Dublin City University at the TweetMT 2015 Shared Task\footnotepubrights The real publication is in TweetMT 2015. in SEPLN 2015. The version you see here does not include the Spanish translations.

Antonio Toral    Xiaofeng Wu    Tommi Pirinen   
Zhengwei Qiu
   Ergun Bicici    Jinhua Du
ADAPT Centre
   School of Computing    Dublin City University    Ireland
{atoral
   xwu    tpirinen    zhengwei.qiu2    ebicici    jdu}@computing.dcu.ie
Abstract

We describe our participation in TweetMT for three language pairs in both directions: Spanish from/to Catalan, Basque and Portuguese. We used a range of techniques: statistical and rule-based MT, morph segmentation, data selection with ParFDA and system combination. As for resources, our focus was on crawling vast amounts of tweets to perform monolingual domain adaptation. Our system was the best of all systems submitted for five out of the six language directions.

1 Introduction and Objectives

While statistical machine translation (SMT) can be considered a mature technology nowadays, one of its requirements is the availability of considerable amounts of parallel text for the language pair of interest. Ideally, the parallel text to train an SMT system should come from the same domain and genre as the text the system is going to be applied to. Thus, using MT to translate types of text for which no parallel data is available constitutes a challenge. This is the case for tweets and social media in general, the target text of the TweetMT shared task.

The main objective of our participation in the TweetMT 2015 shared task was to build the best MT systems for tweets we could with a clear constraint, i.e. it had to be done in a very short period and, to a large extent, be limited to available resources. We have taken part for three language pairs in both directions: Spanish (ES) from/to Catalan (CA), Basque (EU) and Portuguese (PT).

We decided to focus on making the best possible use of available techniques, tools and resources. Regarding techniques and tools, we rely on state-of-the-art SMT, morph segmentation for morphologically rich languages (EU), data selection with ParFDA for fast development of accurate SMT systems [1] and domain adaptation [3], the use of available open-source rule-based systems and, finally, system combination to take advantage of the strengths of the different systems we built. As for resources, we crawl vast amounts of tweets to perform monolingual domain adaptation and complement this with publicly available general-domain monolingual and parallel corpora.

The rest of the paper is organised as follows. Sections 2 and 3 detail the systems built and the resources used, respectively. Section 4 presents the evaluation and, finally, Section 5 outlines conclusions and lines of future work.

2 Architecture and Components of the System

Here we describe the components used in our translation pipeline. First, we pre-process the datasets (Section 2.1), then we use a set of MT systems (Section 2.2) that can incorporate additional functionality (Sections 2.3 and 2.4). Finally, we combine MT systems (Section 2.5).

2.1 Data Preprocessing

Prior to be used, all the datasets used in our systems are preprocessed, as follows:

  1. 1.

    Punctuation normalisation, with Moses’ [9] script.

  2. 2.

    Sentence splitting and tokenisation, with Freeling [14].

  3. 3.

    Normalisation (only for tweets). We sort the vocabulary of a tweet corpus by word frequency and inspect the words that occur in at least 0.5% of the tweets, creating rules to convert informal words to their formal equivalent. This leads to just a handful of rules. E.g. in Spanish, “q”, occurring in 2.62% of the tweets, is converted to its formal equivalent “que”.

  4. 4.

    Truecasing, with a modified version of Moses’ script. We added a set of start-of-sentence characters commonly used in Spanish: ”-”, ”—”, ”¿”, ”“” and ”‘”.

2.2 MT Systems

We build SMT systems using two paradigms: phrase-based with Moses [9] and hierarchical with cdec [5]. In both cases we use default settings. We also use off-the-shelf open-source rule-based MT (RBMT) systems. Namely, Apertium [6] for ESCA, ESPT and EUES,11Revisions 60356, 60384, and 60356, respectively. and Matxin [12] for ESEU.22API at http://ixa2.si.ehu.es/glabaka/Matxin.xml

The SMT systems use 5-gram LMs with Knesser-Ney smoothing [8] except for ParFDA Moses SMT systems, which use LMs of order 8 to 10. We build LMs on individual monolingual corpora (cf. Section 3.2) and interpolate them with SRILM [18] to minimise the perplexity on the dev set. Each target language and its corpora used to build LMs together with their interpolation weights are shown in Table 4. We observe that tweets are given very high weights even if they are not the biggest corpora in the mixes.

