KnowledgeAnswersIntentQueryFunctionCall

A list of arguments of this function call.

The corresponding meaning catalog version that was used to generate this FunctionCall.

Information about the corresponding Concept for this MRF Expression. For more details see go/concepts-roofshot.

Contextual Sensitivity (go/contextual-sensitivity) metadata indicating that a policy- or privacy- sensitive conversation context (previous queries, rewritten user queries, and previous system responses, like attentional entities) is used to generate this FunctionCall. This needs propagation (1) from Interp to DialogIntentState (DIS) (see b/148479837) and (2) from QRewrite/QUS down to Assistant Server's memory finalizer. One can use knowledge/answers/sensitivity/sensitivity_reader.h to parse this proto. NOTE(b/149091449): This is part of the migration of contextual sensitivity protos from Interp sensitivity extension. (nlp::semantic_parsing::sensitivity) to this proto field. See the bug to track the migration progress and for more details.

Contains data about which remodelings are being used for this funcall. For more information see go/meaning-remodeling-framework.

A list of token lists that were ignored during parsing because they are known context phrases for this interpretation. For example, for query [tell me how tall height of Obama really], assuming "tell me", "tall", "of", "really" are explained and thus ignored, with corresponding prior 0.9, 0.8, 0.7, 0.6, following ignored tokens will be populated: { ignored_tokens { token { ngram: "tell me", prior: 0.9} score: 0.9 } ignored_tokens { token { ngram: "tall", prior: 0.8} token { ngram: "of", prior: 0.7} score: 0.56 // currently score = prior1 prior2 ... * prior N } ignored_tokens { token { ngram: "really", prior: 0.6} score: 0.6 } } These tokens can serve two purpose: 1) debug info to show why an interpretation is generated; 2) carry the signals for downstream usage. Note: This field is under active development, and significant changes could happen. Please contact porky-pig@ if you want to use it.

The primary key for this FunctionCall. Note: This is still under development and not available for general use. Contact meaning-platform-eng@ for questions.

The marker specifies the purpose of this meaning struct / function call: Is it asking a question, and if yes, for which slot(s)? Is it a command, statement, etc? This corresponds to a (very coarse) notion of dialog acts. In the absence of this field, it will be inferred using the following algorithm (subject to marker applicability rules specified in marker.proto, see also go/requested-slots): 1) Use the underlying Meaning Schema's default marker, if applicable. 2) Assume that the Meaning Struct is polar or a statement. The presence of a marker will affect the value type of this function call.

A flattened representation of all intent modifiers that apply to this function call.

Name of this function call. The name must be present. If it is omitted, the FunctionCall is not well-formed.

Intent level query sensitivity (go/sensitive-intents). This metadata comes directly from Intent Catalog, indicating a single-shot query sensitivity without putting context into considerations. Therefore, this requires a one-to-one match with each intent registered in Intent Catalog.

Signals at the function call level

A list of tokens that were ignored during parsing that cannot be explained by context phrases.