Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
Tu, Mei and Zhou, Yu and Zong, Chengqing

Article Structure

Abstract

Transitional expressions provide glue that holds ideas together in a text and enhance the logical organization, which together help improve readability of a text.

Introduction

During the last decade, great progress has been made on statistical machine translation (SMT) models.

Chinese Compound-Complex Sentence Structure

To acquire the functional relationships of a Chinese compound-complex sentence, Zhou (2004) proposed a well-annotated scheme to build the Compound-complex Sentence Structure (CSS).

A semantic span can include one or more eus.

3 Modelling

Experiments

4.1 Experimental Setup

Related Work

Improving cohesion for complex sentences or discourse translation has attracted much attention in recent years.

Conclusion

In this paper, we focus on capturing cohesion information to enhance the grammatical cohesion

Topics

translation model

Appears in 10 sentences as: Translation Model (1) translation model (7) translation models (2)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Our models include a CSS-based translation model , which generates new CSS-based translation rules, and a generative transfer model, which encourages producing transitional expressions during decoding.
    Page 1, “Abstract”
  2. One is a new translation model that is utilized to generate new translation rules combined with the information of source functional relationships.
    Page 2, “Introduction”
  3. 1) CSS-based translation model : following formula (1), we obtain the cohesion information by modifying the translation rules with their probabilities P(es Ift) based on word align-
    Page 3, “A semantic span can include one or more eus.”
  4. 3.1 CSS-based Translation Model
    Page 3, “A semantic span can include one or more eus.”
  5. For the existing translation models , the entire training process is conducted at the lexical or syntactic level without grammatically cohesive information.
    Page 3, “A semantic span can include one or more eus.”
  6. Most translation systems adopt the features from a translation model , a language model, and sometimes a reordering model.
    Page 4, “A semantic span can include one or more eus.”
  7. The bilingual training data for translation model and CSS-based transfer model is FBIS corpus with approximately 7.1 million Chinese words and 9.2 million English words.
    Page 6, “Experiments”
  8. For this work, we use an in-house decoder to build the SMT baseline; it combines the hierarchical phrase-based translation model (Chiang, 2005; Chiang, 2007) with the BTG (Wu, 1996) reordering model (Xiong et al., 2006; Zens and Ney, 2006; He et al., 2010).
    Page 7, “Experiments”
  9. To further evaluate the effectiveness of the proposed models, we also conducted an experiment on a larger set of bilingual training data from the LDC corpus7 for translation model and transfer model.
    Page 8, “Experiments”
  10. In the future, we will extend our methods to other translation models , such as the syntax-based model, to study how to further improve the performance of SMT systems.
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention translation model.

See all papers in Proc. ACL that mention translation model.

Back to top.

BLEU

Appears in 8 sentences as: BLEU (8)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. In Table 3, almost all BLEU scores are improved, no matter what strategy is used.
    Page 7, “Experiments”
  2. In particular, the best performance marked in bold is as high as 1.24, 0.94, and 0.82 BLEU points, respectively, over the baseline system on NIST04, CWMT08 Development, and CWMT08 Evaluation data.
    Page 7, “Experiments”
  3. BLEU 35
    Page 8, “Experiments”
  4. The final BLEU scores on NIST05 and NIST06 are given in Table 4.
    Page 8, “Experiments”
  5. The best performance with bold marking scored as high as 0.83 and 0.64 BLEU points, respectively over the
    Page 8, “Experiments”
  6. BLEU scores on the large-scale training data.
    Page 8, “Experiments”
  7. They added the labels assigned to connectives as an additional input to an SMT system, but their experimental results show that the improvements under the evaluation metric of BLEU were not significant.
    Page 9, “Related Work”
  8. To the best of our knowledge, our work is the first attempt to exploit the source functional relationship to generate the target transitional expressions for grammatical cohesion, and we have successfully incorporated the proposed models into an SMT system with significant improvement of BLEU metrics.
    Page 9, “Related Work”

See all papers in Proc. ACL 2014 that mention BLEU.