2.3 Morphological Segmentation

Morphological segmentation is a popular method to deal with SMT for morphologically differing languages by simply splitting words into sub-word units. The main benefits of morphological segmentation are to reduce the out-of-vocabulary (OOV) rate and to increase the percentage of 1 to 1 word alignments between morphosyntactically different languages; e.g. in our case, by matching inflectional suffixes in EU to syntactic prepositions in ES, we expect to improve the MT quality for the EU–ES language pair. The segmentation and de-segmentation is able to create word-forms not present in the training data by matching a translated stem with a correct suffix.

In our participation, morphological segmentation was only used for EU–ES on the EU side, since EU’s morphology is significantly more complex than that of ES. For the remaining languages of the shared task, there is no such big difference in morphology complexity (all of them are closely-related as they belong to the same family) so the expected gains do not outweigh the added complexity of segmentation.

We use unsupervised statistical segmentation as provided by Morfessor 2.0 Baseline [20].33http://www.cis.hut.fi/projects/morpho/morfessor2.shtml The basic setup for segmentation is the same as in the Abu-MaTran project submission to the WMT 2015 translation task [16]. However, some minor Twitter-related pre-processing has been added in order to keep URLs and hashtags intact. The parameters used for Morfessor training are the default of version 2.0.2-alpha and the data for training is the EU side of the ES–EU parallel training data (cf. Section 3.1).

To gauge the effects of our method as well as the morphological complexity of EU as compared to ES we show in Table 1 the OOV rates and vocabulary sizes of the ES and EU sides of the ES–EU training corpus, and EU corpora after morphological segmentation. Segmentation reduces the type-to-token ratio by a factor of 6 and the OOV rate by almost a factor of 10.

  Corpora Tokens Types OOV
  ES 30,532,489 296,612 14.5 %
EU 24,966,862 605,207 25.4 %
EU morphs 35,293,220 100,990 2.6 %
Table 1: Size of ES–EU training corpus in word tokens (ES and EU sides) and in morph tokens (EU).
5-gram OOV perplexity ST trainC trainFDA LMFDA %red trainC trainFDA LMFDA %red CA–ES 2948 2957 2324 .21 332 336 294 .11 EU–ES 3021 3046 2443 .19 462 483 546 -.18 PT–ES 2871 2896 1951 .32 633 623 486 .23 ES–CA 3338 3345 2890 .13 325 330 338 -.04 ES–EU 4110 4129 3349 .19 745 761 63744ES–EU LM is recomputed after the task, removing duplicates, which slightly decrease BLEU, increase NIST. .15\footrefes-eu_LM ES–PT 3087 3117 2216 .28 993 941 746 .25
Table 2: LM comparison built from training corpus (C train), ParFDA selected training data (FDA train), ParFDA selected LM data (FDA LM). %red is reduction proportion.

2.4 ParFDA

ParFDA parallelizes instance selection with an optimized parallel implementation of FDA5 and significantly reduces the time to deploy accurate SMT systems especially in the presence of large training data and still achieve state-of-the-art SMT performance [1, 2]. Detailed composition of the available corpora, which is referred to as constrained (C), are provided in \Crefsec:resources. For ES, we also included LDC Gigaword corpora [13]. The size of the LM corpora includes both the LDC and the monolingual LM corpora provided. ParFDA selected training and LM data obtains accurate translation outputs with the selected LM data reducing the number of OOV tokens by up to 32% and the perplexity by up to 25% and allows us to model higher order dependencies (\CrefLMPerplexityComparison).

2.5 System Combination

For each language direction we have built up to five systems, as detailed in Sections 2.2 to 2.4: (i) phrase-based and (ii) hierarchical SMT, (iii) phrase-based with morph segmentation, (iv) phrase-based with ParFDA and (v) RBMT. We hypothesise these systems to have complementary strengths, and thus we decide to perform system combination. To that end we use MEMT [7], with default settings, except for the parameter length, for which we use its default (7) for all directions except for ESEU, for which we use 5 according to empirical results on the development set.

