Microsoft

Language Understanding (LUIS)

Welcome to the LUIS API Forum

(Language Understanding Intelligent Service)

Categories

Documentation – Any ideas or suggestions for the API Reference or Documentation.

Language Support – Submit a request to have a particular language supported.

luis.ai WebApp – Ideas and suggestions for luis.ai and functionality used through the WebApp.

Programmatic API – Any ideas or feedback pertaining to features or enhancements to LUIS programmatic API.

Samples & SDK Request – Let us know if you would like to see a tutorial/sample or a SDK provided.

  • Hot ideas
  • Top ideas
  • New ideas
  • My feedback
  1. Increase child item upper limit for list entities

    List Entity is advertised as the optimal entity type to use in scenarios where specific terms need to be detected exactly with high accuracy.

    Within such scenarios, however, the number of terms under a given category can be extremely high. The current limit of 20,000 child items would be unable to handle such cases. Handling a set of fixed terms may be associated in usages where the margin of error must be minimal, and thus, using other types of entities cannot be considered as a workaround.
    As such, an increase in the upper limit is requested.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Service Limits  ·  Flag idea as inappropriate…  ·  Admin →
  2. Music domain deleted

    There was a Music pre-built domain which has intents like PlayMusic, Decrease/Increase Volume and SkipBack/SkipForward. But this week the entire pre-built domain is gone. When could you bring it back? Thanks!

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  Intents  ·  Flag idea as inappropriate…  ·  Admin →
  3. Numerical entity detected as Datetime by LUIS when using the French language

    Hi all,

    I notice that simple numerical entity such as 2, 3, 4 have detected as datetime entity , and not only as builtin.number, by LUIS when using the French language.

    This behaviour occurs with the API version 2.0, below a JSON answer returned by LUIS as example.

    Thanks

    {
    "query": "station 2",
    "topScoringIntent": {

    "intent": "Station by index",
    
    "score": 0.9843828

    },
    "intents": [

    {
    
    "intent": "Station by index",
    "score": 0.9843828
    },
    {
    "intent": "Station by code",
    "score": 0.01580255
    },
    {
    "intent": "Inline open Checklist question",
    "score": 0.00675197
    },
    {
    "intent": "Scroll down",
    "score": 0.006016401
    },
    {
    "intent": "Force answer",
    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  Language Support  ·  Flag idea as inappropriate…  ·  Admin →
  4. Some numbers are not detected by LUIS as number entity when using the Italian language

    Hi all,

    I would like to highlight some issues about numberical entity identification with the Italian language (API ver 2.0).

    Utterances containing numbers such as "uno" (one), 15, 4, sei (six), due (two), have not correctly handled by Microsoft Language Understanding.

    The output JSON provided by the cloud service for the examples listed above, does not contain entity (see entities segment below).

    Good behaviour example:

    {
    "query": "stazione due",
    "topScoringIntent": {

    "intent": "Station by index",
    
    "score": 0.9663627

    },
    "intents": [

    {
    
    "intent": "Station by index",
    "score": 0.9663627
    },
    {
    "intent": "Station by code",
    "score": 0.0247762073
    },
    {
    "intent": "Help",
    "score":
    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Language Support  ·  Flag idea as inappropriate…  ·  Admin →
  5. Utterances accents and punctuation marks validation

    I have been working side by side with Partners and Customers in LATAM and I have seen a lot of confusion using utterances in Spanish with the accents and punctuation marks such as the period, question mark, exclamation point, comma, semicolon, colon, dash, hyphen, parentheses, brackets, braces, apostrophe, quotation marks, and ellipsis.

    I have seen Alexa Developer Console prevents this kind of characters in some way it reduces the possible errors generated during the training steps.

    Is it possible to add a validation directly to the LUIS portal in the Utterances field to prevent and normalize the way training should…

    9 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  Utterances  ·  Flag idea as inappropriate…  ·  Admin →
  6. application not found with luis api version 3.0 preview

    0

    I was testing the new preview 3.0 version for LUIS on the test console available at "https://westus.dev.cognitive.microsoft.com/docs/services/luis-endpoint-api-v3-0-preview/operations/5cb0a9459a1fe8fa44c28dd8/console" .

    Despite I am using the same key and app id that correctly work for version 2.0 I get as response the error message "The application wasn't found. Please make sure that the application is published to this region".

    The ID for the application I inserted in the console form is published for westus region, the error message seems not truthful.

    Anybody else is having the same issue?

    Thanks

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  On-Premises Support  ·  Flag idea as inappropriate…  ·  Admin →
  7. Utterance Limit Reported When I have 0 Utterances in my Application

    BadArgument: The number of examples per application version cannot exceed the limit 15000. This is what I get when I try to add new utterances (examples) to an intent. The version of the app I am using has 0 utterances currently. I have repeatedly added and deleted intents, utterances and entities programmatically in the past, but I don't think that utterances I have deleted should count. Shouldn't I be allowed to add new utterances?

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  3 comments  ·  Utterances  ·  Flag idea as inappropriate…  ·  Admin →
  8. Batch Test Results - Overall doesn't total correctly

    I have 61 intents in a dataset in the batch test results. When I push run, in the right scroll panel with all of the intents (and entities listed) I see at the top "overview (44/61 utterances passed)" However if I view each intent individually, I see just 1 FP and 1 FN across all the intents. Therefore, the total should be much higher than 44, perhaps 59 I would have think about the exact total. This should be fixed or at least clarified to the user doing the testing how the 44 was arrived at since it seems to…

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  luis.ai WebApp  ·  Flag idea as inappropriate…  ·  Admin →
  9. Show which utterances were ignored due to missing intents

    "Note: 1 utterance(s) were ignored due to missing intents." I cannot find which intent Luis thinks is missing. It would save so much time if I could quickly display it. It could be something as simple as a stray type which takes forever to find.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  luis.ai WebApp  ·  Flag idea as inappropriate…  ·  Admin →
  10. Persistence of Batch Test Screen

    When I got to batch test my document sets, I see the results in the scatter graph, but when I come back say a day later it looks like I am forced to rerun the batch test. Is there a way to persist these scatter graphs? I have a similar request on this forum to export these results in text format.

