Introducing MELI: the Mandarin-English Language Interview Corpus
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-05-19 17:12 UTCgrok-4.3pith:YIBO5FN4record.jsonopen to challenge →
The pith
The MELI Corpus provides matched Mandarin and English speech from 51 bilingual speakers to enable acoustic comparisons linked to language attitudes.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The authors present the MELI Corpus as a resource of 29.8 hours of speech from 51 bilingual speakers, with ~14.7 hours in Mandarin and ~15.1 hours in English. Each speaker completed read sentence tasks and spontaneous interviews in both languages. All audio is recorded at 44.1 kHz stereo, transcribed, force-aligned at word and phone levels, and anonymized. The corpus is designed to support acoustic comparisons across speakers and languages while linking those measurements to the speakers' expressed language attitudes.
What carries the argument
The MELI Corpus, which integrates matched bilingual recordings with content on language attitudes and provides transcriptions and alignments.
If this is right
- Acoustic features can be compared for the same speaker in Mandarin versus English.
- Code-switching patterns can be examined in relation to attitudes.
- Both read and spontaneous styles are available for style-based comparisons.
- The data is released with metadata and map scans for further study.
- Quantitative and qualitative methods can be combined using the same speakers.
Where Pith is reading between the lines
- The corpus could reveal how language attitudes influence phonetic variation in bilinguals.
- It might be used to improve speech technology for code-switched Mandarin-English.
- Future studies could test if attitudes predict specific pronunciation traits.
- Connections to other bilingual corpora could be explored for cross-linguistic patterns.
Load-bearing premise
The 51 speakers and their sessions represent typical Mandarin-English bilingual speech and attitudes without significant biases from recruitment or self-reporting.
What would settle it
If the speakers' acoustic data shows no reliable links to their stated attitudes or if the sample is found to be skewed by recruitment methods, the corpus's utility for the intended comparisons would be reduced.
read the original abstract
We introduce the Mandarin-English Language Interview (MELI) Corpus, an open-source resource of 29.8 hours of speech from 51 Mandarin-English bilingual speakers. MELI combines matched sessions in Mandarin and English with two speaking styles: read sentences and spontaneous interviews about language varieties, standardness, and learning experiences. Audio was recorded at 44.1 kHz (16-bit, stereo). Interviews were fully transcribed, force-aligned at word and phone levels, and anonymized. Descriptively, the Mandarin component totals ~14.7 hours (mean duration 17.3 minutes) and the English component ~15.1 hours (mean duration 17.8 minutes). We report token/type statistics for each language and document code-switching patterns (frequent in Mandarin sessions; more limited in English sessions). The corpus design supports within-/cross-speaker, within/cross-language acoustic comparison and links acoustics to speakers' stated language attitudes, enabling both quantitative and qualitative analyses. The MELI Corpus will be released with transcriptions, alignments, metadata, scans of labelled maps and documentation under a CC BY-NC 4.0 license.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript introduces the MELI Corpus, an open-source resource of 29.8 hours of speech from 51 Mandarin-English bilingual speakers. It features matched sessions in Mandarin and English, each combining read sentences and spontaneous interviews on language varieties, standardness, and learning experiences. Audio is recorded at 44.1 kHz (16-bit, stereo), with full transcriptions, word- and phone-level force alignments, anonymization, and metadata on speakers' language attitudes. Descriptive statistics cover component durations (~14.7 h Mandarin, ~15.1 h English), token/type counts, and code-switching patterns (frequent in Mandarin sessions, more limited in English). The design is presented as enabling within-/cross-speaker and within-/cross-language acoustic comparisons as well as linkages between acoustics and stated attitudes for quantitative and qualitative analyses. The corpus will be released under CC BY-NC 4.0 with transcriptions, alignments, metadata, and documentation.
Significance. If released as described, the corpus fills a notable gap in publicly available Mandarin-English bilingual speech resources by providing matched-language sessions and attitude metadata. This structure directly supports controlled acoustic comparisons and sociolinguistic investigations that are difficult with existing unbalanced or single-language corpora. The open licensing and inclusion of alignments and scans of labelled maps strengthen its potential for reuse in phonetics, code-switching studies, and attitude-acoustic correlation research.
minor comments (3)
- [Abstract] Abstract: the mean session durations (17.3 min Mandarin, 17.8 min English) are reported without clarifying whether these are per-speaker averages or totals; adding this detail would improve clarity of the descriptive statistics.
- The manuscript would benefit from an explicit comparison table or paragraph situating MELI against existing bilingual corpora (e.g., in terms of language pair balance, style matching, and attitude metadata) to highlight its distinctive contributions.
- Ensure the full release includes the promised scans of labelled maps and documentation files; a brief appendix listing all released components with file names and formats would aid users.
Simulated Author's Rebuttal
We thank the referee for their positive review of the MELI Corpus manuscript and for recommending acceptance. The assessment accurately captures the corpus design, its matched-language sessions, attitude metadata, and potential for acoustic and sociolinguistic analyses.
Circularity Check
No significant circularity
full rationale
This is a descriptive corpus-introduction paper with no derivations, equations, fitted parameters, predictions, or uniqueness theorems. The central claims enumerate corpus features (matched Mandarin/English sessions, read+interview styles, transcriptions, alignments, attitude metadata) that directly enable the listed analyses by construction of the resource itself; no step reduces to a prior fitted quantity or self-citation chain. The 51-speaker sample is presented as the actual collected data rather than a modeled prediction, and representativeness is not required for the design claims to hold.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.