METHOD AND SYSTEM FOR AUTOMATIC DETECTING MORPHEMES IN A TASK CLASSIFICATION SYSTEM USING LATTICES
The invention concerns a method and system for detecting morphemes in a user's communication. The method may include recognizing a lattice of phone strings from the user's input communication, the lattice representing a distribution over the phone strings, and detecting morphemes in the user's input communication using the lattice. The morphemes may be acoustic and/or non-acoustic. The morphemes may represent any unit or sub-unit of communication including phones, diphones, phone-phrases, syllables, grammars, words, gestures, tablet strokes, body movements, mouse clicks, etc. The training speech may be verbal, non-verbal, a combination of verbal and non-verbal, or multimodal.
1 . A method of detecting morphemes using lattices in an automated task classification system which operates on one or more task objectives of a user, comprising:
recognizing a lattice of phone strings representing morphemes in the user's input communication, the lattice representing a distribution over the phone strings;
detecting morphemes in the user's input communication using the recognized lattice; and
making task-type classification decisions based on the detected morphemes in the user's input communication.
2 . The automated task classification method of claim 1 , wherein the morphemes include at least one of verbal speech and non-verbal speech.
3 . The automated task classification method of claim 2 , wherein the non-verbal speech includes the use of at least one of gestures, body movements, head movements, non-responses, text, keyboard entries, keypad entries, mouse clicks, DTMF codes, pointers, stylus, cable set-top box entries, graphical user interface entries and touchscreen entries.
4 . The automated task classification method of claim 1 , wherein the morphemes are expressed in multimodal form.
5 . The automated task classification method of claim 1 , wherein the user's input communication is derived from the verbal and non-verbal speech and the user's environment.
6 . The automated task classification method of claim 1 , wherein the morphemes in the user's input communication are derived from the user's actions, including the user's focus of attention.
7 . The automated task classification method of claim 1 , further comprising entering into a dialog with the user to obtain a feedback response from the user.
8 . The automated task classification method of claim 1 , wherein the user is prompted to provide a feedback response includes additional information with respect to the user's initial input communication.
9 . The automated task classification method of claim 1 , wherein the user is prompted to provide a feedback response that includes confirmation with respect to at least one of the set of task objectives determined in the classification decision.
10 . The automated task classification method of claim 1 , wherein the input communication is routed based on the classification decision.
11 . The automated task classification method of claim 1 , wherein the task objective is performed after the user's input communication is routed.
12 . The automated task classification method of claim 1 , wherein the method operates in conjunction with one or more communication networks, the one or more communication networks including at least one of a telephone network, the Internet, an intranet, Cable TV network, a local area network (LAN), and a wireless communication network.
13 . The automated task classification method of claim 1 , wherein the method is used for customer care purposes.
14 . The automated task classification method of claim 1 , wherein the classification decisions and corresponding user input communications are collected for automated learning purposes.
15 . The automated task classification method of claim 1 , wherein the relationship between the generated morphemes and the one or more task objectives includes a measure of usefulness of a one of the morphemes to a specified one of the predetermined task objectives.
16 . The automated task classification method of claim 15 , wherein the usefulness measure is a salience measure.
17 . The automated task classification method of claim 16 , wherein the salience measure is represented as a conditional probability of the task objective being requested given an appearance of the morpheme in the input communication, the conditional probability being a highest value in a distribution of the conditional probabilities over the set of predetermined task objectives.
18 . The automated task classification method of claim 16 , wherein each of the plurality of generated morphemes has a salience measure exceeding a predetermined threshold.
19 . The automated task classification method of claim 1 , wherein the relationship between the generated morphemes and the predetermined set of task objectives includes a measure of commonality within a language of the morphemes.
20 . The automated task classification method of claim 19 , wherein the commonality measure is a mutual information measure.
21 . The automated task classification method of claim 20 , wherein each of the plurality of generated morphemes has a mutual information measure exceeding a predetermined threshold.
22 . The automated task classification method of claim 1 , wherein the input communication from the user represents a request for at least one of the set of predetermined task objectives.
23 . The automated task classification method of claim 1 , wherein the input communication is responsive to a query of a form “How may I help you?”.
24 . The automated task classification method of claim 2 , wherein each of the verbal and non-verbal speech are directed to one of the one or more task objectives and each of the verbal and non-verbal speech is labeled with the one task objective to which it is directed.
25 . The automated task classification method of claim 1 , wherein the morphemes are generated by clustering selected ones of salient sub-morphemes from training speech which are semantically and syntactically similar, stored in a database which is used to detect morphemes in the recognized lattice.