Balancing Quality and Human Involvement: An Effective Approach to Interactive Neural Machine Translation

Zhao, Tianxiang; Liu, Lemao; Huang, Guoping; Li, Huayang; Liu, Yingling; Liu, Guiquan; Shi, Shuming

doi:10.1609/aaai.v34i05.6514

Cited by 13 publications

(7 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Peris et al (2019) [21] have implemented online learning techniques in the system, in order to train the model using the error-free translations corrected by the user. Lam et al (2019) [22] and Zhao et al (2020) [23] opted to use reinforcement learning techniques to increase the quality of the translations generated by the MT models, resulting in a reduction in human effort. Both simplified the interaction method by only selecting a set of incorrect words from the translation, or by only checking some critical words, and used the feedback provided by the user not only for correcting the current translation, but also to improve the quality of future translation by using it to train the MT model.…”

Section: Related Workmentioning

confidence: 99%

On the Use of Mouse Actions at the Character Level

Navarro

Casacuberta

2022

Information

View full text Add to dashboard Cite

Neural Machine Translation (NMT) has improved performance in several tasks up to human parity. However, many companies still use Computer-Assisted Translation (CAT) tools to achieve perfect translation, as well as other tools. Among these tools, we find Interactive-Predictive Neural Machine Translation (IPNMT) systems, whose main feature is facilitating machine–human interactions. In the most conventional systems, the human user fixes a translation error by typing the correct word, sending this feedback to the machine which generates a new translation that satisfies it. In this article, we remove the necessity of typing to correct translations by using the bandit feedback obtained from the cursor position when the user performs a Mouse Action (MA). Our system generates a new translation that fixes the error using only the error position. The user can perform multiple MAs at the same position if the error is not fixed, each of which increases the correction probability. One of the main objectives in the IPNMT field is reducing the required human effort, in order to optimize the translation time. With the proposed technique, an 84% reduction in the number of keystrokes performed can be achieved, while still generating perfect translations. For this reason, we recommend the use of this technique in IPNMT systems.

show abstract

Section: Related Workmentioning

confidence: 99%

On the Use of Mouse Actions at the Character Level

Navarro

Casacuberta

2022

Information

View full text Add to dashboard Cite

show abstract

“…A common use case in this line is interactive machine translation (IMT) (Green et al, 2014;Cheng et al, 2016;Peris et al, 2017;Knowles and Koehn, 2016;Santy et al, 2019). IMT systems utilize MT systems to complete the rest of a translation after human translators editing a prefix translation (Alabau et al, 2014;Zhao et al, 2020). For most IMT systems, the core to achieve this completion is prefix-constrained decoding (Wuebker et al, 2016).…”

Section: Related Workmentioning

confidence: 99%

GWLAN: General Word-Level AutocompletioN for Computer-Aided Translation

Liu

Huang

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Computer-aided translation (CAT), the use of software to assist a human translator in the translation process, has been proven to be useful in enhancing the productivity of human translators. Autocompletion, which suggests translation results according to the text pieces provided by human translators, is a core function of CAT. There are two limitations in previous research in this line. First, most research works on this topic focus on sentence-level autocompletion (i.e., generating the whole translation as a sentence based on human input), but word-level autocompletion is under-explored so far. Second, almost no public benchmarks are available for the autocompletion task of CAT. This might be among the reasons why research progress in CAT is much slower compared to automatic MT. In this paper, we propose the task of general word-level autocompletion (GWLAN) from a real-world CAT scenario, and construct the first public benchmark 1 to facilitate research in this topic. In addition, we propose an effective method for GWLAN and compare it with several strong baselines. Experiments demonstrate that our proposed method can give significantly more accurate predictions than the baseline methods on our benchmark datasets.

show abstract

“…Deploying machine learning models [15,23,26,29] in real-world systems often presents the challenge of distribution shift. This occurs when the statistical characteristics of newly incoming data differ from those observed by the model in a dynamically changing environment [3,13,17,21,32].…”

Section: Introductionmentioning

confidence: 99%

T-SaS: Toward Shift-aware Dynamic Adaptation for Streaming Data

Ren,

Zhao,

Qin

et al. 2023

Proceedings of the 32nd ACM International Conference on Information and Knowledge Management

Self Cite

View full text Add to dashboard Cite

In many real-world scenarios, distribution shifts exist in the streaming data across time steps. Many complex sequential data can be effectively divided into distinct regimes that exhibit persistent dynamics. Discovering the shifted behaviors and the evolving patterns underlying the streaming data are important to understand the dynamic system. Existing methods typically train one robust model to work for the evolving data of distinct distributions or sequentially adapt the model utilizing explicitly given regime boundaries. However, there are two challenges: (1) shifts in data streams could happen drastically and abruptly without precursors. Boundaries of distribution shifts are usually unavailable, and ( 2) training a shared model for all domains could fail to capture varying patterns. This paper aims to solve the problem of sequential data modeling in the presence of sudden distribution shifts that occur without any precursors. Specifically, we design a Bayesian framework, dubbed as T-SaS, with a discrete distribution-modeling variable to capture abrupt shifts of data. Then, we design a model that enable adaptation with dynamic network selection conditioned on that discrete variable. The proposed method learns specific model parameters for each distribution by learning which neurons should be activated in the full network. A dynamic masking strategy is adopted here to support inter-distribution transfer through the overlapping of a set of sparse networks. Extensive experiments show that our proposed method is superior in both accurately detecting shift boundaries to get segments of varying distributions and effectively adapting to downstream forecast or classification tasks. CCS CONCEPTS• Computing methodologies → Neural networks; Learning latent representations; Classification and regression trees.

show abstract

Balancing Quality and Human Involvement: An Effective Approach to Interactive Neural Machine Translation

Cited by 13 publications

References 16 publications

On the Use of Mouse Actions at the Character Level

On the Use of Mouse Actions at the Character Level

GWLAN: General Word-Level AutocompletioN for Computer-Aided Translation

T-SaS: Toward Shift-aware Dynamic Adaptation for Streaming Data

Contact Info

Product

Resources

About