    4 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  luis.ai WebApp  ·  Flag idea as inappropriate…  ·  Admin →
    Planned  ·  Luke Bayler responded

    Hello,

    We are currently working on some batch testing APIs, and are working on adding support to the UI as well. The results would be available as JSON.

    Thanks,
    Luke

  11. Speech on LUIS Dashboard

    I have subscribed to speech services and can integrate speech utterance results into a call to a LUIS app to predict the instance. My question is can I test speech with my LUIS app from the Dashboard? DialogFlow provides this capability and it would very useful here too.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  luis.ai WebApp  ·  Flag idea as inappropriate…  ·  Admin →
  12. Export Results from Batch Test Screen

    While it is great to have the scatter graph to visualize the errors, it is far more effect to have say a csv file where I can quickly view the same information. Can you please allow export of the test results (in text format) so that they can be analyzed faster?

    6 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  luis.ai WebApp  ·  Flag idea as inappropriate…  ·  Admin →
    Planned  ·  Luke Bayler responded

    Hello,

    We are currently working on some batch testing APIs, and are working on adding support to the UI as well. The results would be available as JSON.

    Thanks,
    Luke

  13. Github integration for revisions

    GitHub integration for both experimentation and security.


    1. Commit. We would like to take snapshots of our Luis apps and easily commit those changes into GitHub. Ideally this would be a simple matter of "export" followed by a "commit + push" and then all the necessary files associated with the app would be exported.


    2. Restore. We could also easily restore snapshots (of course). If there is a work around, would love to hear about it.


    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Programmatic API  ·  Flag idea as inappropriate…  ·  Admin →
  14. Phrase List Recommendations Dutch (nl-NL)

    Phrase List recommendations in Dutch would be really helpful to improve the performance of simple entities. There aren't a lot of prebuilt entities in Dutch. So you have to train a lot of them yourself and phrase list recommendations would help to improve them.

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Language Support  ·  Flag idea as inappropriate…  ·  Admin →
  15. include SmallTalk Intent in Pre-Built

    I would like to suggest LUIS should include the Pre-Built Intent for SmallTalk.

    4 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  1 comment  ·  Intents  ·  Flag idea as inappropriate…  ·  Admin →
  16. About new parttern in Chinese

    I could type new parttern in Chinese tomorrow. But today, I can't type it in Chinese.

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Language Support  ·  Flag idea as inappropriate…  ·  Admin →
  17. Provide Shareable Authoring Keys

    We deploy LUIS as part of a CI/CD pipeline for our application. Currently, this requires manually storing a developer's authoring key in our devops service, and using that key to create and manage our production LUIS apps.

    This is bad for two reasons:


    1. The developer whose key is used has the ability to modify or delete our production LUIS apps, bypassing our change control system.

    2. If this developer leaves, we'll need to use the authoring key of somebody else on the team, which will require creating new apps and potentially disrupting our service.

    It would be great if we got…

    3 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    Under Review  ·  0 comments  ·  Programmatic API  ·  Flag idea as inappropriate…  ·  Admin →
  18. Allow unlabelling of recognized entities in utterances.

    When an entity is recognized and labelled in an utterance, we should be able to unlabel it so that LUIS does not pass it through as a recognised entity.

    This is important as you might have a phrase that is an entity in one scenario, but is not in another scenario. Currently the entity is just recognised and passed on from LUIS in every scenario.

    As an example, see this Stack Overflow post: https://stackoverflow.com/questions/54226528/how-to-remove-luis-entity-marker-from-utterance

    2 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Entities  ·  Flag idea as inappropriate…  ·  Admin →
    Resolved  ·  Luke Bayler responded

    Hello,

    Prebuilt entities are exact matches from a list of regular expressions.

    If you are unable to label something because it is labelled, remove the prebuilt entity, add the label, and then add the prebuilt entity again.

    Thanks,
    Luke

  19. One docker container for dispatcher and related multiple luis apps

    Hi,
    You have introduced an amazing feature of having Luis as docker container. This is a differentiation compared to most of NLUs.
    However, is a a problematic to use if we have a concept of dispatcher - for example in our case we have 10 apps and 1 dispatcher, meaning we need to have 11 containers running.
    ideally to be able to wrap it just in 1 container and this 1 container will be accessing internet for transactions count, not all of them separately.

    5 votes
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    1 comment  ·  Programmatic API  ·  Flag idea as inappropriate…  ·  Admin →
  20. Improve datetimeV2 date range understanding

    When using the datetimeV2 entity, the phrase "from april to may" returns a date range of 4/01 - 5/01. However the phrase "from april through may" returns the same date range. It would make more sense if the latter returned 4/01 - 5/31

    1 vote
    Sign in
    (thinking…)
    Sign in with: Facebook Google
    Signed in as (Sign out)

    We’ll send you updates on this idea

    0 comments  ·  Entities  ·  Flag idea as inappropriate…  ·  Admin →
  • Don't see your idea?

Feedback and Knowledge Base