Human and Machine Judgements for Russian Semantic Relatedness

Panchenko, Alexander; Ustalov, Dmitry; Arefyev, Nikolay; Paperno, Denis; Konstantinova, Natalia; Loukachevitch, Natalia; Biemann, Chris

doi:10.1007/978-3-319-52920-2_21

Cited by 40 publications

(24 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To address this concern, the final two conditions presented primetarget pairs that bore only a semantic or phonological relation, respectively. Semantic relations between Russian prime-target pairs were established based on the results of a cognitive association experiment (Panchenko et al, 2016; accessed online 1/1/2017).…”

Section: Methodmentioning

confidence: 99%

Suffix interference in Russian

Comstock

2018

Proc Ling Soc Amer

View full text Add to dashboard Cite

Abstract. The phenomenon of "suffix interference" has been used as evidence for a distinction between inflectional and derivational processes (e.g. Pinker & Prince, 1988;Pinker, 1999;Pinker & Ullman, 2002). Yet much of the work on affix priming exists in English, a morphologically poor language, and suffix interference appears inconsistently in cross-linguistic data. The greater reliance on morphological complexity in Russian, and its use of an infinitival suffix and aspectual affixes that may bridge the distinction between traditional definitions of inflectional and derivational word forms, call into question how generalizable the original findings on suffix interference may be for morphologically-complex languages. Investigating these questions, this paper provides unexpected findings: suffix interference is absent in Russian, inflectional suffixes reveal significantly more robust priming effects, and the infinitival suffix is best considered a special case of affix priming, failing to pattern with either inflectional or derivational suffixes. Thus, Russian appears to defy the assumption that inflections are "stripped" during morphological parsing; instead, verbal inflections prove the greatest facilitators of morphological priming. A linear mixed effects model indicates these effects cannot be explained by frequency alone.

show abstract

Section: Methodmentioning

confidence: 99%

Suffix interference in Russian

Comstock

2018

Proc Ling Soc Amer

View full text Add to dashboard Cite

show abstract

“…We produce a collection of pseudo-documents using the window of size 5 and subsampling. For evaluation we use HJ testset [34] with human judgments on 398 word pairs translated to Russian from the widely used English testsets: MC [28], RG [37], and WordSim353 [10]. We also use SimLex-999 testset translation [18].…”

Section: Document Similarity Taskmentioning

confidence: 99%

Interpretable Probabilistic Embeddings: Bridging the Gap Between Topic Models and Neural Networks

Potapenko¹,

Popov²,

Vorontsov³

2017

Communications in Computer and Information Science

View full text Add to dashboard Cite

We consider probabilistic topic models and more recent word embedding techniques from a perspective of learning hidden semantic representations. Inspired by a striking similarity of the two approaches, we merge them and learn probabilistic embeddings with online EM-algorithm on word co-occurrence data. The resulting embeddings perform on par with Skip-Gram Negative Sampling (SGNS) on word similarity tasks and benefit in the interpretability of the components. Next, we learn probabilistic document embeddings that outperform para-graph2vec on a document similarity task and require less memory and time for training. Finally, we employ multimodal Additive Regularization of Topic Models (ARTM) to obtain a high sparsity and learn embeddings for other modalities, such as timestamps and categories. We observe further improvement of word similarity performance and meaningful inter-modality similarities.

show abstract

“…However, in this paper, we limit ourselves to measuring correlation of semantic similarity scores with human judgments, which is also the most established one. For more information on the semantic similarity and relatedness task, especially in Russian context, we refer the reader to [4].…”

Section: Related Workmentioning

confidence: 99%

Size vs. Structure in Training Corpora for Word Embedding Models: Araneum Russicum Maximum and Russian National Corpus

Kutuzov

Kunilovskaya

2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

In this paper, we present a distributional word embedding model trained on one of the largest available Russian corpora: Araneum Russicum Maximum (over 10 billion words crawled from the web). We compare this model to the model trained on the Russian National Corpus (RNC). The two corpora are much different in their size and compilation procedures. We test these differences by evaluating the trained models against the Russian part of the Multilingual SimLex999 semantic similarity dataset. We detect and describe numerous issues in this dataset and publish a new corrected version. Aside from the already known fact that the RNC is generally a better training corpus than web corpora, we enumerate and explain fine differences in how the models process semantic similarity task, what parts of the evaluation set are difficult for particular models and why. Additionally, the learning curves for both models are described, showing that the RNC is generally more robust as training material for this task.

show abstract

Human and Machine Judgements for Russian Semantic Relatedness

Cited by 40 publications

References 20 publications

Suffix interference in Russian

Suffix interference in Russian

Interpretable Probabilistic Embeddings: Bridging the Gap Between Topic Models and Neural Networks

Size vs. Structure in Training Corpora for Word Embedding Models: Araneum Russicum Maximum and Russian National Corpus

Contact Info

Product

Resources

About