Jenna Kanerva

Post-hoc Manipulations of Vector Space Models with Application to Semantic Role Labeling

In this paper, we introduce several vector space manipulation methods that are applied to trained vector space models in a post-hoc fashion, and present an application of these techniques in semantic role labeling for Finnish and English. Specifically, we show that the vectors can be circularly shifted to encode syntactic information and subsequently averaged to produce representations of predicate senses and arguments. Further, we show that it is possible to effectively learn a linear transformation between the vector representations of predicates and their arguments, within the same vector space.

CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies

The Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data sets. In 2017, the task was devoted to learning dependency parsers for a large number of languages, in a real-world setting without any gold-standard annotation on input. All test sets followed a unified annotation scheme, namely that of Universal Dependencies. In this paper, we define the task and evaluation methodology, describe how the data sets were prepared, report and analyze the main results, and provide a brief categorization of the different approaches of the participating systems.

Morphological Segmentation and OPUS for Finnish-English Machine Translation

This paper describes baseline systems for Finnish-English and English-Finnish machine translation using standard phrasebased and factored models including morphological features. We experiment with compound splitting and morphological segmentation and study the effect of adding noisy out-of-domain data to the parallel and the monolingual training data. Our results stress the importance of training data and demonstrate the effectiveness of morphological pre-processing of Finnish.

SETS: Scalable and Efficient Tree Search in Dependency Graphs

We present a syntactic analysis query toolkit geared specifically towards massive dependency parsebanks and morphologically rich languages. The query language allows arbitrary tree queries, including negated branches, and is suitable for querying analyses with rich morphological annotation. Treebanks of over a million words can be comfortably queried on a low-end netbook, and a parsebank with over 100M words on a single consumer-grade server. We also introduce a web-based interface for interactive querying. All contributions are available under open licenses.

The Finnish Proposition Bank

We present the Finnish PropBank, a resource for semantic role labeling (SRL) of Finnish based on the Turku Dependency Treebank whose syntax is annotated in the well-known Stanford Dependency (SD) scheme. The contribution of this paper consists of the lexicon of the verbs and their arguments present in the treebank, as well as the predicate-argument annotation of all verb occurrences in the treebank text. We demonstrate that the annotation is of high quality, that the SD scheme is highly compatible with PropBank annotation, and further that the additional dependencies present in the Turku Dependency Treebank are clearly beneficial for PropBank annotation. Further, we also use the PropBank to provide a strong baseline for automated Finnish SRL using a machine learning SRL system developed for the SemEval’14 shared task on broad-coverage semantic dependency parsing. The PropBank as well as the SRL system are available under a free license at http://bionlp.utu.fi/.

TurkuNLP: Delexicalized Pre-training of Word Embeddings for Dependency Parsing

We present the TurkuNLP entry in the CoNLL 2017 Shared Task on Multilingual Parsing from Raw Text to Universal Dependencies. The system is based on the UDPipe parser with our focus being in ex- ploring various techniques to pre-train the word embeddings used by the parser in order to improve its performance especially on languages with small training sets. The system ranked 11th among the 33 participants overall, being 8th on the small treebanks, 10th on the large treebanks, 12th on the parallel test sets, and 26th on the sur- prise languages.

Phrase-Based SMT for Finnish with More Data, Better Models and Alternative Alignment and Translation Tools

This paper summarises the contributions of the teams at the University of Helsinki, Uppsala University and the University of Turku to the news translation tasks for translating from and to Finnish. Our models address the problem of treating morphology and data coverage in various ways. We introduce a new efficient tool for word alignment and discuss factorisations, gappy language models and reinflection techniques for generating proper Finnish output. The results demonstrate once again that training data is the most effective way to increase translation performance.

Cross-Lingual Pronoun Prediction with Deep Recurrent Neural Networks.

In this paper we present our winning system in the WMT16 Shared Task on CrossLingual Pronoun Prediction, where the objective is to predict a missing target language pronoun based on the target and source sentences. Our system is a deep recurrent neural network, which reads both the source language and target language context with a softmax layer making the final prediction. Our system achieves the best macro recall on all four language pairs. The margin to the next best system ranges between less than 1pp and almost 12pp depending on the language pair.

Turku: Semantic Dependency Parsing as a Sequence Classification

This paper presents the University of Turku entry to the SemEval-2015 task on BroadCoverage Semantic Dependency Parsing. The system uses an existing transition-based parser as a sequence classifier to jointly predict all arguments of one candidate predicate at a time. Compared to our 2014 entry, the 2015 system gains about 3pp in terms of F-score for a fraction of the development time. Depending on the subtask, the difference between our entry and the winning system ranges between 1 and 5pp.