A computing system for determining a user intent from a speech input to effect a user intended action, the computer system comprising:
- a set of processing nodes, wherein each processing node is capable of understanding only a subset of words directly relevant to a particular context associated with that processing node, the processing nodes of the set being arranged to receive a same speech input, wherein each processing node is configured to attempt to interpret the same speech input based on the subset of words directly relevant to its associated context, to extract therefrom an output indicative of user intent, whereby each processing node is unable to interpret any portion of the same speech input containing a word outside of its subset of words, whereby a portion of the same speech input relating to the particular context of a first of the processing nodes is interpretable to the first processing node but is not interpretable to a second of the processing nodes, wherein each processing node comprises:
- an automated speech recognition component configured to generate a speech recognition result from the same speech input based on the subset of words understood by that processing node;
- and
- an intent recognition component configured to generate the output indicative of user intent based on the speech recognition result, wherein the speech recognition output is restricted to only any portion or portions of the same speech input that the automated speech recognition component was able to interpret;
- and
- a controller configured to receive the outputs from the set of processing nodes and determine a most likely user intent based on the outputs;
- wherein the controller comprises a further processing node, wherein the further processing node is capable of understanding only a subset of words derived from the outputs indicative of user intent of each of the processing nodes of the set of processing nodes, the further processing node being arranged to receive the same speech input, wherein the further processing node is configured to:
- attempt to interpret the same speech input based on the subset of words derived from the outputs indicative of user intent of each of the processing nodes;
- and
- provide an indication of a most relevant output indicative of user intent, the most relevant being the output indicative of user intent which is determined to be most likely to represent the speech input;
- and
- wherein the subset of words of the further processing node are selected based on the speech recognition results of each of the processing nodes of the set of processing nodes, wherein the further processing node comprises an automated speech recognition component and an intent recognition component, wherein the indication of the most relevant output indicative of user intent generated by the further processing node comprises a comparable confidence score, the comparable confidence score denoting an extent to which the speech recognition output generated by the intent recognition component of the further processing node is relevant to the speech input.