See all papers in Proc. ACL that mention BLEU.

Back to top.

machine translation

Appears in 7 sentences as: Machine Translation (1) machine translation (6)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. However, in most current statistical machine translation (SMT) systems, the outputs of compound-complex sentences still lack proper transitional expressions.
    Page 1, “Abstract”
  2. During the last decade, great progress has been made on statistical machine translation (SMT) models.
    Page 1, “Introduction”
  3. 5 http://www.speech.sri.com/projects/srilm/ 6 The China Workshop on Machine Translation
    Page 6, “Experiments”
  4. In (Xiong et al., 2013a), three different features were designed to capture the lexical cohesion for document-level machine translation .
    Page 8, “Related Work”
  5. (Xiong et al., 2013b) incorporated lexical-chain-based models (Morris and Hirst, 1991) into machine translation .
    Page 8, “Related Work”
  6. (Meyer and Popescu-Belis, 2012) used sense-labeled discourse connectives for machine translation from English to French.
    Page 9, “Related Work”
  7. of machine translation .
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention machine translation.

See all papers in Proc. ACL that mention machine translation.

Back to top.

word alignment

Appears in 6 sentences as: word alignment (7)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Instead, we reserve the cohesive information in the training process by converting the original source sentence into tagged-flattened CSS and then perform word alignment and extract the translation rules from the bilingual flattened source CSS and the target string.
    Page 3, “A semantic span can include one or more eus.”
  2. We then perform word alignment on the modified bilingual sentences, and extract the new translation rules based on the new alignment, as shown in Figure 3(b) to Figure 3(c).
    Page 4, “A semantic span can include one or more eus.”
  3. bound by the word alignment , the alignment complies with EUC only if there is no overlap between pSA and pSB.
    Page 5, “A semantic span can include one or more eus.”
  4. If A is the word alignment of (f, e), then the goal is to construct the maximum subset A*g A under the condition that A * is the word alignment with the constraint of EU.
    Page 5, “A semantic span can include one or more eus.”
  5. We obtain the word alignment with the grow-diag-final-and strategy with GIZA++.
    Page 6, “Experiments”
  6. The merits of “Flattened Rule” are twofold: 1) In training process, the new word alignment upon modified sentence pairs can align transitional expressions to flattened CSS tags; 2) In decoding process, the CSS-based rules are more discriminating than the original rules, which is more flexible than “TFS”.
    Page 7, “Experiments”

See all papers in Proc. ACL 2014 that mention word alignment.

See all papers in Proc. ACL that mention word alignment.

Back to top.

language model

Appears in 5 sentences as: language model (5)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Most translation systems adopt the features from a translation model, a language model , and sometimes a reordering model.
    Page 4, “A semantic span can include one or more eus.”
  2. The process of training this transfer model and smoothing is similar to the process of training a language model .
    Page 4, “A semantic span can include one or more eus.”
  3. formula (6) are estimated in the same way as a factored language model , which has the advantage of easily incorporating various linguistic information.
    Page 4, “A semantic span can include one or more eus.”
  4. A 5-gram language model is trained with SRILM5 on the combination of the Xinhua portion of the English Giga-word corpus combined with the English part of FBIS.
    Page 6, “Experiments”
  5. probabilities, the BTG reordering features, and the language model feature.
    Page 7, “Experiments”

See all papers in Proc. ACL 2014 that mention language model.

See all papers in Proc. ACL that mention language model.

