{"id":"31d52a60-708c-4d42-9ff8-8bafb73adfb2","arxiv_id":"2605.04583","paper_version":3,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"TajikNLP is an open-source toolkit that supplies the first comprehensive text processing pipeline for Tajik Cyrillic, accompanied by four new linguistic datasets.","lead":"The paper introduces TajikNLP, an open-source Python library offering the first full NLP pipeline for Tajik text in Cyrillic, covering cleaning, tokenization, POS tagging, morphology, sentiment, and embeddings. A smart generalist might read it to understand how basic language tools are built for severely under-resourced languages.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Claim that novel morphology engine 'significantly improves' agglutinative handling rests only on internal test coverage, not accuracy metrics or baselines.","rationale":"Reader's weakest_assumption directly identifies the missing quantitative validation for the morphology component. This is the load-bearing point for any claim that the toolkit meaningfully advances processing capability rather than merely packaging existing rules. Full text (if it contains unreported metrics) could resolve it; absent that, the concern stands.","tokens_in":1809,"tokens_out":308,"duration_ms":12248,"concrete_test":"Add a results table reporting token-level accuracy or F1 for morphemic segmentation and lemmatization on a 20% held-out split of the released POS-tagged corpus (or on 500 manually annotated inflected forms); if scores fall below 85% or show no gain over a simple suffix-stripping baseline, the improvement claim does not hold.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that the unified morphology engine (controlled/deep modes) actually delivers measurable gains on Tajik inflections. Abstract states this improvement occurs, yet validation is limited to '616 automated tests achieving 93% source code coverage' with no reported precision/recall, F1 on the 52.5k POS corpus, or comparison against prior Tajik stemmers/lemmatizers. Without these numbers the 'significantly improve' assertion is unsupported, weakening the infrastructure claim for users who need reliable performance.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces TajikNLP, the first open-source Python toolkit for comprehensive Tajik (Cyrillic) text processing. It describes a modular pipeline built around a unified Doc object that chains cleaning, normalization, BPE tokenization, morphemic segmentation, POS tagging, stemming, lemmatization, and sentence splitting. A novel unified morphology engine with controlled and deep modes is claimed to significantly improve handling of agglutinative inflections. The release includes a lexicon-based sentiment analyzer, pre-trained Word2Vec/FastText embeddings from Hugging Face, and four new datasets (52.5k POS-tagged corpus, 3.5k sentiment lexicon, 5.6k toponym gazetteer, 3.8k personal names). Reliability is asserted via 616 automated tests achieving 93% source-code coverage. The work positions itself as foundational infrastructure for low-resource Tajik NLP.","tokens_in":1909,"tokens_out":512,"duration_ms":14318,"significance":"If the morphology engine and pipeline components deliver reliable performance, the release would provide the first publicly available, end-to-end toolkit and accompanying datasets for Tajik, directly addressing a documented gap in Cyrillic-script low-resource languages. The permissive licensing of code and data, the Hugging Face integration, and the extensive test suite constitute concrete, reusable contributions that could accelerate both academic research and downstream applications.","major_comments":[{"comment":"Abstract: the central claim that the unified morphology engine 'significantly improve handling of Tajik's agglutinative nominal and verbal inflections' is unsupported by any reported accuracy, precision/recall, or F1 figures on the 52.5k POS corpus, any error analysis of inflection cases, or any comparison against prior Tajik stemmers/lemmatizers. Only the existence of 616 tests at 93% coverage is stated; coverage alone does not establish measurable gains on the agglutinative phenomena highlighted as the novel contribution.","section":"Abstract"}],"minor_comments":[{"comment":"The manuscript would benefit from an explicit section or table listing the exact test cases that exercise the morphology engine's controlled vs. deep modes and the specific inflection phenomena they target.","section":null},{"comment":"Dataset release statements should include direct links or DOIs in addition to license information to facilitate immediate reuse.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for highlighting the need for empirical support of the morphology engine's performance claims. We address the concern directly below and commit to revisions that strengthen the manuscript's evidential basis.","responses":[{"response":"We agree the current version lacks quantitative evaluation (accuracy, F1, error analysis, or baselines) for the morphology engine on agglutinative inflections, and that the 616 tests measure code coverage rather than task performance. The 52.5k POS corpus exists but was not used for morphology-specific metrics in the submitted manuscript. In revision we will add a dedicated evaluation subsection reporting precision/recall/F1 on a held-out split of the corpus for stemming and lemmatization (with focus on nominal/verbal agglutination), plus any available comparisons to prior Tajik stemmers. If time constraints prevent full new experiments, we will moderate the abstract language to remove the unsupported 'significantly improve' phrasing while retaining the description of the engine's design.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the central claim that the unified morphology engine 'significantly improve handling of Tajik's agglutinative nominal and verbal inflections' is unsupported by any reported accuracy, precision/recall, or F1 figures on the 52.5k POS corpus, any error analysis of inflection cases, or any comparison against prior Tajik stemmers/lemmatizers. Only the existence of 616 tests at 93% coverage is stated; coverage alone does not establish measurable gains on the agglutinative phenomena highlighted as the novel contribution."