Bridging the Gap: An Intermediate Language for Enhanced and Cost-Effective Grapheme-to-Phoneme Conversion with Homographs with Multiple Pronunciations Disambiguation

Abbas Bertina; Elham Esmaeilnia; Hossein Biniazian; Mahdi Pirnia; Shahab Beirami; Soheil Shahi

REVIEW 5 major objections 5 minor 9 references

Bridging the Gap: An Intermediate Language for Enhanced and Cost-Effective Grapheme-to-Phoneme Conversion with Homographs with Multiple Pronunciations Disambiguation

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that converting Persian text to a compact Romanized intermediate language, then transliterating with a sequence-to-sequence model, reaches Persian grapheme-to-phoneme accuracy of 1.96 percent phoneme error and BLEU 94.6.

desk verdict A sensible intermediate-language idea drowned by missing results: the four result tables are empty, the promised formal concept analysis is absent, and the single PER number is measured against LLM-generated labels. read the letter →

arxiv 2505.06599 v1 pith:KLA6B6ND submitted 2025-05-10 cs.CL

Abbas Bertina , Shahab Beirami , Hossein Biniazian , Elham Esmaeilnia , Soheil Shahi , Mahdi Pirnia This is my paper · ORCID

classification cs.CL

keywords grapheme-to-phonemeconversionPersianintermediaterepresentationPinglishhomographdisambiguationEzafesequence-to-sequencetransliterationpolyphones

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to show that Persian grapheme-to-phoneme conversion becomes more accurate and cheaper if the problem is split in two: first turn Persian script into "Pinglish," a deliberately compact Romanized intermediate in which every Persian phoneme has exactly one character, then use a small sequence-to-sequence transliteration model to read Persian and write Pinglish. On this pipeline the authors report a BLEU score of 94.6 and a phoneme error rate of 1.96 percent, and they report strong performance on the two phenomena that trip up prior Persian G2P tools: homographs with multiple pronunciations and the unwritten Ezafe vowel. The system is trained on about 64,000 Persian–Pinglish sentence pairs, expanded to 195,000 by merging and splitting at non-Ezafe boundaries, covering formal and informal Persian. If the numbers are taken as stated, the framework would be a new benchmark for Persian G2P and a template for other homograph-rich languages.

What carries the argument

The load-bearing object is the Pinglish intermediate representation: a modified romanization of Persian with a one-to-one mapping from phonemes to Latin characters, obtained by excluding the ambiguous letters {c, u, q, w, x} and introducing special characters such as "ķ" for the Persian "kh" sound (as in "ķAb" for "sleep"). Around this sits a pipeline whose other moving parts are a rule-based post-processing step that enforces the one-to-one mapping, a custom BPE tokenizer (vocabulary 2,372, maximum subword length 3), and a five-layer encoder-decoder sequence-to-sequence model (8 attention heads, feed-forward dimension 1,024, hidden size 512) trained on Persian-to-Pinglish pairs. The machinery works by turning G2P into a closed-vocabulary transliteration task: because the target alphabet is small, consistent, and phonemically complete, the model can learn the mapping with far fewer labels than a direct Persian-to-IPA system would need.

What would settle it

Take a held-out set of Persian sentences that includes known homographs and Ezafe contexts, transcribe them into phonemes by hand or from an independent pronunciation dictionary, and compare the model's Pinglish output against those human transcriptions; if the phoneme error rate against human labels is materially higher than 0.0196, the reported accuracy is measuring agreement with the generator rather than pronunciation accuracy.

Watch

Extended reading notes

Core claim

The central claim is that representing Persian pronunciation through a one-to-one phoneme-to-character intermediate language, rather than through IPA or direct Persian-to-IPA mapping, removes most of the ambiguity a neural model has to learn. The authors construct this intermediate ("Pinglish") by prompting a large language model with few-shot examples to romanize Persian sentences, then applying a rule-based post-processing step that eliminates ambiguous characters and assigns each Persian phoneme a unique Latin letter. A five-layer encoder-decoder transliteration model with a custom byte-pair-encoding tokenizer is then trained to map Persian graphemes to Pinglish. According to the paper, this yields a BLEU score of 94.6 and a phoneme error rate of 0.0196 on its own test set, with homograph disambiguation and Ezafe detection accuracy reported above the prior model on the SentenceBench corpus. The paper further claims the same intermediate-language recipe extends to Arabic and Chinese, which also have frequent homographs.