3 Resources Employed

3.1 Parallel Corpora

Ideally, we would use data in the same domain and genre as the test set, i.e. tweets. We have access to parallel tweets provided by the task for ES–CA and ES–EU (4,000 parallel tweets for each language pair, we use 1,000 for dev and the remaining 3,000 for training). For ES–PT we have access to 999 parallel tweets (we use them for dev) from Brazilator,55http://www.cngl.ie/brazilator a recent project by DCU and Microsoft to translate tweets from the 2014 soccer World Cup across 24 language directions.

As the availability of parallel tweets for the language pairs of TweetMT 2015 is rather limited (at most we have 4,000 per language pair), we use additional sources of parallel data. For ES–CA we use elPeriodico (eP)66http://catalog.elra.info/product_info.php?products_id=1122 and a selection of contemporary novels. For ES–EU, translation memories (TMs) provided by the shared task77http://komunitatea.elhuyar.org/tweetmt/resources/ and two corpora from Opus [19]:88http://opus.lingfil.uu.se/ Open subtitles 2013 and Tatoeba. Finally, for ES–PT we use Europarl v799http://www.statmt.org/europarl/ and two corpora from Opus: news-commentary and Tatoeba. Table 3 provides details on these corpora.

  Pair Corpus # s. # tokens
ES–CA tweets 3K 48k, 48k
eP 0.6M 13.5M, 14M
novels 47K .78M, .86M
ES–EU tweets 3K 42K, 38K
TMs 1.1M 28.9M, 23.5M
OpenSubs 0.16M 1.2M, 1.0M
Tatoeba 902 6.7K, 5.5K
ES–PT EU 1.9M 54M, 53M
NC 9K .26M, .25M
Tatoeba 53K .42M, .41M
Table 3: Parallel corpora used for training. For each corpus we provide its number of sentence pairs (# s.) and tokens on both sides (# tokens).

3.2 Monolingual Corpora

Our main source of monolingual data is in-domain and comes from crawled tweets. We use TweetCat [11] and crawl tweets for all the target languages (CA, ES, EU and PT) during March and April 2015.

For each language we create two lists of words as required by the crawler: (i) most common discriminating words (up to 100), these are words that are unique to the language and they are used to seed the crawler so that it can find candidate tweets; and (ii) most common words of the language (200), these are used to determine the language of crawled tweets. These two lists are derived from a list of the most common words found in a corpus of subtitles.1010https://onedrive.live.com/?cid=3732e80b128d016f&id=3732E80B128D016F!3584

The tweets crawled are post-processed with langid1111https://github.com/saffsd/langid.py to identify their language. We keep the tweets whose langid’s confidence score is above a certain threshold, which is set empirically at 0.7 by inspecting tweets.

In addition to crawled tweets, we use the target sides of the parallel corpora (cf. Section 3.1 and a set of monolingual corpora as follows. For CA we use caWaC [10], a corpus crawled from the .cat top level domain. For ES, news crawl and news-commentary from WMT’13.1212http://www.statmt.org/wmt13/ For EU, a dump from Wikipedia (20150407). For PT, the news sources CETEMPublico,1313http://www.linguateca.pt/cetempublico/ and CETENFolha,1414http://www.linguateca.pt/cetenfolha/ and a dump from Wikipedia (20150510).

Table 4 shows details on these corpora including their interpolation weights (cf. Section 2.2).

  Lang Corpus # tokens Weights
CA tweets 29M 0.60
caWaC 0.5G 0.33
eP 14M 0.07
ES tweets 129.2M 0.75
news 0.4G 0.21
europarl 60M 0.04
EU tweets 11.3M 0.97
Wikipedia 11.5M 0.01
TMs 23M 0.02
PT tweets 33M 0.93
Wikipedia 166M 0.02
Others 286M 0.05
 
Table 4: Monolingual corpora used for training. For each corpus we show its number of tokens (# tokens) and its weight in LM interpolation.

4 Evaluation

We report our results on the development set (all systems built) and then on the test set (systems submitted).

4.1 Evaluation on Development Data

Table 5 presents the results obtained on the devset by the individual systems and a set of combinations for the three language pairs we covered: ES–CA, ES–EU and ES–PT. The scores were obtained on raw MT output (i.e. tokenised and truecased) as calculated by us with BLEU [15] (multibleu cased as included in Moses version 3) and TER [17] (as implemented in TERp version 0.1). Due to time constraints not all the possible combinations were tried. The scores of the best individual system and combination are shown in bold.

