Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Translation from the Information Bottleneck Perspective: an Efficiency Analysis of Spatial Prepositions in Bitexts

Antoine Taroni, Ludovic Moncla, Frederique Laforest

Abstract

Efficient communication requires balancing informativity and simplicity when encoding meanings. The Information Bottleneck (IB) framework captures this trade-off formally, predicting that natural language systems cluster near an optimal accuracy-complexity frontier. While supported in visual domains such as colour and motion, linguistic stimuli such as words in sentential context remain unexplored. We address this gap by framing translation as an IB optimisation problem, treating source sentences as stimuli and target sentences as compressed meanings. This allows IB analyses to be performed directly on bitexts rather than controlled naming experiments. We applied this to spatial prepositions across English, German and Serbian translations of a French novel. To estimate informativity, we conducted a pile-sorting pilot-study (N=35) and obtained similarity judgements of pairs of prepositions. We trained a low-rank projection model (D=5) that predicts these judgements (Spearman correlation: 0.78). Attested translations of prepositions lie closer to the IB optimal frontier than counterfactual alternatives, offering preliminary evidence that human translators exhibit communicative efficiency pressure in the spatial domain. More broadly, this work suggests that translation can serve as a window into the cognitive efficiency pressures shaping cross-linguistic semantic systems.

Translation from the Information Bottleneck Perspective: an Efficiency Analysis of Spatial Prepositions in Bitexts

Abstract

Efficient communication requires balancing informativity and simplicity when encoding meanings. The Information Bottleneck (IB) framework captures this trade-off formally, predicting that natural language systems cluster near an optimal accuracy-complexity frontier. While supported in visual domains such as colour and motion, linguistic stimuli such as words in sentential context remain unexplored. We address this gap by framing translation as an IB optimisation problem, treating source sentences as stimuli and target sentences as compressed meanings. This allows IB analyses to be performed directly on bitexts rather than controlled naming experiments. We applied this to spatial prepositions across English, German and Serbian translations of a French novel. To estimate informativity, we conducted a pile-sorting pilot-study (N=35) and obtained similarity judgements of pairs of prepositions. We trained a low-rank projection model (D=5) that predicts these judgements (Spearman correlation: 0.78). Attested translations of prepositions lie closer to the IB optimal frontier than counterfactual alternatives, offering preliminary evidence that human translators exhibit communicative efficiency pressure in the spatial domain. More broadly, this work suggests that translation can serve as a window into the cognitive efficiency pressures shaping cross-linguistic semantic systems.
Paper Structure (21 sections, 4 equations, 8 figures, 1 table)

This paper contains 21 sections, 4 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Theoretical framework
  3. General methodology
  4. Naming data
  5. Similarity judgements
  6. Experiments
  7. Naming data: detection and alignment of spatial prepositions
  8. Models of human similarity judgements
  9. Computation of encoders locations in the information plane
  10. Discussion
  11. Characterising cross-linguistic use of spatial prepositions
  12. Estimation of similarity between spatial relations evoked by prepositions
  13. Efficiency of attested systems
  14. Conclusion
  15. Ethics Statement
  16. ...and 6 more sections

Figures (8)

  • Figure 1: Excerpt of corpus. Spatial prepositions in the French source sentences are emphasised. id refers to the original identifier as released in the corpus.
  • Figure 2: French to English excerpt of computed alignments of spatial prepositions from corpus. Each row is a one-hot encoding of the source preposition over the target lexicon. The source preposition is emphasised in bold.
  • Figure 3: Empirical similarity data from the pile-sorting task.
  • Figure 4: Attested (stars) and counterfactual (dots) translations of spatial prepositions. As the proportion of random permutations applied to the attested translations increases, counterfactual systems deviate further from the IB curve. Points are slightly jittered along the Complexity axis for visualisation.
  • Figure 5: Deviation from optimality of attested translation, random and counterfactual ones. The black error bars stand for standard deviations.
  • ...and 3 more figures