Load-bearing premise

The load-bearing assumption is that the machine-created Pinglish labels used for training are correct pronunciations, because the model is trained and tested on random splits of that same machine-made data.

Editorial extensions

If this is right

If the reported PER of 0.0196 is accurate, Persian text-to-speech preprocessing can use this compact transliteration pipeline with lower inference cost than dictionary-heavy or parsing-heavy G2P tools.
The paper's claimed superiority on SentenceBench implies the intermediate-language approach transfers to sentences with homographs and Ezafe better than LLM-prompting alone.
The data-generation recipe—LLM romanization plus rule-based normalization plus sentence merging—could cut the cost of building G2P training data for other low-resource languages.
The one-to-one phoneme-character design principle, if correct, predicts that extending the same framework to Chinese and Arabic would improve polyphone disambiguation in those languages.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

An unstated but immediate test is to score the same model on human-transcribed Persian phonemes or an independent pronunciation dictionary; that would separate genuine pronunciation accuracy from fidelity to the machine-generated labels.
A direct ablation the paper does not report is comparing the encoder-decoder with Pinglish targets against the same architecture with IPA targets; the paper's own rationale predicts Pinglish should win on PER, and that comparison would quantify the benefit of the intermediate representation.
The random train/test split of a single generated corpus leaves open the risk that repeated words or sentence templates appear in both halves; a word-disjoint or sentence-template-disjoint split would be a stronger test of generalization.
If the method is as language-agnostic as claimed, the same two-stage design could be applied to Arabic first, where vowel diacritics are also largely unwritten, and compared against existing Arabic G2P systems.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this paper has a reasonable core idea — a one-to-one Latin intermediate representation ('Pinglish') for Persian G2P, with a BPE tokenizer and a small encoder-decoder trained to transliterate — but it is not in a reviewable state. The headline numbers (BLEU 94.6, PER 0.0196) sit on empty evidence.

What is genuinely new: the Pinglish intermediate language, with a defined phoneme-to-character mapping and excluded ambiguous letters, is a legitimate engineering choice. Training a seq2seq model on LLM-generated Romanized Persian is a plausible low-cost strategy, and the paper describes the architecture and tokenization details clearly enough to reproduce if the data were released. The related work on Persian G2P tools is real and relevant.

Where it falls apart: Tables 2, 3, and 4 are empty placeholders. There is no baseline comparison, no variance, no significance test, and no error analysis. The only reported numbers appear inline in Section 4. The test set is a 1,000-entry random split of the same 64,000-entry corpus that produced the training labels. Those labels were generated by GPT-4o and post-processed by rules; no native-speaker or expert validation is reported. So the PER measures how well the model reproduces GPT-4o's transliterations, not how accurately it pronounces Persian. That is the load-bearing weakness.

Also, the abstract claims a formal concept analysis component for homograph disambiguation. The methodology contains no trace of FCA — no equations, no algorithm, no mention outside the abstract. Either the section is missing or the claim is unsupported. Either way, the paper as written does not deliver what it promises.

The B-Plus podcast data is mentioned as collected, but the dataset section describes only the 64,000 Pinglish entries; it is unclear how the podcast subtitles contributed. Minor compared to the empty tables.

Who is this for? Someone working on low-resource G2P might find the Pinglish idea worth considering, but only as a proposal. The empirical claims are not usable.

Recommendation: desk reject in current form, with an invitation to resubmit when Tables 2-4 actually contain numbers, the test set is externally validated (or at least human-checked), baselines such as Rabiee's Persian_g2p are run on the same data, and the FCA claim is either implemented or removed. It is not ready for referee time yet.

Referee Report

5 major / 5 minor

Summary. The paper proposes an intermediate language called "Pinglish" for Persian grapheme-to-phoneme (G2P) conversion. The pipeline consists of GPT-4o-generated Romanized Persian labels refined by rule-based post-processing, a custom BPE tokenizer, and an EncoderDecoder transliteration model trained on 62,000 entries (augmented to 195,000). The authors report a BLEU score of 94.6 and a phoneme error rate (PER) of 0.0196, and claim that this establishes a new benchmark for Persian G2P, with particular strengths in Ezafe detection and homograph disambiguation. The experimental results section, however, contains only empty placeholders for Tables 2, 3, and 4, and the evaluation is conducted against a random split of the same LLM-generated corpus, with no human validation and no reported comparison against existing systems.