  System BLEU TER
ESCA Moses (1) 82.21 0.1102
cdec (2) 81.45 0.1128
ParFDA (3) 82.37 0.1062
Apertium (4) 78.17 0.1310
1+2 81.71 0.1102
1+4 82.37 0.1057
1+2+4 81.93 0.1085
CAES Moses (1) 82.52 0.1086
cdec (2) 81.76 0.1118
ParFDA (3) 82.16 0.1063
Apertium (4) 77.96 0.1329
1+2 82.38 0.1088
1+4 82.58 0.1077
1+2+4 82.38 0.1083
1+3+4 82.45 0.1074
ESEU Moses (1) 22.57 0.6116
cdec (2) 23.7 0.5863
ParFDA (3) 21.59 0.6181
Matxin (4) 12.66 0.7436
Morph (5) 5.20 0.8812
1+2 23.18 0.5796
1+4 18.36 0.6112
1+2+4 23.58 0.5771
1+2+4+5 24.07 0.5741
1+2+3+4+5 24.42 0.5777
EUES Moses (1) 24.21 0.6228
cdec (2) 24.65 0.5911
ParFDA (3) 22.25 0.6346
Apertium (4) 18.36 0.6918
Morph (5) 11.25 0.9655
1+2 24.18 0.5883
1+4 24.33 0.6076
1+2+4 24.94 0.5831
1+2+4+5 25.21 0.5792
ESPT Moses (1) 29.21 0.6052
cdec (2) 28.14 0.5962
ParFDA (3) 27.74 0.6164
Apertium (4) 24.96 0.6272
1+2 28.76 0.5891
1+4 26.58 0.6082
1+2+4 27.00 0.5878
PTES Moses (1) 30.47 0.5267
cdec (2) 29.42 0.5254
ParFDA (3) 29.63 0.5338
Apertium (4) 27.52 0.5335
1+2 29.9 0.5230
1+4 30.01 0.5131
1+2+4 29.89 0.5089
Table 5: Results on the dev set.

At least one of the combinations obtains better scores (both in terms of BLEU and TER) than the best individual system (except for ESPT with BLEU and for CAES with TER), supporting our hypothesis that the individual systems built are complementary. Although SMT systems outperform RBMT systems for all directions,1515When interpreting the results, it should be taken into account that automatic metrics are known to be biased towards statistical MT approaches [4]. the addition of RBMT in system combinations has a positive impact (except for ESPT). Phrase-based SMT outperforms hierarchical SMT for related language pairs (ES–CA and ES–PT), but the opposite is true for the unrelated language pair ES–EU. We hypothesise this is due to the fact that ES and EU follow different word orders (SVO and SOV, respectively), and this leads to pervasive long reorderings in translation, that are better modelled with a hierarchical approach.

4.2 Evaluation on Test Data

Table 6 presents the results on the test set of the systems we submitted. The scores shown are the ones reported by the organisers (case-insensitive BLEU and TER) on post-processed MT outputs (detokenised and detruecased). For each language direction we submitted the three systems that obtained the best performance on the dev set. The scores of the best submitted system are shown in bold.

  System BLEU TER
ESCA DCU1 (1+4) 0.7669 0.1740
DCU2 (1) 0.7899 0.1626
DCU3 (1+2+4) 0.7630 0.1738
CAES DCU1 (1+4) 0.7826 0.1506
DCU2 (1+2+4) 0.7816 0.1500
DCU3 (1+3+4) 0.7943 0.1431
ESEU DCU1 (1+2+4) 0.2455 0.6533
DCU2 (1+2+3+4+5) 0.2636 0.6469
DCU3 (1+2+4+5) 0.2493 0.6553
EUES DCU1 (2) 0.2687 0.6512
DCU2 (1+2+4) 0.2698 0.6406
DCU3 (1+2+4+5) 0.2728 0.6363
ESPT DCU1 (1) 0.3595 0.5290
DCU2 (1+2) 0.3711 0.5157
DCU3 (1+2+4) 0.3687 0.5163
PTES DCU1 (1) 0.4465 0.5767
DCU2 (1+2) 0.4467 0.5627
DCU3 (1+2+4) 0.4524 0.5403
Table 6: Results on the test set.

