---
title: "Call My Net 2: A New Resource for Speaker Recognition (2020)"
description: "This paper describes Call My Net 2 (CMN2), a corpus built by the Linguistic Data Consortium with a partner team in Tunis. 6 references."
authors: ["Karen Jones", "Stephanie Strassel", "Kevin Walker", "Jonathan Wright"]
year: 2020
venue: "Proceedings of the Twelfth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2020.lrec-1.816/"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/jones-2020-call-my-net-2-new/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/jones-2020-call-my-net-2-new/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2020](https://md.research.elamur.ai/free/en/papers/2020/page/8/) / Call My Net 2: A New Resource for Speaker Recognition

# Call My Net 2: A New Resource for Speaker Recognition

**Authors:** Karen Jones, [Stephanie Strassel](https://md.research.elamur.ai/free/en/authors/stephanie-strassel/), [Kevin Walker](https://md.research.elamur.ai/free/en/authors/kevin-walker/) and [Jonathan Wright](https://md.research.elamur.ai/free/en/authors/jonathan-wright/) · **Year:** 2020 · **Venue:** [Proceedings of the Twelfth Language Resources and Evaluation Conference](https://md.research.elamur.ai/free/en/venues/lrec/) ([LREC 2020](https://md.research.elamur.ai/free/en/venues/lrec/2020/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2020.lrec-1.816/) · [PDF](https://aclanthology.org/2020.lrec-1.816.pdf) · [OpenAlex](https://openalex.org/W3031484551)

**Citations:** 2 citations (OpenAlex, 2026-10-14)

## Summary
Speaker recognition systems need large, realistic evaluation data. Earlier LDC collections had not targeted Tunisian Arabic and had no VoIP calls. This paper describes Call My Net 2 (CMN2), a corpus built by the Linguistic Data Consortium with a partner team in Tunis. 408 native or highly fluent Tunisian Arabic speakers, called "claques", phoned friends and relatives about topics of their own choosing. They used mobiles, landlines, and Skype/Viber on computers or tablets, from both quiet and noisy places. Placing the recording platform in the same city as the speakers removed the connection failures seen in the first Call My Net. Every call passed automatic checks for duration and speech amount, then a two-stage human audit of signal quality, language and speaker identity. The resulting 4,562 calls supplied test data for NIST SRE18 and SRE19, and the full corpus is planned for the LDC Catalog. CMN2 is useful because it offers natural conversations between acquaintances across varied channels and acoustic conditions, in an Arabic dialect with little spoken data.

## Key points
- 4,562 calls from 408 speakers, recorded December 2016 to January 2018. Each speaker made at least 10 calls of up to 10 minutes. SRE18 used 2,306 recordings from 213 speakers, and the other speakers were held back for SRE19.
- Channel mix:
  - Devices: 4,341 mobile, 136 computer and 85 landline calls.
  - VoIP: 322 Skype and 826 Viber calls against 3,414 non-VoIP calls, meeting the 20–30% VoIP target.
- Noise: speakers were asked to make half their calls in noisy settings, and 46% of claque call sides were reported noisy (2,087 noisy vs 2,475 quiet calls).
- The platform was built in Tunis to LDC specifications and could record up to 15 two-party calls at once. It used a-law throughout, avoided the North American network, and had an LDC mirror for remote monitoring. A study website handled enrollment, call setup and live progress reports.
- A call counted as provisionally successful if it lasted at least 7 minutes and had at least 3 minutes of claque speech, as detected by the LDC HMM SAD. Delivery is single-channel 8-kHz a-law files with metadata for calls, sides, subjects and auditor judgements.

## Interesting ideas
- The audit uses two layers: each call is first judged on its own for clarity, speech amount and speaker count, then one native annotator compares all of a speaker's calls for identity and language consistency. This catches errors that no single-item check can see.
- An opening "reference segment" (15–40 s from the first minute, rich in personal greetings) serves as an anchor for confirming who is speaking later in the call.
- Instead of sampling at random, auditors split the rest of each call into thirds and take the densest 45-second speech stretch from each, so the sample covers the whole call while maximizing content.
- Cheap automatic gates (duration, SAD speech amount) decide which calls are worth sending to expensive human review. This saves annotator time for plausible candidates only.
- The setup form enforces dependent options (choosing VoIP removes landline from the menu, choosing non-VoIP removes computer), so contradictory metadata cannot be entered at all.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): Silver parses could be checked per sentence first, then for consistency across a whole document or across witness parsers, to expose systematic errors that per-sentence agreement hides.
- **I2** (frames): Coreference could treat the first, most descriptive mention of an entity as its reference profile and match later pronouns and short mentions against it.
- **I3** (learn): Evaluation samples from long texts could pick the densest sentence region in each third of the document, so spot checks cover beginning, middle and end efficiently.
- **I4** (learn): Run cheap structural validity checks on outputs before invoking LLM teacher tiers, so costly judgments are spent only on structurally sound candidates.
- **I5** (world): The text world could use the same constraint style, where asserting one property removes incompatible values from the ontology before a contradiction is ever recorded.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [University of Washington](https://md.research.elamur.ai/free/en/universities/university-of-washington/) — University, Seattle, United States · 263 works · 180,581 citations
- [University of Pennsylvania](https://md.research.elamur.ai/free/en/universities/university-of-pennsylvania/) — University, Philadelphia, United States · 170 works · 60,739 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 798 works · [💭 Experimental and Cognitive Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/experimental-and-cognitive-psychology/) · 463 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 1,484 works