Significance. If the reported results were supported by a verified reference standard and a fair baseline comparison, the intermediate-language approach would be a useful contribution for low-resource G2P conversion, especially for handling Ezafe and homographs. The authors provide some concrete architectural details, a reasonably sized dataset, and a clear motivation for avoiding IPA in favor of a compact one-to-one phoneme-to-character representation. However, the central empirical claims are currently unsupported: the three main result tables are empty, the reference labels are generated by the same LLM used to create training data, and no named baseline is evaluated on the same test set. The work therefore cannot presently substantiate the stated "new benchmark" conclusion.

major comments (5)

[Section 4, Tables 2-4] Tables 2, 3, and 4, which are supposed to report Ezafe detection accuracy, homograph disambiguation results, BLEU/PER, and the SentenceBench comparison, are empty placeholders in the manuscript. Since the conclusion's central claim of achieving BLEU 94.6 and PER 0.0196 rests on these results, and since the text explicitly states that Table 4 shows superiority over an alternative approach, the absence of all numerical data makes the headline results unverifiable.
[Sections 3.1.2, 3.1.3, and 4] The PER of 0.0196 is computed against GPT-4o-generated Pinglish labels, with no native-speaker or expert validation reported. Because the model is trained on a random split of the same LLM-generated corpus and evaluated on another split of that corpus, the reported error rate measures how well the model reproduces the generator's transliteration behavior rather than how accurately it pronounces Persian. The manuscript provides no evidence that the Pinglish labels constitute a correct phonemic reference, so the central benchmark claim is not established.
[Abstract and Section 3] The abstract states that the homograph lexical database is constructed "utilizing formal concept analysis" (FCA), yet no FCA component appears anywhere in the method described in Section 3. The dataset construction pipeline (Sections 3.1.1-3.1.3) uses LLM prompting and rule-based post-processing only. This discrepancy means a claimed contribution is never actually implemented or evaluated.
[Section 5, Table 4] No baseline system is evaluated on the authors' test set. The related work mentions PersianG2P, persian_phonemizer, and Rabiee's Persian_g2p with a PER of 3.9%, but the experimental section reports no comparison against any of these systems on a common benchmark. The only proposed external evaluation, the SentenceBench comparison, is an empty table, so the claim of "superior performance compared to existing state-of-the-art approaches" is unsupported.
[Section 4, data augmentation] The augmentation procedure merges sentences and splits them at non-Ezafe positions to expand the training set from 64,000 to 195,000 samples. The paper asserts that this preserves "linguistic integrity," but no analysis or experiment validates that the augmented examples preserve valid phonemic context. Since the augmented data are used to train the final model, this assumption is load-bearing for the reported performance and needs direct verification.

minor comments (5)

[Section 3.3] The tokenization description is ambiguous: it says "maximum token length of less than 4" and later sets "maximum subword length is limited to 3," but the relationship between these statements and the BPE hyperparameters (vocabulary 2372, minimum frequency 100) is not explained.
[Figures 1-4] Figures 1 through 4 are referenced in the text as schematic overviews of the methodology, dataset construction, mapping examples, and architecture, but the manuscript does not contain the actual figure images, only captions. This makes the methodological description difficult to follow.
[Section 3.1.1] The B-Plus podcast data are mentioned as a source of informal Persian, but the paper never states how many sentences were obtained, how they were preprocessed, or how they were combined with the LLM-generated formal and informal data. The dataset description in Section 3.1.3 mentions only the 64,000 generated entries.
[Throughout] The writing contains frequent grammatical and formatting issues, such as "conversion(G2P)" with no space, "grapheme-to-phoneme conversion with homographs with multiple pronunciations disambiguation" in the title, and the sentence "also the findings of this research are generalizable" in Section 6. A thorough editorial pass is needed.
[Section 2] The references to prior work are often imprecise: for example, Kaamel-Dict and Sentence-Bench are attributed to reference [1] in one place and to [2] and [3] in the references list, and the relation between [1] and [2]/[3] is not clearly stated.