Out of six directions, our best submission is the top performing system for five of them (indicated with ). For most directions, the addition of a RBMT system leads to better performance. Similarly, for the directions where we have used segmentation (ESEU) and ParFDA (CAES and ESEU), the addition of systems based on these techniques had a positive impact on the results.

We now delve deeper into the results obtained by SMT systems based on ParFDA (cf. Section 2.4). Although ParFDA systems were submitted to the shared task only as part of system combinations, we have evaluated a posteriori the performance of this technique by means of standalone systems on the test set. ParFDA Moses SMT system obtains top results in CAES and ESCA and close to top results in other language pairs with 1.21 BLEU points average difference to the top (\CrefParFDA_TranslationResults). An interesting feature of ParFDA regards its ability to build and deploy SMT systems in a quick manner. In the specific case of TweetMT, ParFDA took about 8 hours to build for ESCA and 28 hours for PTES taking about 11 GB and 27 GB disk space in total, respectively.

TweetMT CA–ES EU–ES PT–ES
ParFDA .8012 .2713 .4374
Top .7942 .3109 .4519
diff -.007 .0396 .0145
LM order 8 8 8
ES–CA ES–EU ES–PT
ParFDA .7926 .2482 .3589
Top .7907 .2636 .3711
diff -.0019 .0154 .0122
LM order 8 10 8
Table 7: BLEU results for ParFDA standalone systems on the test set, their difference to the top, and ParFDA LM order used. ParFDA obtains top results in CAES and ESCA and 1.21 BLEU points average difference.

5 Conclusions and Future Work

This paper has described our participation in the TweetMT 2015 shared task. Our focus has been on rapid development of MT systems adapted to tweets by making the best possible use of available techniques, tools and resources. Our best submissions have been the ones that combine different MT systems (except for ESCA), supporting our hypothesis that the techniques we have used are complementary.

As for future work, we consider several possible avenues. First, we would like to analyse in detail the translations produced by our systems in order to derive findings beyond the ones we can extract from the automatic evaluation metrics used in the task. Second, most of the tweets in the test set use formal language,1616This is due to the fact that they are extracted from twitter accounts that publish tweets in multiple languages, and such accounts belong, to a large extent, to institutions that use formal language. and thus we would like to test our systems in a more representative set of tweets where informal language would be expected to be more pervasive.

Acknowledgments

This research is supported by the EU 7th Framework Programme FP7/2007-2013 under grant agreement PIAP-GA-2012-324414 (Abu-MaTran), by SFI as part of the ADAPT research center (07/CE/I1142) at Dublin City University and the project “Monolingual and Bilingual Text Quality Judgments with Translation Performance Prediction” (13/TIDA/I2740). We also thank the SFI/HEA Irish Centre for High-End Computing (ICHEC) for the provision of computational facilities and support. Finally, we would like to thank Mikel L. Forcada and Iacer Calixto for their advice on normalising tweets for Basque and Portuguese, respectively, and Gorka Labaka for his help with Matxin’s API.