}],"tokens_in":1493,"tokens_out":344,"duration_ms":15469,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper's main deliverable is a Python library that chains cleaning, BPE tokenization, morphemic segmentation, POS tagging, stemming, lemmatization, and sentence splitting for Tajik Cyrillic text, plus a sentiment analyzer and pre-trained embeddings. It also releases four new datasets: a 52.5k POS-tagged corpus, a 3.5k sentiment lexicon, a 5.6k toponym gazetteer, and a 3.8k names list. That combination is new for this language.\n\nThe release itself is straightforward and useful. Releasing everything under permissive licenses with 616 automated tests at 93% coverage shows basic engineering care. For anyone who needs to start working with Tajik text without building from scratch, the code and data lower the barrier.\n\nThe soft spot is exactly where the stress-test note points: the abstract says the unified morphology engine (controlled and deep modes) significantly improves handling of agglutinative inflections, yet the only validation mentioned is test coverage. There are no precision/recall figures on the POS corpus, no error analysis, and no comparison against prior Tajik stemmers or lemmatizers. Without those numbers the performance claim stays unsupported.\n\nThis is infrastructure work, not a methods discovery. It is aimed at researchers and developers who need practical Tajik support in low-resource Cyrillic settings. A reader already working on similar languages would find the datasets and modular design worth looking at.\n\nI would send it to peer review. The contribution is the release and data, not the unbacked performance assertion, so referees can focus on whether the components are actually usable.","headline":"TajikNLP ships the first open toolkit and four datasets for Tajik Cyrillic processing, but the 'significantly improves' claim on the morphology engine has no accuracy numbers or baselines to support it.","tokens_in":2412,"tokens_out":419,"would_cite":false,"duration_ms":15033,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"TajikNLP supplies the first open-source Python library for full processing of Tajik text in Cyrillic script.","keywords":["Tajik language","NLP toolkit","Cyrillic script","morphology engine","open-source library","POS tagging","sentiment analysis","low-resource languages"],"falsifier":"Run the morphology engine on a sample of Tajik sentences not used in development and compare its segmentations and analyses against independent human annotations to check whether the claimed handling of inflections holds.","tokens_in":2698,"feed_emoji":"🧰","tokens_out":510,"duration_ms":18030,"temperature":0.7,"pith_summary":"The paper introduces TajikNLP as an open-source toolkit that assembles a complete set of text-processing steps for Tajik while keeping the original Cyrillic orthography. The library centers on a single Doc object that chains cleaning, normalization, subword tokenization, morphemic segmentation, part-of-speech tagging, stemming, lemmatization, and sentence splitting. A new unified morphology engine supplies two analysis modes tuned to Tajik's agglutinative nominal and verbal forms. Four supporting datasets and pre-trained embeddings are released alongside the code, which is checked by 616 automated tests. A sympathetic reader would care because Tajik has lacked any comparable public infrastructure, so the toolkit removes the need to build these components from zero for research or applications.","feed_headline":"TajikNLP delivers first full NLP pipeline for Cyrillic Tajik text","feed_subtitle":"Library adds tokenization, morphology analysis, sentiment tools, and four datasets for this under-resourced language.","key_machinery":"The unified Doc object that chains the processing steps, anchored by the novel unified morphology engine that supplies controlled and deep analysis modes for Tajik agglutinative inflections.","core_discovery":"TajikNLP is an open-source Python library that provides the first comprehensive pipeline for processing authentic Tajik text while preserving the original Cyrillic orthography, built around a modular architecture with a unified Doc object and a novel unified morphology engine offering controlled and deep analysis modes that improve handling of agglutinative inflections, together with a lexicon-based sentiment analyser, pre-trained embeddings, and four openly published linguistic datasets.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["TajikNLP toolkit processes Tajik Cyrillic text","TajikNLP includes morphology engine for Tajik inflections","TajikNLP library adds sentiment tools and Tajik datasets","TajikNLP Python pipeline for tokenization and stemming","Modular Doc object enables TajikNLP Cyrillic processing"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The novel morphology engine actually improves handling of Tajik agglutinative inflections as described, an assumption resting on the internal test suite rather than external accuracy benchmarks.","fun_headline_variants_meta":{"raw":{"variants":["TajikNLP toolkit processes Tajik Cyrillic text","TajikNLP includes morphology engine for Tajik inflections","TajikNLP library adds sentiment tools and Tajik datasets","TajikNLP Python pipeline for tokenization and stemming","Modular Doc object enables TajikNLP Cyrillic processing"]},"model":"grok-4.3","cost_usd":0.006579,"raw_usage":{"total_tokens":3111,"prompt_tokens":744,"num_sources_used":0,"completion_tokens":71,"cost_in_usd_ticks":65787000,"prompt_tokens_details":{"text_tokens":744,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2296,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":744,"tokens_out":71,"duration_ms":17428,"temperature":1.0,"reasoning_tokens":2296,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T23:57:31.823399+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Run the morphology engine on a sample of Tajik sentences not used in development and compare its segmentations and analyses against independent human annotations to check whether the claimed handling of inflections holds.","supporting_citations":[],"review_version":2}