP'2011 conference. pp.10-18. Edinburgh, Scotland. Association for Computational Linguistics
• Panchenko A. (2010): Can We Automatically Reproduce Semantic Relations of an Information Retrieval Thesaurus?. In Proceedings of the Young Scientists Conference of the 4th Russian Summer School in Information Retrieval (RuSSIR'2010). pp.36–51. Voronezh, Russia. VSU

Software

• wsd: A system for unsupervised, knowledge-free, and interpretable word sense disambiguation. Main contributor: Fide Marten. 
• Serelex: Lexico-semantic search engine, a web application for search and visualization of semantically related words. The system provides access to word similarity graphs for English, French, Russian, and Portuguese languages and features a RESTful API. Main contributors: Alexey Romanov and Pavel Romanov.
• PatternSim: A tool for (1) calculation semantic relatedness of words and (2) extraction of hypernyms based on lexical-syntactic patterns also known as Hearst patterns. The tool implements extraction grammars in form of Unitex FSTs for English, French, Russian, and Portuguese. The tool is capable of processing in parallel large amounts of texts ( tested on corpora up to 50Gb). Co-developed with Hubert Naets.
• Chinese Whispers: A fast Java implementation of the Chinese Whispers graph clustering algorithm (Biemann, 2006). The tool is able to perform global or local clustering of the input graph (the latter is also known as ego-network clustering and can be useful for graph-based word sense induction). Main contributor: Johannes Simon.    
• PyMystem3: A morphological analyzer of Russian language for Python based on Yandex MyStem 3.0 (a wrapper around the binary version of the MyStem, which enable a convenient use of the tool from Python). Main contributor: Denis Sukhonin.
• SenseGram: A tool for learning of word sense embeddings on the basis of existing word embeddings. The system also provides word sense disambiguation functionality on the basis of learned word sense embeddings. The approach received the best paper award at the representation learning workshop at the ACL'2016 conference. Main contributor: Maria Pelevina.
• TAXI:  TAXonomy Induction system based on Lexico-Syntactic Patterns, Substrings and Focused Crawling. This system was ranked first at the SemEval 2016 shared task on taxonomy extraction from text. 
• lefex: This project contains Hadoop jobs for extraction of word, feature and word-feature counts from texts. These counts can be helpful for computing distributional thesauri and word graphs. Several types of features are available including syntactic and trigram features. Co-developed with Johannes Simon.
• STC: Short Text Categorizer is a lightweight and fast text categorization library written in C++ based on linear models and vector space model. The library relies on the LibLINEAR library for supervised categorization.
• Lexical-semantic language map of the Russian language. This visualization was obtained by clustering of the graph of semantically related words obtained using the Serelex (see above). The visualization groups semantically similar words, e.g. countries, names, etc.
• russe-evaluation: A collection of scripts and datasets for evaluation of semantic similarity/relatedness measures for the Russian language. This collection of tools contains several datasets that are based on human judgments (HJ), cognitive associations (AE), and semantic relations (synonyms and hypernyms) extracted from thesauri (RT).
• JoSimText: An implementation of the JoBimText approach (Biemann and Riedl, 2013) in Apache Spark. 

Datasets

• RDT: Russian Distributional Thesaurus is a large-scale resource that consists of a distributional thesaurus and word vectors. In order to build the distributional thesaurus, we used the skip-gram model (Mikolov et al., 2013) trained on a 12.9 billion word collection of books in Russian (around 150Gb of plain text). According to the results of our participation in the shared task on Russian semantic similarity (Panchenko et al., 2015), this approach scored in the top submissions showing the state-of-the-art results in word relatedness for Russian. Due to its high coverage, the resource can be used in various applications.  
• RUSSE: four semantic relatedness resources for Russian, each being a list of triples (word_i, word_j, similarity_ij) designed for evaluation of semantic relatedness, each complementing another in terms of relation type. These benchmarks were used in a shared task on Russian semantic similarity. 
• Serelex lexical-semantic networks are distributional thesauri weighted with semantic relatedness scores based on the PatternSim approach. These are the resources used in the Serelex lexical semantic search engine.
  • English: This network was extracted from a concatenation of Wikipedia and ukWaC corpora. 
  • French: The network was extracted from a concatenation of Wikipedia and frWaC corpora. Another version is available extracted from a lower-cased corpus.
  • Russian: a distributional thesaurus weighted with semantic relatedness scores based on the PatternSim approach.  The network was extracted from a concatenation of Wikipedia and ruWaC corpora. Another version is available extracted only from Wikipedia corpus.
• A large-scale corpus of books in Russian: this 12.9 billion tokens corpus was assembled from various books from lib.rus.ec website.  
• Data for sentiment analysis for the Russian language: 
  • Posts from the lovehate.ru web site: This is a complete crawl by 2015 of this website, which is organized in the way so that each topic can receive "love" and "hate" comments i.e. positive or negative opinions of users on the topic. The collection can be useful for training of statistical sentiment classifiers.
  • Dictionary of abusive words: a comprehensive collection/compilation of freely available dictionaries and lists of abusive words. Abusive words are usually denote highly negative emotions and can be helpful for feature extraction. 
  • Sentiment dictionary: this sentiment lexicon contains over 11K of words carefully verified by several annotators. This dataset can be useful for building a dictionary-based sentiment analysis system or for generation of features in a supervised system. 

Supervision of MA Theses

I supervised research-oriented Master theses, usually also aiming to publish a conference paper on the basis of the produced materials.

• Matthias Schildwächter (2019, University of Hamburg): An Open-Domain System for Retrieval and Visualization of Comparative Arguments from Text.
• Alvin Rindra Fazrie (2019, University of Hamburg): Visual Information Management with Compound Graphs. Main supervisor: Steffen Remus. 
• Dahmash Ibrahim (2018, University of Hamburg): Question Answering using Dynamic Neural Networks. Main supervisor: Benjamin Milde.
• Mirco Franzek (2018, University of Hamburg). Comparative Argument Mining. Co-supervised with Chris Biemann. 
• Marten Fide (2017, TU Darmstadt). Predicting hypernyms in contexts with JoBimText. Co-supervised with Chris Biemann. Now at TU Darmstadt.
• Maria Pelevina (2016, TU Darmstadt). Unsupervised Word Sense Disambiguation with Sense Embeddings. Co-supervised with Chris Biemann. Now at Deutsche Bahn R&D.
• Simon Dif (2015, TU Darmstadt). Statistical Models of Semantics with Structured Topics. Co-supervised with Chris Biemann. A dual degree Masters program with ENSIMAG, Grenoble, France. Now at Altran R&D.
• Alexey Romanov (2012, Moscow State Technical University). Graph Algorithms in the Lexical Semantic Search Engine 'Serelex'. Co-supervised with Andrew Philippovich. Now at the University of Massachusets Lowell.

Internships & Visiting Researchers

I help to write research proposals to funding organizations which let researchers visit our faculty and do interesting short-term research project together. 

• Dmitry Ustalov (2016): Graph Clustering for Word Sense Induction. Funded by DAAD
• Artem Chernodub (2017): Recurrent Neural Networks for Argument Mining. Funded by DAAD.
• Andrey Kutuzov (2018): Learning graph embeddings via node similarities. Funded by the University of Oslo. 
• Shantanu Acharya (2018): Taxonomy induction using word sense representations. Funded by DAAD.

Organization of Events

traveling, photography, speleology