References

  • [1] E. Biçici, Q. Liu and A. Way (2015-09) ParFDA for fast deployment of accurate statistical machine translation systems, benchmarks, and statistics. In Proceedings of the EMNLP 2015 Tenth Workshop on Statistical Machine Translation, Lisbon, Portugal. Cited by: 1, 2.4.
  • [2] E. Biçici and D. Yuret (2015) Optimizing instance selection for statistical machine translation with feature decay algorithms. IEEE/ACM Transactions On Audio, Speech, and Language Processing (TASLP) 23, pp. 339–350. External Links: Document Cited by: 2.4.
  • [3] E. Biçici (2015) Domain adaptation for machine translation with instance selection. The Prague Bulletin of Mathematical Linguistics 103, pp. 5–20. External Links: ISSN 1804-0462, Document Cited by: 1.
  • [4] C. Callison-Burch, M. Osborne and P. Koehn (2006) Re-evaluation the role of bleu in machine translation research. In 11th Conference of the European Chapter of the Association for Computational Linguistics, pp. 249–256. External Links: Link Cited by: 4.1.
  • [5] C. Dyer, A. Lopez, J. Ganitkevitch, J. Weese, F. Ture, P. Blunsom, H. Setiawan and V. E. and (2010) Cdec: a decoder, alignment, and learning framework for finite-state and context-free translation models. In Proceedings of the Association for Computational Linguistics (ACL), Cited by: 2.2.
  • [6] M. L. Forcada, M. Ginestí-Rosell, J. Nordfalk, J. O’Regan, S. Ortiz-Rojas, J. A. Pérez-Ortiz, G. R. Felipe Sánchez-Martínez and F. M. Tyers (2011) Apertium: a free/open-source platform for rule-based machine translation. Machine Translation 25 (2), pp. 127–144. Note: Special Issue: Free/Open-Source Machine Translation Cited by: 2.2.
  • [7] K. Heafield and A. Lavie (2010) Combining machine translation output with open source: the carnegie mellon multi-engine machine translation scheme. The Prague Bulletin of Mathematical Linguistics 93, pp. 27–36. Cited by: 2.5.
  • [8] R. Kneser and H. Ney (1995) Improved backing-off for m-gram language modeling. In Acoustics, Speech, and Signal Processing, 1995. ICASSP-95., 1995 International Conference on, Vol. 1, pp. 181–184. Cited by: 2.2.
  • [9] P. Koehn, H. Hoang, A. Birch, C. Callison-Burch, M. Federico, N. Bertoldi, B. Cowan, W. Shen, C. Moran, R. Zens, C. Dyer, O. Bojar, A. Constantin and E. Herbst (2007) Moses: open source toolkit for statistical machine translation. In Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions, ACL ’07, Stroudsburg, PA, USA, pp. 177–180. External Links: Link Cited by: 1., 2.2.
  • [10] N. LjubeÅ¡ić and A. Toral (26-31) CaWaC - a web corpus of catalan and its application to language modeling and machine translation. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), Reykjavik, Iceland. Cited by: 3.2.
  • [11] N. Ljubešić, D. Fišer and T. Erjavec (2014) TweetCaT: a Tool for Building Twitter Corpora of Smaller Languages. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), Reykjavik, Iceland. Cited by: 3.2.
  • [12] A. Mayor, I. Alegria, A. D. de Ilarraza Sánchez, G. Labaka, M. Lersundi and K. Sarasola (2011) Matxin, an open-source rule-based machine translation system for basque. Machine Translation 25 (1), pp. 53–82. External Links: Link, Document Cited by: 2.2.
  • [13] Â. Mendonça, D. Jaquette, D. Graff and D. DiPersio (2011) Spanish Gigaword third edition, Linguistic Data Consortium. Philadelphia, USA. Cited by: 2.4.
  • [14] L. Padró and E. Stanilovsky (2012) FreeLing 3.0: towards wider multilinguality. In Proceedings of the Language Resources and Evaluation Conference (LREC 2012), Istanbul, Turkey. Cited by: 2..
  • [15] K. Papineni, S. Roukos, T. Ward and W. Zhu (2002) BLEU: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting on association for computational linguistics, pp. 311–318. Cited by: 4.1.
  • [16] R. Rubino, T. Pirinen, M. Esplà-Gomis, N. LjubeÅ¡ić, S. Ortiz-Rojas, V. Papavassiliou and P. P. and (2015) Abu-MaTran at WMT 2015 Translation Task: Morphological Segmentation and Web Crawling. In Proceedings of the Tenth Workshop on Statistical Machine Translation, Cited by: 2.3.
  • [17] M. Snover, B. Dorr, R. Schwartz, L. Micciulla and J. Makhoul (2006) A study of translation edit rate with targeted human annotation. In Proceedings of Association for machine translation in the Americas, pp. 223–231. Cited by: 4.1.
  • [18] A. Stolcke (2002) SRILM-an extensible language modeling toolkit.. In INTERSPEECH, Cited by: 2.2.
  • [19] J. Tiedemann (23-25) Parallel data, tools and interfaces in opus. In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC’12), N. C. (. Chair), K. Choukri, T. Declerck, M. U. Dogan, B. Maegaard, J. Mariani and J. O. and (Eds.), Istanbul, Turkey (english). External Links: ISBN 978-2-9517408-7-7 Cited by: 3.1.
  • [20] S. Virpioja, P. Smit, S. Grönroos and M. Kurimo (2013) Morfessor 2.0: python implementation and extensions for morfessor baseline. Cited by: 2.3.