Back to top.

proposed models

Appears in 5 sentences as: proposed models (5)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. To test the effectiveness of the proposed models , we have compared the translation quality of different integration strategies.
    Page 7, “Experiments”
  2. To further evaluate the effectiveness of the proposed models , we also conducted an experiment on a larger set of bilingual training data from the LDC corpus7 for translation model and transfer model.
    Page 8, “Experiments”
  3. The results in Table 4 further verify the effectiveness of our proposed models .
    Page 8, “Experiments”
  4. To the best of our knowledge, our work is the first attempt to exploit the source functional relationship to generate the target transitional expressions for grammatical cohesion, and we have successfully incorporated the proposed models into an SMT system with significant improvement of BLEU metrics.
    Page 9, “Related Work”
  5. The experimental results show that significant improvements have been achieved on various test data, meanwhile the translations are more cohesive and smooth, which together demonstrate the effectiveness of our proposed models .
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention proposed models.

See all papers in Proc. ACL that mention proposed models.

Back to top.

translation quality

Appears in 5 sentences as: translation qualities (1) translation quality (4)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. In general, according to formula (3), the translation quality based on the log-linear model is related tightly with the features chosen.
    Page 4, “A semantic span can include one or more eus.”
  2. According to the statistics in Table 1, we see that CSS is really widely distributed in the NIST and CWMT corpora, which implies that the translation quality may benefit substantially from the CSS information, if it is well considered in SMT.
    Page 6, “Experiments”
  3. To test the effectiveness of the proposed models, we have compared the translation quality of different integration strategies.
    Page 7, “Experiments”
  4. From the table, we cannot conclude that the EUC constraint will certainly promote translation quality , but the transfer model performs better with the constraint on most testing sets.
    Page 7, “Experiments”
  5. Different translation qualities along with different n-grams for transfer model.
    Page 8, “Experiments”

See all papers in Proc. ACL 2014 that mention translation quality.

See all papers in Proc. ACL that mention translation quality.

Back to top.

conditional probabilities

Appears in 4 sentences as: conditional probabilities (2) conditional probability (2)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Therefore, theoretically, the conditional probability of a target translation es conditioned on the source CSS-based tree ft is given by P(es | fl) ,
    Page 3, “A semantic span can include one or more eus.”
  2. The conditional probabilities of the new translation rules are calculated following (Chiang, 2005).
    Page 4, “A semantic span can include one or more eus.”
  3. transfer model can be represented as a conditional probability : P(w | CSS) (4) By deriving each node of the CSS, we can obtain a factored formula: P(w I CSS) = Hm.
    Page 4, “A semantic span can include one or more eus.”
  4. Following (Bilmes and Kirchhoff, 2003), the conditional probabilities P(wk Iwok‘1,Ri,RPj) in
    Page 4, “A semantic span can include one or more eus.”

See all papers in Proc. ACL 2014 that mention conditional probabilities.

See all papers in Proc. ACL that mention conditional probabilities.

Back to top.

log-linear

Appears in 4 sentences as: log-linear (4)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Following Och and Ney (2002), our model is framed as a log-linear model:
    Page 3, “A semantic span can include one or more eus.”
  2. courage the decoder to generate transitional words and phrases; the score is utilized as an additional feature hk (es, ft) in the log-linear model.
    Page 3, “A semantic span can include one or more eus.”
  3. In general, according to formula (3), the translation quality based on the log-linear model is related tightly with the features chosen.
    Page 4, “A semantic span can include one or more eus.”
  4. Our contributions can be summarized as: l) the new translation rules are more discriminative and sensitive to cohesive information by converting the source string into a CSS-based tagged-flattened string; 2) the new additional features embedded in the log-linear model can encourage the decoder to produce transitional expressions.
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention log-linear.

See all papers in Proc. ACL that mention log-linear.

Back to top.

log-linear model

Appears in 4 sentences as: log-linear model (3) log-linear model: (1)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. Following Och and Ney (2002), our model is framed as a log-linear model:
    Page 3, “A semantic span can include one or more eus.”
  2. courage the decoder to generate transitional words and phrases; the score is utilized as an additional feature hk (es, ft) in the log-linear model .
    Page 3, “A semantic span can include one or more eus.”
  3. In general, according to formula (3), the translation quality based on the log-linear model is related tightly with the features chosen.
    Page 4, “A semantic span can include one or more eus.”
  4. Our contributions can be summarized as: l) the new translation rules are more discriminative and sensitive to cohesive information by converting the source string into a CSS-based tagged-flattened string; 2) the new additional features embedded in the log-linear model can encourage the decoder to produce transitional expressions.
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention log-linear model.