Circularity Check

1 steps flagged · score 6.0 of 10

PER 0.0196 measures fidelity to the LLM-generated Pinglish labels, not independently validated Persian pronunciation.

fitted input called prediction [Sections 3.1.2, 3.1.3 and 4 (Experimental Setup)]
"Using a few-shot learning approach with an LLM, we generated a Romanized version of Persian called 'Pinglish.' ... After post-processing, we amassed a dataset of 64,000 entries, each containing: The original Persian text. Its corresponding Pinglish representation. ... From the initial 64,000 entries, we allocated 1,000 samples each for validation and test evaluation, with the remaining 62,000 samples designated for training data augmentation. ... The model achieved a BLEU score of 94.6 and a PER of 0.0196."

The training targets and the test references are both drawn from the same LLM-generated Pinglish corpus: GPT-4o produced the Romanized text, a rule-based post-processor normalized it, and the 1,000 test samples are a random slice of that same 64,000-entry dataset. No independent phonemic transcription, native-speaker check, or external gold standard is reported for the Pinglish labels. Consequently, the reported PER and BLEU measure how well the seq2seq model imitates GPT-4o's transliteration behavior, not how accurately it pronounces Persian. The evaluation is circular by construction because the reference itself was manufactured by the same generator that defined the target representation; a model trained on this distribution will score well on a held-out portion of it.

full rationale

The paper's central quantitative claims, BLEU 94.6 and PER 0.0196, are computed against references that are generated by the same process that created the training labels. Section 3.1.2 states that Pinglish was produced by prompting an LLM, with rule-based post-processing to enforce one-to-one phoneme-to-character mapping; Section 3.1.3 says the resulting 64,000 entries are pairs of original Persian text and its Pinglish representation; Section 4 then splits this dataset into 62,000 training, 1,000 validation, and 1,000 test samples. There is no external validation of the Pinglish labels anywhere in the manuscript. Thus the headline PER is a measure of the model's fidelity to GPT-4o's output distribution, not a measure of Persian pronunciation accuracy. This is the core circular step: the prediction target is defined by the same LLM whose behavior the model is trained to reproduce, and the test set is a random partition of that same generated corpus. The promised independent check via SentenceBench is not actually presented: Table 4 is an empty placeholder and no SentenceBench numbers appear in the text, so no external benchmark is available to break the circularity. The related-work comparison to Rabiee's PER of 3.9% is also not a controlled comparison because the two PERs are computed on different references. For these reasons, the central empirical claim is partially circular: it reduces to self-consistency with an unvalidated generated standard, warranting a score of 6. No author self-citation chain is load-bearing here, and the model architecture and tokenization choices are not themselves circular; the circularity is localized to the evaluation target.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The paper contains no derivation from first principles; the central claims are empirical. The hand-chosen hyperparameters and LLM-generated target labels are the main unverified inputs. The missing tables and unreleased artifacts mean every numerical claim rests on assumptions about data quality and comparability that the manuscript does not supply.

free parameters (3)

BPE minimum frequency threshold = 100
Hand-chosen in Section 3.3; determines the 2372-token vocabulary and thus the input segmentation, but is not fitted to the PER objective.
Maximum subword token length = 3 (max token length less than 4)
Hand-chosen in Section 3.3 to keep source and target subword distributions similar; affects transliteration accuracy.
Train/validation/test split sizes = 62,000 / 1,000 / 1,000
Split chosen in Section 4; the test set is a random sample of the same LLM-generated corpus, so the reported PER depends on this internal split.

assumptions (4)

domain assumption GPT-4o's Pinglish output and the rule-based post-processing produce a faithful one-to-one phoneme-to-character mapping for formal and informal Persian.
Introduced in Section 3.1.2; if the LLM-generated transliterations contain systematic phonemic errors, the entire training set and PER measure are corrupted.
domain assumption The B-Plus podcast subtitles are representative informal Persian and follow the same phonemic mapping.
Used in Section 3.1.1 as the informal corpus; no manual transcription or alignment check is described.
ad hoc to paper Augmentation by merging sentences and splitting at non-Ezafe positions preserves phonemic context and creates valid training examples.
Described in Section 4; this expansion from 64,000 to 195,000 samples assumes the generated sentence boundaries do not introduce unnatural contexts.
standard math BLEU and PER are computed on the same tokenization and phoneme inventory as used in training.
Equations (1) and (2) in Section 4 define the metrics, but the exact reference phoneme alignment and tokenization are not specified.

invented entities (1)

