# Use OpenStreetMap labels for sentences to pronounce

**URL:** <https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220>\
**Category:** Common Voice\
**Tags:** sentence-collection\
**Created:** [October 12, 2021, 7:15pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220 "2021-10-12T19:15:03Z")\
**Posts on this page:** 11\
**Page:** 1

<div class="post-metadata">

**Author:** ![guerda](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/guerda/32/52065_2.png) [@guerda](https://discourse.mozilla.org/u/guerda)\
**Post date:** [October 12, 2021, 7:15pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/1 "2021-10-12T19:15:03Z")

</div>

OpenStreetMap offers a lot of labels in different languages, for points of interest, addresses, etc.

[https://wiki.openstreetmap.org/wiki/Names](https://wiki.openstreetmap.org/wiki/Names)

[https://wiki.openstreetmap.org/wiki/Multilingual\_names](https://wiki.openstreetmap.org/wiki/Multilingual_names)

Maybe it’s worthwhile to discuss if Common Voice could source those for pronunciation “sentences”. It might be positive for transcription of addresses or other POIs.

To be honest, I’m not sure if the license is compatible with Common Voice’s public domain policy. Maybe the labels itself are not a level where copyright applies, but IANAL.

Any thoughts welcome.

---

<div class="post-metadata">

**Author:** ![david-song](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/david-song/32/33437_2.png) [@david-song](https://discourse.mozilla.org/u/david-song)\
**Post date:** [October 14, 2021, 2:53am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/2 "2021-10-14T02:53:13Z")

</div>

Probably not compatible with cc0, but this data set is:

[https://geonames.nga.mil/gns/html/index.html](https://geonames.nga.mil/gns/html/index.html)

---

<div class="post-metadata">

**Author:** ![guerda](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/guerda/32/52065_2.png) [@guerda](https://discourse.mozilla.org/u/guerda)\
**Post date:** [October 14, 2021, 4:03am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/3 "2021-10-14T04:03:53Z")

</div>

Yeah, this is my concern, too. Not sure if just the labels alone meet the threshold of originality (Schöpfungshöhe in German law).

---

<div class="post-metadata">

**Author:** ![david-song](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/david-song/32/33437_2.png) [@david-song](https://discourse.mozilla.org/u/david-song)\
**Post date:** [October 14, 2021, 4:32am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/4 "2021-10-14T04:32:31Z")

</div>

Maybe but there’s database rights in Europe, so unlike the US, collections of unoriginal facts can be protected by copyright law.

---

<div class="post-metadata">

**Author:** ![bozden](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/bozden/32/51381_2.png) [@bozden](https://discourse.mozilla.org/u/bozden)\
**Post date:** [October 14, 2021, 1:01pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/5 "2021-10-14T13:01:27Z")

</div>

I’ve been informed multiple times in the past that synthetic sentences and non-conversational single words are not good for the corpus.

I also need to feed common proper names, city/location names into the corpus thou…

---

<div class="post-metadata">

**Author:** ![guerda](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/guerda/32/52065_2.png) [@guerda](https://discourse.mozilla.org/u/guerda)\
**Post date:** [October 14, 2021, 1:48pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/6 "2021-10-14T13:48:41Z")

</div>

I think it makes sense to only include complete sentences, not just names. I thought about navigation use cases, where place names might be helpful to add to the training set, in order to boost that specific use case.  
Of course if the DeepSpeech generalizes enough, including specific words would not be needed.

---

<div class="post-metadata">

**Author:** ![david-song](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/david-song/32/33437_2.png) [@david-song](https://discourse.mozilla.org/u/david-song)\
**Post date:** [October 15, 2021, 1:31am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/7 "2021-10-15T01:31:46Z")

</div>

Maybe pull the Wikipedia database and look for geonames records that have a Wikipedia page, extract sentences from those pages, and filter them by ones that contain the page title?

---

<div class="post-metadata">

**Author:** ![bozden](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/bozden/32/51381_2.png) [@bozden](https://discourse.mozilla.org/u/bozden)\
**Post date:** [October 15, 2021, 8:15am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/8 "2021-10-15T08:15:37Z")

</div>

We can not use Wikipedia data directly, as Wikipedia data is not CC-0. There is a legal agreement between Mozilla and Wikipedia, also some scripts and a process for this purpose, taking 3 sentences per page.  
Please search for related posts, I’m on mobile now.

---

<div class="post-metadata">

**Author:** ![david-song](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/david-song/32/33437_2.png) [@david-song](https://discourse.mozilla.org/u/david-song)\
**Post date:** [October 17, 2021, 7:10pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/9 "2021-10-17T19:10:42Z")

</div>

Yeah I meant as a starting point for the script. If we whitelist those pages and add a “sentences must include $TEXT” rule, where $TEXT is the place name, then it’s the same process under the same agreement, right?

---

<div class="post-metadata">

**Author:** ![irvin](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/irvin/32/59482_2.png) [@irvin](https://discourse.mozilla.org/u/irvin)\
**Post date:** [October 19, 2021, 9:30am UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/10 "2021-10-19T09:30:03Z")

</div>

wikidata is CC0, perhaps we can extract all the instances of [road](https://www.wikidata.org/wiki/Q34442) (such as [Ketagalan Boulevard](https://www.wikidata.org/wiki/Q697197)) into an CC0 database.

I believe this data can benefit common voice in some way such as some characters of road names appear less in daily dialogs but is essential when creating a general purpose stt model.

---

<div class="post-metadata">

**Author:** ![irvin](https://sea1.discourse-cdn.com/flex001/user_avatar/discourse.mozilla.org/irvin/32/59482_2.png) [@irvin](https://discourse.mozilla.org/u/irvin)\
**Post date:** [October 19, 2021, 12:58pm UTC](https://discourse.mozilla.org/t/use-openstreetmap-labels-for-sentences-to-pronounce/87220/11 "2021-10-19T12:58:11Z")

</div>

local contributor had help write a query for roads in wikidata, althought it’s for local region, you may take it as reference - [https://w.wiki/4E$t](https://w.wiki/4E%24t)