See all papers in Proc. ACL that mention log-linear model.

Back to top.

SMT system

Appears in 4 sentences as: SMT system (3) SMT systems (1)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. In Section 3, we present our models and show how to integrate the models into an SMT system .
    Page 2, “Introduction”
  2. They added the labels assigned to connectives as an additional input to an SMT system , but their experimental results show that the improvements under the evaluation metric of BLEU were not significant.
    Page 9, “Related Work”
  3. To the best of our knowledge, our work is the first attempt to exploit the source functional relationship to generate the target transitional expressions for grammatical cohesion, and we have successfully incorporated the proposed models into an SMT system with significant improvement of BLEU metrics.
    Page 9, “Related Work”
  4. In the future, we will extend our methods to other translation models, such as the syntax-based model, to study how to further improve the performance of SMT systems .
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention SMT system.

See all papers in Proc. ACL that mention SMT system.

Back to top.

BLEU scores

Appears in 3 sentences as: BLEU scores (3)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. In Table 3, almost all BLEU scores are improved, no matter what strategy is used.
    Page 7, “Experiments”
  2. The final BLEU scores on NIST05 and NIST06 are given in Table 4.
    Page 8, “Experiments”
  3. BLEU scores on the large-scale training data.
    Page 8, “Experiments”

See all papers in Proc. ACL 2014 that mention BLEU scores.

See all papers in Proc. ACL that mention BLEU scores.

Back to top.

n-grams

Appears in 3 sentences as: N-grams (1) n-grams (2)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. 4.4 Analysis of Different Effects of Different N-grams
    Page 7, “Experiments”
  2. To evaluate the effects of different n-grams for our proposed transfer model, we compared the uni-/bi-/tri-gram transfer models in SMT, and illustrate the results in Fig-
    Page 7, “Experiments”
  3. Different translation qualities along with different n-grams for transfer model.
    Page 8, “Experiments”

See all papers in Proc. ACL 2014 that mention n-grams.

See all papers in Proc. ACL that mention n-grams.

Back to top.

significant improvements

Appears in 3 sentences as: significant improvement (1) significant improvements (2)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. The experimental results show that significant improvements are achieved on various test data meanwhile the translations are more cohesive and smooth.
    Page 1, “Abstract”
  2. To the best of our knowledge, our work is the first attempt to exploit the source functional relationship to generate the target transitional expressions for grammatical cohesion, and we have successfully incorporated the proposed models into an SMT system with significant improvement of BLEU metrics.
    Page 9, “Related Work”
  3. The experimental results show that significant improvements have been achieved on various test data, meanwhile the translations are more cohesive and smooth, which together demonstrate the effectiveness of our proposed models.
    Page 9, “Conclusion”

See all papers in Proc. ACL 2014 that mention significant improvements.

See all papers in Proc. ACL that mention significant improvements.

Back to top.

translation system

Appears in 3 sentences as: translation system (2) translation systems (1)
In Enhancing Grammatical Cohesion: Generating Transitional Expressions for SMT
  1. The two models are integrated into a hierarchical phrase-based translation system to evaluate their effectiveness.
    Page 1, “Abstract”
  2. Most translation systems adopt the features from a translation model, a language model, and sometimes a reordering model.
    Page 4, “A semantic span can include one or more eus.”
  3. First, we adopted only the tagged-flattened rules in the hierarchical translation system .
    Page 7, “Experiments”

See all papers in Proc. ACL 2014 that mention translation system.

See all papers in Proc. ACL that mention translation system.

Back to top.