Pinglish intermediate language
purpose: One-to-one Latin-character representation of Persian phonemes used as the training target for the seq2seq G2P model
Defined in Section 3.1.2 through LLM generation and rule-based post-processing; its validity is only asserted on the same LLM-generated data, with no external dictionary or human verification.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Bridging the Gap: An Intermediate Language for Enhanced and Cost-Effective Grapheme-to-Phoneme Conversion with Homographs with Multiple Pronunciations Disambiguation." pith.science (2026). https://pith.science/paper/KLA6B6ND

@misc{pith2026250506599,
  author       = {Pith},
  title        = {Pith review of: Bridging the Gap: An Intermediate Language for Enhanced and Cost-Effective Grapheme-to-Phoneme Conversion with Homographs with Multiple Pronunciations Disambiguation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KLA6B6ND}},
  note         = {Machine review of arXiv:2505.06599}
}

read the original abstract

Grapheme-to-phoneme (G2P) conversion for Persian presents unique challenges due to its complex phonological features, particularly homographs and Ezafe, which exist in formal and informal language contexts. This paper introduces an intermediate language specifically designed for Persian language processing that addresses these challenges through a multi-faceted approach. Our methodology combines two key components: Large Language Model (LLM) prompting techniques and a specialized sequence-to-sequence machine transliteration architecture. We developed and implemented a systematic approach for constructing a comprehensive lexical database for homographs with multiple pronunciations disambiguation often termed polyphones, utilizing formal concept analysis for semantic differentiation. We train our model using two distinct datasets: the LLM-generated dataset for formal and informal Persian and the B-Plus podcasts for informal language variants. The experimental results demonstrate superior performance compared to existing state-of-the-art approaches, particularly in handling the complexities of Persian phoneme conversion. Our model significantly improves Phoneme Error Rate (PER) metrics, establishing a new benchmark for Persian G2P conversion accuracy. This work contributes to the growing research in low-resource language processing and provides a robust solution for Persian text-to-speech systems and demonstrating its applicability beyond Persian. Specifically, the approach can extend to languages with rich homographic phenomena such as Chinese and Arabic

Figures

Figures reproduced from arXiv: 2505.06599 by the authors.

**Figure 2.** Fig.2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** Fig.4 [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

9 extracted references · 7 canonical work pages

[1]

ه -ﮐﺴﺮه" and

Introduction Grapheme-to-phoneme (G2P) conversion represents a significant challenge in speech processing and natural language understanding, particularly for languages with complex phonological systems. As a morphologically rich and phonetically ambiguous language, Persian presents unique difficulties due to the absence of diacritics in standard writing,...

work page
[2]

words with multiple pronunciations

Related Work Recent advances in G2P conversion have demonstrated various innovative approaches and methodologies. The Persian phonemizer represents a significant advancement in Persian language processing, offering an automated solution for converting Persian text to the International Phonetic Alphabet ( IPA) [22] notation through an integrated approach c...

work page
[3]

ﺷﯿﺮ" is not considered a homograph in our study, despite having multiple meanings, because its pronunciation remains the same across different contexts. However, the word

Methodology This section introduces our methodology, including dataset construction for Persian G2P conversion, data collection, proposed intermediate language, and model architecture. The architectural overview of the proposed system is depicted in Figure 1, demonstrating the pipeline from data collection to phoneme conversion. Fig.1. Schematic overview ...

work page
[4]

From the initial 64,000 entries, we allocated 1,000 samples each for validation and test evaluation, with the remaining 62,000 samples designated for training data augmentation

Experimental Setup Our experimental framework utilized the dataset described in the previous section, implementing a strategic split and augmentation approach. From the initial 64,000 entries, we allocated 1,000 samples each for validation and test evaluation, with the remaining 62,000 samples designated for training data augmentation. Through sentence me...

work page
[5]

Results and Analysis To evaluate the effectiveness of our proposed model, we conducted extensive experiments on both our custom dataset and the SentenceBench dataset [3]. Table 2 presents the performance of our model in detecting Ezafe constructions over 50 epochs on our test set, demonstrating its ability to accurately infer implicit phonetic markers tha...

work page
[6]

Our research makes several significant contributions to the field of Persian language processing

Conclusion This paper presents a novel approach to Persian grapheme-to-phoneme conversion by developing an intermediate language system that effectively addresses the challenges of homographs with multiple pronunciations disambiguation and Ezafe detection. Our research makes several significant contributions to the field of Persian language processing. Fi...