## Cite

Cite as: Karen Jones, Stephanie Strassel, Kevin Walker and Jonathan Wright. 2020. Call My Net 2: A New Resource for Speaker Recognition. Proceedings of the Twelfth Language Resources and Evaluation Conference.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{jones2020call,
  title = {{Call My Net 2: A New Resource for Speaker Recognition}},
  author = {Karen Jones, Stephanie Strassel, Kevin Walker and Jonathan Wright},
  booktitle = {Proceedings of the Twelfth Language Resources and Evaluation Conference},
  year = {2020},
  url = {https://aclanthology.org/2020.lrec-1.816/}
}
```

- [← previous A Small Universal Dependencies Treebank for Hittite](https://md.research.elamur.ai/free/en/papers/andersen-2020-small-universal-dependencies-treebank-hittite/)
- [next → Clause Structure Variation in Biblical Hebrew: A Quantitative Approach](https://md.research.elamur.ai/free/en/papers/naaijer-2020-clause-structure-variation-biblical-hebrew/)
- [↑ back to papers of 2020](https://md.research.elamur.ai/free/en/papers/2020/page/8/)

## References

6 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. 14.1 Bibliographical References Jones, K., Strassel, S., Walker, K. and Wright, J. “Call My Net Corpus: A Multi-lingual Corpus for Evaluation of Speaker Recognition Technology,” in INTERSPEECH 2017: 18th Annual Conference of the International Speech Communication Association, August 20-24, Stockholm, Sweden, Proceedings, 2017, pp. 2621-2624. [find](https://scholar.google.com/scholar?q=14%2E1+Bibliographical+References+Jones%2C+K%2E%2C+Strassel%2C+S%2E%2C+Walker%2C+K%2E+and+Wright%2C+J%2E+%E2%80%9CCall+My+Net+Corpus%3A+A+Multi%2Dlingual+Corpus+for+Evaluation+of+Speaker+Recognition+Technology%2C%E2%80%9D+in+INTERSPEECH+2017%3A+1)
2. Masmoudi, A., Khmekhem, M.E., Yannick, E., Belguith, L.H. and Habash, N. (2014). A Corpus and Phonetic Dictionary for Tunisian Arabic Speech Recognition. In Nicoletta Calzolari (Conference Chair), et al., editors, Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 306-310, Reykjavik, Iceland, may. European Language Resource Association (ELRA). [find](https://scholar.google.com/scholar?q=Masmoudi%2C+A%2E%2C+Khmekhem%2C+M%2EE%2E%2C+Yannick%2C+E%2E%2C+Belguith%2C+L%2EH%2E+and+Habash%2C+N%2E+%282014%29%2E+A+Corpus+and+Phonetic+Dictionary+for+Tunisian+Arabic+Speech+Recognition%2E+In+Nicoletta+Calzolari+%28Conference+Chair%29%2C+et+)
3. NIST. (2018). “NIST 2018 Speaker Recognition Evaluation Plan,” [https://www.nist.gov/system/files/documents/2018/08/1](https://www.nist.gov/system/files/documents/2018/08/1) 7/sre18_eval_plan_2018-05-31_v6.pdf. [Online; accessed 20-Feb-2020]. [↗](https://www.nist.gov/system/files/documents/2018/08/1)
4. NIST. (2019). “NIST 2019 Speaker Recognition Evaluation: CTS Challenge,” [https://www.nist.gov/system/files/documents/2019/07/2](https://www.nist.gov/system/files/documents/2019/07/2) 2/2019_nist_speaker_recognition_challenge_v8.pdf, 2019 [Online; accessed 20-Feb-2020] [↗](https://www.nist.gov/system/files/documents/2019/07/2)
5. Sadjadi, S.O., Greenberg, C., Singer, E., Reynolds, D., Mason, L. and Hernandez-Cordero, J. (2019). The 2018 NIST Speaker Recognition Evaluation. In Interspeech 2019: 20 th Annual Conference of the International Speech Communication Association, September 15-19, Graz, Austria, Proceedings, 2019, pp. 1483-1487 [find](https://scholar.google.com/scholar?q=Sadjadi%2C+S%2EO%2E%2C+Greenberg%2C+C%2E%2C+Singer%2C+E%2E%2C+Reynolds%2C+D%2E%2C+Mason%2C+L%2E+and+Hernandez%2DCordero%2C+J%2E+%282019%29%2E+The+2018+NIST+Speaker+Recognition+Evaluation%2E+In+Interspeech+2019%3A+20+th+Annual+Conference+of+the+In)
6. Zribi, I., Ellouze, M., Belguith, L.H. and Blache, P. (2015). Spoken Tunisian Arabic Corpus “STAC”: Transcription and Annotation. In Research in Computing Science, pp.123-135 14.2 Language Resource References Ryant, N. (2013). “LDC HMM Speech Activity Detector (v1.0.5).” LDC, University of Pennsylvania [find](https://scholar.google.com/scholar?q=Zribi%2C+I%2E%2C+Ellouze%2C+M%2E%2C+Belguith%2C+L%2EH%2E+and+Blache%2C+P%2E+%282015%29%2E+Spoken+Tunisian+Arabic+Corpus+%E2%80%9CSTAC%E2%80%9D%3A+Transcription+and+Annotation%2E+In+Research+in+Computing+Science%2C+pp%2E123%2D135+14%2E2+Language+Resource+Re)

*Updated 24 October 2026*