Archboard feedback for Question Answering

[heaths]

• [x] Put all models in the root namespace.
• [x] Move projectName and deploymentName parameters to new input model named QuestionAnsweringProject that requires them both in the constructor. This should be the second required parameter after string question.
• [x] QueryKnowledgeBase() ->
  • AnswersResult GetAnswers(string question, QuestionAnsweringProject project, AnswersOptions options = null, …)
• [x] QueryText() ->
  • AnswersFromTextResult GetAnswersFromText(string question, IEnumerable<string> textDocuments, …)
  • AnswersFromTextResult GetAnswersFromText(string question, IEnumerable<TextDocument> textDocuments, …)
  • AnswersFromTextResult GetAnswersFromText(AnswersFromTextOptions options, …)
• [x] TextRecord -> TextDocument
• [x] AnswerSpanRequest -> ShortAnswerOptions
• [x] ShortAnswerOptions.Enable should be hidden and always set to true
• [x] ShortAnswerOptions.TopAnswersWithSpan -> Top
• [x] QueryKnowledgeBaseOptions -> AnswersOptions
• [x] AnswersOptions.AnswerSpanRequest -> ShortAnswerOptions (same as type name)
• [x] Keep AnswerSpan model but rename property to ShortAnswer
  • [x] KnowledgeBaseAnswer.AnswerSpan -> ShortAnswer
• [x] ConfidenceScore -> Confidence
• [x] ConfidenceScoreThreshold -> ConfidenceThreshold
• [x] KnowledgeBaseAnswerRequestContext.PreviousUserQuery -> KnowledgeBaseAnswerRequestContext.PreviousQuestion
• [x] Hide StringIndexType
• [x] RankerType -> RankerKind
• [x] KnowledgeBaseAnswer.Id -> KnowledgeBaseAnswer.QnaId for consistency
• [x] [C# only] Rename Top -> Size (at least 2 instances)
• [x] Stick with a string for language for now, per offline discussion with @rokulka and @tg-msft, but XML doc comments should link to https://docs.microsoft.com/azure/cognitive-services/qnamaker/overview/language-support anywhere we use language or DefaultLanguage.

Investigate:

• ~Use KeyValuePair<string, string> instead of MetadataRecord~
• ~Use NameValueCollection instead of IList<MetadataRecord> for MetadataFilter.Metadata~

After discussion since these are both input models and how they are used only for serializing, it's okay to leave as-is.

[heaths]

Updated with changes from subsequent discussions.

[heaths]

@annatisch went through the list of property name renames and came up with this for python:

AnswerSpan.confidence_score -> AnswerSpan.confidence
AnswerSpanRequest.confidence_score_threshold -> ShortAnswerOptions.confidence_threshold
AnswerSpanRequest.top_answers_with_span -> ShortAnswerOptions.top
KnowledgeBaseAnswer.answer_span -> KnowledgeBaseAnswer.short_answer
KnowledgeBaseAnswer.id -> KnowledgeBaseAnswer.qna_id
KnowledgeBaseAnswer.confidence_score -> KnowledgeBaseAnswer.confidence
KnowledgeBaseAnswerRequestContext.previous_user_query -> KnowledgeBaseAnswerContext.previous_question
QueryKnowledgeBaseOptions.confidence_score_threshold -> AnswersOptions.confidence_threshold
QueryKnowledgeBaseOptions.answer_span_request -> AnswersOptions.short_answer_options
QueryKnowledgeBaseOptions.ranker_type -> AnswersOptions.ranker_kind
QueryTextOptions.records -> AnswersFromTextOptions.text_documents
TextAnswer.confidence_score -> TextAnswer.confidence
TextAnswer.answer_span -> TextAnswer.short_answer

[heaths]

Per discussion with @KrzysztofCwalina and @tg-msft I'll just move all models to the root via transform. If we add authoring to the same package (likely, given it's much smaller than Search) we can add the client and models - with their own .Models namespace - to a separate authoring namespace like we did for Search.