Recoverable Identifier

arXiv:2604.26230 · detector doi_compliance · incontrovertible · 2026-05-19 20:23:25.081824+00:00

Paper page Integrity report arXiv Try DOI

Evidence text

I analyze the relationship between the correlation coefficient and perplexity scores. If strong correlation is found between them, perplexity scores can be used to optimize 9 The LWIC dictionaries are created in three steps: (1) collect candidate words from English dictionaries and thesauri, (2) select only relevant words from the candidate words by employing three independent judges, (3) remove words that occur infrequently or cause inconsistency by applying the dictionaries to example texts (Pennebaker et al., 2015). 10 See Appendix for the list of seed words used in the evaluation. 11 In word2vec, the windows size 𝑑 is 5 for CBOW and 10 for SG. The size of the input layer 𝑘 is 50, 100, 150, 200, 250 or 300; the learning rate is 0.05; and the number of iterations is 10 for both algorithms. In LSS, seed words are initialized with uniform polarity scores 𝑝 that sum to one. hyperparameters such as the size of the input layer or the training algorithm in the absence of gold standard. Results The results of evaluation show that word2vec-based probabilistic models are most strongly correlated with the full dictionaries in both ‘achievement’ and ‘health’ (Figure 2). The correlation is particularly strong when the SG algorithm is used to train them, reaching r=0.51 and r=0.66 in the respective concepts. The spatial models were also trained using the same algorithm, but they achieved significantly weaker correlations (r=0.47 and r=0.54, respectively). The SVD-based models performed 

Evidence payload

{
  "printed_excerpt": "I analyze the relationship between the correlation coefficient and perplexity scores. If strong correlation is found between them, perplexity scores can be used to optimize 9 The LWIC dictionaries are created in three steps: (1) collect can",
  "reconstructed_doi": "10.1109/BTS-I2C63534.2024.1094221213",
  "ref_index": 5,
  "resolved_title": null,
  "verdict_class": "incontrovertible"
}