work page
[7]

LLM-powered grapheme-to-phoneme conversion: Benchmark and case study,

Future work Several promising research directions emerge from this work. The primary focus will be extending the intermediate language framework to morphologically rich languages, investigating language-specific adaptations for phonological complexities, and developing universal phonological mapping strategies. We plan to explore advanced contextual proce...

arXiv 2024
[13]

Encoder-Decoder Models,

Hugging Face, “Encoder-Decoder Models,” [Online]. Available: https://huggingface.co/docs/transformers/v4.46.0/mo del_doc/encoder-decoder. Accessed: Mar. 25, 2025. [14] Hugging Face, “Transformers-based Encoder-Decoder Models,” 2020. [Online]. Available: https://huggingface.co/ blog/encoder-decoder. Accessed: Mar. 25, 2025. [15] K. Papineni, S. Roukos, T. ...

work page 2025

Show all 9 references

[21]

Persian G2P,

A. Rabiee, “Persian G2P,” GitHub, 2019. [Online]. Available: https://github.com/AzamRabiee/Persian_G2P. Accessed: Mar. 25, 2025. [22] International Phonetic Association, “The International Phonetic Alphabet,” 1888. [Online]. Available: https://www. internationalphoneticalphabe...

2019 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

ه -ﮐﺴﺮه" and

Introduction Grapheme-to-phoneme (G2P) conversion represents a significant challenge in speech processing and natural language understanding, particularly for languages with complex phonological systems. As a morphologically rich and phonetically ambiguous language, Persian presents unique difficulties due to the absence of diacritics in standard writing,...

work page

[2] [2]

words with multiple pronunciations

Related Work Recent advances in G2P conversion have demonstrated various innovative approaches and methodologies. The Persian phonemizer represents a significant advancement in Persian language processing, offering an automated solution for converting Persian text to the International Phonetic Alphabet ( IPA) [22] notation through an integrated approach c...

work page

[3] [3]

ﺷﯿﺮ" is not considered a homograph in our study, despite having multiple meanings, because its pronunciation remains the same across different contexts. However, the word

Methodology This section introduces our methodology, including dataset construction for Persian G2P conversion, data collection, proposed intermediate language, and model architecture. The architectural overview of the proposed system is depicted in Figure 1, demonstrating the pipeline from data collection to phoneme conversion. Fig.1. Schematic overview ...

work page

[4] [4]

From the initial 64,000 entries, we allocated 1,000 samples each for validation and test evaluation, with the remaining 62,000 samples designated for training data augmentation

Experimental Setup Our experimental framework utilized the dataset described in the previous section, implementing a strategic split and augmentation approach. From the initial 64,000 entries, we allocated 1,000 samples each for validation and test evaluation, with the remaining 62,000 samples designated for training data augmentation. Through sentence me...

work page

[5] [5]

Results and Analysis To evaluate the effectiveness of our proposed model, we conducted extensive experiments on both our custom dataset and the SentenceBench dataset [3]. Table 2 presents the performance of our model in detecting Ezafe constructions over 50 epochs on our test set, demonstrating its ability to accurately infer implicit phonetic markers tha...

work page

[6] [6]

Our research makes several significant contributions to the field of Persian language processing

Conclusion This paper presents a novel approach to Persian grapheme-to-phoneme conversion by developing an intermediate language system that effectively addresses the challenges of homographs with multiple pronunciations disambiguation and Ezafe detection. Our research makes several significant contributions to the field of Persian language processing. Fi...

work page

[7] [7]

LLM-powered grapheme-to-phoneme conversion: Benchmark and case study,

Future work Several promising research directions emerge from this work. The primary focus will be extending the intermediate language framework to morphologically rich languages, investigating language-specific adaptations for phonological complexities, and developing universal phonological mapping strategies. We plan to explore advanced contextual proce...

arXiv 2024

[8] [13]

Encoder-Decoder Models,

Hugging Face, “Encoder-Decoder Models,” [Online]. Available: https://huggingface.co/docs/transformers/v4.46.0/mo del_doc/encoder-decoder. Accessed: Mar. 25, 2025. [14] Hugging Face, “Transformers-based Encoder-Decoder Models,” 2020. [Online]. Available: https://huggingface.co/ blog/encoder-decoder. Accessed: Mar. 25, 2025. [15] K. Papineni, S. Roukos, T. ...

work page 2025